PDA

View Full Version : A good entry level computer system?



Edz
06-26-2011, 06:42 AM
Hi, Ive been doing a bit of research & I would like advice on building a system that will give good work capacity with upgradable potential. Ive also had a look at some of the recommendations in this forum.
Ill build the system myself saving cash; and Im favouring the Intel DX58OS (http://ark.intel.com/Product.aspx?id=36888) mother board paired with a Xeon 5540 CPU (http://ark.intel.com/Product.aspx?spec=SLBF6).... Id start with about 12GB of DDR3 ram 1066MHz because the parity is ideally matched takes a bit of a hit in MHz compared to other CPU.

I like this motherboard because its reasonably priced even in my country; has 1366 socket and I can build up to 48 GB of RAM; has X58 north-bridge but the QPI speed of the CPU is slightly down compared to other CPUs. And the CPU is also reasonably priced, about as much as the Intel i7 compared below.

The processor can handle a maximum of 144 GB of memory which is miles above the Intel i7. So If Im understanding this correctly, the Xeon will be able to use all 48 GB of motherboards RAM to process, compared to only 24 GB by the Intel i7. Compare (http://ark.intel.com/Product.aspx?spec=SLBUZ)

The Xeon is a quad, able to calculate 8 threads which if I understand it correctly should speed up rendering and ray-tracing...

I think this setup will give great performance and upgrading potential for a really good price.

What other entry level configurations are other members using or recommending?

Perhaps some of the members will even recommend high end workstations?

Hieron
06-26-2011, 07:05 AM
Few points:

-Why go with a Q1 2009 Xeon processor. It's expensive, outdated and not fast. It costs 600 euro here or something bizarre.
-The motherboard is equally outdated and expensive.
-Why go with "pro" server equipment at all? If you are going 1 socket anyway.
-What are you going to need 48 GB ram for?

If I were you I would get:
-100-150 euro Asus Motherboard. Personally I'd go for the ~100 euro uATX board:ASUS P8P67-M B3 but perhaps normal ATX is better for OC'ing. Didn't have issues myself before though.
-250 euro Intel 2600K Sandy Bridge. It's new, fast, nice.
-16 GB, 4x4 GB memory, perhaps 200 euro total. If you truly run out of memory at 16GB, you can easily buy a insane new rig, or need to work on memory management :P Just make sure the memory configuration is on the QVL if possible.

So that brings it up to 600 euro for the heart of your machine. Some 128 GB SSD OS drive, a few TB's data drive, a case and a 560 GTX Ti to top it off or so. That's imho the very best cost/efficient way to build a machine now. Ow and ofc, do the easy thing, slap a decent cooler on it and overclock it to 4 - 4.5 ghz ofcourse. :) That will blow any Xeon like that clear out of the water for a much lower cost.

"High end" would be 980x 6 core or going to 2 sockets Xeon. The 980x 6 core or equivalent (got it myself) will be outdated in say 5 months when the new high level procs come out so I wouldn't purchase it now. A dual socket is always nice, but then you are looking at a serious cost increase and a lack of overclock options unless you go the EVGA SR-2 way.. but overclocking Xeons is a bit of a... dunno, perhaps try with something as easy as the 2600k. And that is *easy*.


ps: forget about "upgrade potential". I always considered it over the last 15 years and it never was the best option for me. Especially not when you start now, buying 2 years old equipment.

pps: not LW but informative anyway:
http://www.cpubenchmark.net/cpu_lookup.php?cpu=Intel+Xeon+E5540+%40+2.53GHz

See the 2600k high up on the top for it's measly 300$? See how the E5540 dangles loooooow down at perhaps 40% of it's speed. And this is at stock speeds.. the 2600k can and must be overclocked. That's what the k stands for :). It will push the E5540 down to a relative <33%. Keep in mind, the E5540 is way more expensive, old CPU's like that don't drop in price, they just get replaced.

Lightwolf
06-26-2011, 07:35 AM
What Hieron said...

I found that the only thing I upgrade over time might be RAM or HDD space.

In just about any other case it's more economical to a get a new box after 2-3 years (especially for CPU bound applications like rendering).

P.S. The new Sandy-Bridge CPUs do allow for 32GB in 4 memory sockets but the matching RAM modules aren't readily available yet (and will also be very expensive initially).

Cheers,
Mike

Red_Oddity
06-26-2011, 07:43 AM
What Hieron said indeed, we run dual socket Xeons at work because, well, we need the power it delivers (that might change though, seeing how power full the 2500/2600 Intels are)

At home i updated my workstation to a I5 2500K with a Gigabyte GA-P67A-UD3-B3 with 8GB Corsair and Crucial M4, it cost around 700 euros including VAT. The power and components controls that board give are insane (and this isn't even a real OC board), it has a dual bios/firmware, even runs OS-X.

And it is blazingly fast, even with the 2500K Intel.

Seriously, if you go single socket, don't bother with Xeons, it really isn't worth the extra cash it will cost you.

bazsa73
06-26-2011, 08:30 AM
I would say just get something decent but not the ubermegapro sith because eventually it is you what counts (who)

Edz
06-27-2011, 12:20 AM
Hi guys,

Thanks 4 the replies.

I agree bazsa..

---

Hieron,
good to know that your performance requirements were satisfied. This suggest good things to me regarding the setup I listed above... and it should be noted that all the xeon E55xx and xeon L55xx CPUs have a maximum bandwidth of 25.6 GB/s because of the QPI and the quad cores (So there are cheaper options @ about 300 euro.

Sure, the motherboard is 3 years old, and the CPU 2yrs, but I shall show you this isn’t necessarily a bad thing in real terms. I only want a system for handling LW 10... maybe sound also. The age isn’t a factor to me, but trusted execution technology would have made that mother board better.

For me, the upgradeability is actually about the memory and perhaps future 1366 socket CPUs that.

I looked over the sandybridge processor, as well as other potential boards with various chip sets. I agree with the set up you listed for what it is - the ATX version paired with the sandybridge CPU makes a lot of sense. But in my country, its a little expensive.

BUT the overclocking business IMO isn’t all its cracked up to be.

Lets look at the capacity of your set up and consider operations at the bandwidth maximum then beyond to the OC maximum . let m = max and c = oc max then I’ll assume a linear bandwidth B; scalar s = (c-m)/c = 1-m/c then the new bandwidth is (s+1)B =21+(4.6-3.8)21/4.6 =24.7GB/s

See... the CPU maximum bandwidth 21 GB/s calculates to a theoretical max increase in bandwidth of 24.7 GB/s, and don’t forget that’s assuming linearity - and disregarding the decrease in benefit due to heat (the exponential reality)

The extra memory is for the asymmetrical nature of QPI throughput, plus extra space to hold data not used in calculation when operating at maximum bandwidth... and potentially, there may be future CPUs that might need more memory to increase bandwidth capacity.

For the original core setup (of this post), the CPU operates in recommended parameters... no dangers of damage and potential data corruption due to overheating.

----
Red_Odity
You might want to hold off on the 2600 .... X79 chip set has been designed... and the bandwidth and new processors that use it might be worth the wait.

----
Lightwolf... isn't it strange that DDR RAM 1066MHz has the potential to out perform DDR RAM 2000+ MHz...

Edz
06-27-2011, 12:31 AM
Perhaps I should have made note that I used the 3.8 GHz over-clocked to 4.6 GHs to work out the linear benefit.

Lightwolf
06-27-2011, 12:33 AM
For me, the upgradeability is actually about the memory and perhaps future 1366 socket CPUs that.
Socket 1366 is dead. Sandy-Bridge E (out by the end of this year) will need a new socket.

I'd also forget any thoughts about bandwidth. Especially for applications like 3D rendering it's highly overrated. Current CPU caches manage to hide a lot of the latencies (which tends to make more of a difference than bandwidth anyhow).
Except for some very synthetic tests the difference between standard RAM clock speeds and the next higher bin is negligible, the same is true for dual vs. triple channel (in the case of socket 1366).
Power use may also be a factor of course.

I'd say, unless you have the cash for a (currently fastest when threading) socket 1366 based box, then socket 1155 and the i7-2600K is the one to get, hands down. To make it worse, the i7-2600K will also be faster on any task that doesn't multi-thread (which in the case of LW covers Modeler and most of the operations in Layout - with the grand exception of rendering).

Cheers,
Mike

Hieron
06-27-2011, 01:42 AM
edit: nvm. Good luck with the Xeon! Be sure to post your benchmark results! :p

ps: why did you ask for advice?

Red_Oddity
06-27-2011, 02:52 AM
...
Except for some very synthetic tests the difference between standard RAM clock speeds and the next higher bin is negligible, the same is true for dual vs. triple channel (in the case of socket 1366).
...

Cheers,
Mike

Very true, we tested various memory at various speeds, the speed gains the 1600 gave where negligible when compared to 1333 memory, same with the cash latencies, even though the prices go up at ridiculous amounts.

Edz
06-27-2011, 03:14 AM
edit: nvm. Good luck with the Xeon! Be sure to post your benchmark results! :p

ps: why did you ask for advice?:)

I'm not a computer expert, I only started researching a few days ago, and I only know what my research uncovers, and what others
tell me.

There are members in here that have years of real experience, and I have none.

I only want to buy parts for 1 machine that will last a good while, and I want to make comparison to other entry level machines so that I make the best informed decision I can.

I'm not sure how to do benchmarks

Lightwolf
06-27-2011, 03:18 AM
I only want to buy parts for 1 machine that will last a good while, and I want to make comparison to other entry level machines so that I make the best informed decision I can.
Well, you just got the same information from three different people, there's a clue :hey:

:beerchug:

Cheers,
Mike

JustForFun
06-27-2011, 11:41 AM
Notice this combo on newegg. Since I'm looking for a run of the mill replacement box, how would this fair?

http://www.newegg.com/Product/ComboBundleDetails.aspx?ItemList=Combo.675226

Lightwolf
06-27-2011, 11:56 AM
Notice this combo on newegg. Since I'm looking for a run of the mill replacement box, how would this fair?

http://www.newegg.com/Product/ComboBundleDetails.aspx?ItemList=Combo.675226
That looks solid enough. What's missing is a graphics board (currently I'd get something nVidia 560Ti based) unless you can re-use one you own already.
Oh, and the OS of course ;)
If you feel generous, get a SSD for the OS and programs that you use often (120GB should be plenty).

Cheers,
Mike

Hieron
06-27-2011, 01:30 PM
Notice this combo on newegg. Since I'm looking for a run of the mill replacement box, how would this fair?

http://www.newegg.com/Product/ComboBundleDetails.aspx?ItemList=Combo.675226


Looks good to me, and heartily agree with what Mike suggested. :)
That is alot of power in that box.. sure is the sweet spot now if you ask me. I've been really close to replacing the oldest parts of our renderfarm (Q6600's) for those 2600k's..

High end parts come at the end of the year, but I think it'll be a while before the 2600k's are replaced as the "smart" purchase.

ps: if you're planning on using a box like this as a rendernode, 500W power supply easily suffices. Seeing you are getting a Z-model mobo, you may.

Qexit
06-27-2011, 04:13 PM
I'm not sure how to do benchmarksWith Lightwave a good way to benchmark your system is to render a frame from one of the benchmark scenes included with the content you get as part of your purchase. As it happens, I visited a friend of mine yesterday who has just invested in a shiny new 2600K based PC. He wanted something to run on it to see how much difference it made when he overclocked it, so I went round with my copy of LW 10 to find out. I just picked out the first scene: HDRI_Spheres. On Windows 7 Home 32-bit and 4GB of RAM, I got the following results:

2600K @ basic (3.5GHz turbo) = 123.4 secs
2600K @ 45x (4.5GHz turbo) = 104.3 secs

Back home on my i7 860 PC with Win 7 Pro 64-bit and LW10 64-bit:

i7 860 @ basic (3.4 GHz turbo) = 195.2 secs

Both cpus run 8 threads off their four cores. It's fun seeing all 8 cpus maxed out at 100% in Task Master :D

With a Xeon 5540 maxing out at only 2.8GHz it would be beaten quite convincingly even by my i7 860. Bottom line for me is, I want one of those 2600K systems. Very affordable and very cost-effective. If it's not fast enough for you, buy two and use the other one as a render box :thumbsup:

michaeldejong
06-27-2011, 04:20 PM
My favorite part of this discussion is the opinion of what qualifies as an 'entry-level' machine lol.

:lwicon:

Lightwolf
06-27-2011, 04:35 PM
My favorite part of this discussion is the opinion of what qualifies as an 'entry-level' machine lol.

True, it's more like best bang for the buck...

Cheers,
Mike

Qexit
06-27-2011, 04:45 PM
True, it's more like best bang for the buck...

Cheers,
Mike..which sounds a lot like my definition of an entry level machine :D

Lightwolf
06-27-2011, 04:46 PM
..which sounds a lot like my definition of an entry level machine :D
I assume I buy one entry level machine after the other then... :D

Cheers,
Mike

Red_Oddity
06-27-2011, 04:53 PM
My favorite part of this discussion is the opinion of what qualifies as an 'entry-level' machine lol.

:lwicon:

Entry level PC for a 3D artist, not an entry level PC for your mom to check cooking recipes from the internet on.

Even 'entry level' is completely relative.

michaeldejong
06-27-2011, 08:05 PM
Entry level PC for a 3D artist, not an entry level PC for your mom to check cooking recipes from the internet on.

Even 'entry level' is completely relative.


Yes, well put for sure. What I used a few years ago wouldnt hold up to what I put my new machine through. Its great to read these posts to get the experts thoughts on what hardware is the best. I'll likely be posting something simular to this thread within a year.

Ernest
06-27-2011, 11:11 PM
Entry level PC for a 3D artist, not an entry level PC for your mom to check cooking recipes from the internet on.

Even 'entry level' is completely relative.But still!

If we're calling a Xeon system with 12GB of RAM an entry level system, then what do we need in order to call something high end? Something like THIS? (http://www.thinkmate.com/System/SuperServer_5086B-TRF)

michaeldejong
06-27-2011, 11:37 PM
But still!

If we're calling a Xeon system with 12GB of RAM an entry level system, then what do we need in order to call something high end? Something like THIS? (http://www.thinkmate.com/System/SuperServer_5086B-TRF)

That was sort of my point.

My first comment was from feeling that the computers mentioned above are much 'higher-end' than the machine I use. But also if you had given me the computer I have now (six cores), 3 years ago when I first bought into Lightwave - I'd probably say it was a supercomputer lol. But now with what I am learning and creating, the old computer (two cores) seems below an 'entry-level' computer & is something I would give to my mom for surfing the internet lol .. But at that time it was what I could afford and it served its purpose.

Like it was said, its very time relative too ..

:lwicon:

Red_Oddity
06-28-2011, 03:18 AM
But still!

If we're calling a Xeon system with 12GB of RAM an entry level system, then what do we need in order to call something high end? Something like THIS? (http://www.thinkmate.com/System/SuperServer_5086B-TRF)

Well, yes, or maybe even mid-end, we have a bunch of Supermicro SYS-6016T-MR doing the brunch of our number crunching, a heavy investment, but these puppies run 24/7.
When you do renders and conversion of 2000-6000 frames per shot/scene, you'll be glad you have a bunch of these.

(for example, my last scene on the farm was over 8000 frames, took 40 minutes to render, as opposed to the 12 to 14 hours it takes on a dual Xeon quad system)

Hieron
06-28-2011, 07:19 AM
But still!

If we're calling a Xeon system with 12GB of RAM an entry level system, then what do we need in order to call something high end? Something like THIS? (http://www.thinkmate.com/System/SuperServer_5086B-TRF)

A Xeon and 12 GB Ram says nothing really.. 12 GB is very cheap these days and Xeons come in slow not so sexy forms as well. Such as the one posted above.

The suggested 2600k systems is fast but still quite cheap, it sure is in the sweet spot for many people. Heck, it will even beat quite some sexy dual Xeon setups out there.


Well, yes, or maybe even mid-end, we have a bunch of Supermicro SYS-6016T-MR doing the brunch of our number crunching, a heavy investment, but these puppies run 24/7.
When you do renders and conversion of 2000-6000 frames per shot/scene, you'll be glad you have a bunch of these.

(for example, my last scene on the farm was over 8000 frames, took 40 minutes to render, as opposed to the 12 to 14 hours it takes on a dual Xeon quad system)

Seeing the great work you guys are doing, I'm sure the purchase cost of a decent renderfarm is easily justified. Nice! Wouldn't mind a rack of renderboxxes myself.. :) Does LW play a big factor in your pipeline?

Red_Oddity
06-28-2011, 10:10 AM
Does LW play a big factor in your pipeline?
Less than i like (almost everything we do is largely Maya and range of other software), a couple of one man show commercials are LW only, and most jobs have the occasional model passed through Modeler for cleanup and poly alignment (Maya is notoriously bad at mesh cleanup, and some MR shaders are very dependent on the poly normal direction being correct.)

Lately we've been testing VRay 2 on some gigs, and we like it very very much (a lot more than MR so far)

At the moment we're writing a complete new pipeline that should be able to easily transfer assets and shots from program to program (Lightwave included)

Where are you situated in the Netherlands Hieron? Any shop i know?

JustForFun
06-28-2011, 01:21 PM
The suggested 2600k systems is fast but still quite cheap, it sure is in the sweet spot for many people.

Looks like that system could be built for $8-900. With a budget of $1.5-1.8k in mind (my current budget), is there a system that would give an appropriate increase in performance?

Lightwolf
06-28-2011, 01:24 PM
Looks like that system could be built for $8-900. With a budget of $1.5-1.8k in mind (my current budget), is there a system that would give an appropriate increase in performance?
Nope. You can get a six core 1366 based system, which would give you a better rendering performance but worse single threaded one - coupled with a higher power usage.
At least until Q4 of this year (at the earliest).
I'd rather get a SSD as well.

Cheers,
Mike

JustForFun
06-28-2011, 01:29 PM
I'd rather get a SSD as well.

I like the idea of an SSD for apps. Is it true the type of controler can either make them shine, or drag em down?

Lightwolf
06-28-2011, 01:31 PM
I like the idea of an SSD for apps. Is it true the type of controler can either make them shine, or drag em down?
Some of the new SSDs perform beyond SATA II - but I suspect it doesn't make that much of a practical difference.
Last time I checked there were some niggles with the SATA III ports on AMD based system, but those may be teething problems.

SATA III SSDs really aren't that much more expensive either.

Cheers,
Mike

Hieron
06-28-2011, 02:02 PM
Looks like that system could be built for $8-900. With a budget of $1.5-1.8k in mind (my current budget), is there a system that would give an appropriate increase in performance?

Not really, yet (as Mike said, Q4 brings a new enthusiast proc). Which is quite an interesting thing.. 800$ gets you really high on the cost/price curve.



Where are you situated in the Netherlands Hieron? Any shop i know?

Thanks, very interesting. LW still plays a major role for us, but we're mixing in Maya and Modo as well so it's becoming a mixed pipeline as well. Modo -> LW is doable though ofcourse. (despite layers getting messed up etc)

We're located in Enschede, 2 of the 3 of us studied at the university there. The link is under my nick here but we're: http://www.Nymus3D.nl

Ernest
06-28-2011, 07:27 PM
We're located in Enschede, 2 of the 3 of us studied at the university there.

Go Twente!
(Sorry, couldn't resist.)

Edz
06-29-2011, 04:35 AM
With Lightwave a good way to benchmark your system is to render a frame from one of the benchmark scenes included with the content you get as part of your purchase. As it happens, I visited a friend of mine yesterday who has just invested in a shiny new 2600K based PC. He wanted something to run on it to see how much difference it made when he overclocked it, so I went round with my copy of LW 10 to find out. I just picked out the first scene: HDRI_Spheres. On Windows 7 Home 32-bit and 4GB of RAM, I got the following results:

2600K @ basic (3.5GHz turbo) = 123.4 secs
2600K @ 45x (4.5GHz turbo) = 104.3 secs

Back home on my i7 860 PC with Win 7 Pro 64-bit and LW10 64-bit:

i7 860 @ basic (3.4 GHz turbo) = 195.2 secs

Both cpus run 8 threads off their four cores. It's fun seeing all 8 cpus maxed out at 100% in Task Master :D

With a Xeon 5540 maxing out at only 2.8GHz it would be beaten quite convincingly even by my i7 860. Bottom line for me is, I want one of those 2600K systems. Very affordable and very cost-effective. If it's not fast enough for you, buy two and use the other one as a render box :thumbsup:

:thumbsup: Thanks, this was really helpful. And it lead me to more research. I'm not going with the older options now.

I might do the sandybridge thing.

another ideas is a x58 chip set asus board with a core i7 960 3.2 GHz QPI rated to 4.2 GT/s and 25.6GB/s Bandwidth max, 12 DDR3 2000+ MHz all up @ about $700 to $900 US more or less.

According to QPI calculations, a single processor @ 3.2 GHz processes at a bandwidth of 25.6 GB/s... thats reaches the supremum (the worst time) for sequential potential (Amdahls law) given full use of the maximum bandwidth, and there's still 7 other threads for parallel calculations. so in theory, the entire bandwith is always utilized and should out perform the sandy bridge rated at 21GB/s

And there's always OC...

---

I read a test where 2 xeon 5680 were compared with and i7 980x, it tested lightwave-01 with 40.01 and 41.88 respectively.

but when it came to testing 1 million polygons

there was a clear advantage to having 24 threads. by as much as 4 mins

Edz
06-29-2011, 04:41 AM
Well, yes, or maybe even mid-end, we have a bunch of Supermicro SYS-6016T-MR doing the brunch of our number crunching, a heavy investment, but these puppies run 24/7.
When you do renders and conversion of 2000-6000 frames per shot/scene, you'll be glad you have a bunch of these.

(for example, my last scene on the farm was over 8000 frames, took 40 minutes to render, as opposed to the 12 to 14 hours it takes on a dual Xeon quad system)

Thanks Red... thats good to know.

But I suppose your calculating along 24 threads then?

Lightwolf
06-29-2011, 04:51 AM
According to QPI calculations, a single processor @ 3.2 GHz processes at a bandwidth of 25.6 GB/s... thats reaches the supremum (the worst time) for sequential potential (Amdahls law) given full use of the maximum bandwidth, and there's still 7 other threads for parallel calculations. so in theory, the entire bandwith is always utilized and should out perform the sandy bridge rated at 21GB/s

As mentioned before, those calculations are completely (to 99%) irrelevant today (unless you use niche applications). Forget about bandwidth, honestly. L3 and L2 caches nowadays keep cores quite busy.

CPI is more important (Cycles per instruction) but even then only in conjunction with the clock speed. This is also where the Sandy-Bridge outshines the predecessor ... clock for clock it's roughly 20-25% faster. Which also explains why the single threaded performance is so much better (plus some optimisations in 64-bit mode that the previous gen didn't have - micro-op fusion to be precise).

Cheers,
Mike - who can chuck around buzzwords if he needs to.

Hieron
06-29-2011, 06:10 AM
:thumbsup: Thanks, this was really helpful. And it lead me to more research. I'm not going with the older options now.

I might do the sandybridge thing.

A wise choice I think.



:another ideas is a x58 chip set asus board with a core i7 960 3.2 GHz QPI rated to 4.2 GT/s and 25.6GB/s Bandwidth max, 12 DDR3 2000+ MHz all up @ about $700 to $900 US more or less.

According to QPI calculations,


:) Honestly, you can't derive the result you'll see in LW that way. It just doesn't matter, like Mike said. The caches take care of it all.. You can't calculate it that easily from a single spec. Just check CPU multithreaded benchmarks around the net, the general result is always the same and a very good indication of performance in LW.

Personally I use Wprime alot to compare across machines and my OC's.


I read a test where 2 xeon 5680 were compared with and i7 980x, it tested lightwave-01 with 40.01 and 41.88 respectively.

but when it came to testing 1 million polygons

there was a clear advantage to having 24 threads. by as much as 4 mins

Ow seriously no doubt about that!
You are comparing 2 VERY similar processors here, esp. when not overclocked. If one uses 2x a Xeon X5680 it will surely beat a single 980X. Hands down.

The single 980X is 780 euro here, 2x a X5680 is 2700 euro here. So, by all means, if you have 3500 to spare for a full machine, go for 2x X5680. It will drop in value like a rock though.

If you compare that 3500 to the 700 a 2600K may cost, the xeons system is 500% of the price. But may only be 150% of the performance vs a 4.4 Ghz clocked 2600K. In single threaded parts of LW it will be 75% of the performance of the OC'ed 2600k. That's expensive core's sitting around looking pretty...

The question you then need to ask, is if you are willing to spend 5x more money to get a (if lucky) increase of 50% in speed.

The wise choice is often to get this "sweet spot" 2600k and put the extra cash in: better peripherals, 4 rendernodes! (giving you 3x more total speed for the same price, in network rendering). etc etc All numbers a bit rough btw, you get the idea.

Anyway, I should be working. Good luck with the choice!
Remember: spending 700-800 on a 2600K system is NEVER a bad choice and it will outperform most systems people here on the forums use LW with.

Red_Oddity
06-29-2011, 03:42 PM
Thanks Red... thats good to know.

But I suppose your calculating along 24 threads then?

LW 24 threads, MR 12 (MR doesn't seem to like hyper threading), Fusion often just 1 as it is rather buggy when it comes to multi threading and multi frame rendering.
VRay i don't know, probably 12 as well.

JustForFun
06-30-2011, 07:16 PM
That looks solid enough. What's missing is a graphics board (currently I'd get something nVidia 560Ti based)...

Your opinion on this video card please :) :

http://www.newegg.com/Product/Product.aspx?Item=N82E16814133384&cm_sp=Cat_Video_Cards_%26%2338%3b_Video_Devices-_-Clearance-_-14-133-384

I'm ordering that combo and wondering if this card would fit the bill.

Lightwolf
07-01-2011, 01:57 AM
Your opinion on this video card please :) :

http://www.newegg.com/Product/Product.aspx?Item=N82E16814133384&cm_sp=Cat_Video_Cards_%26%2338%3b_Video_Devices-_-Clearance-_-14-133-384

I'm ordering that combo and wondering if this card would fit the bill.
Yup, it looks good to me. Afaik PNY essentially use nVidia's reference design, which is quite decent for the 560Ti.

Cheers,
Mike

Edz
07-01-2011, 09:15 PM
:D Thanks for the reply Lightwolf & U2 Hieron... once again this was very helpful, and lead me to investigate the claims.
] As mentioned before, those calculations are completely (to 99%) irrelevant today (unless you use niche applications).This is true, but perhaps it is better understood in the context of a normal user; the context of LW 10 though needs to be clarified more. we are talking LW 10: which is niche as. The normal user would definitely struggle to fill even a quarter of the bandwidth.


CPI is more important (Cycles per instruction) but even then only in conjunction with the clock speed. This is also where the Sandy-Bridge outshines the predecessor ... clock for clock it's roughly 20-25% faster. I read that the overall performance improvements were 17% better than the Lynnfield (what ever that means). And don’t forget, even parallel instruction via just in time command scheduling, command overlap, and out of order command scheduling is limited to bandwidth regardless of how fast the processor is.


Which also explains why the single threaded performance is so much better.again this must be the case.

The improvements are great for the average user, and the sandybridge couple with the P67 chipset, and the appropriate memory configuration would make an excellent gaming computer.

I wouldn’t chose the sandybridge for single-quad-processor higher volume type rendering, and the option of x58 asus combined with an core i7 3.2GHz would more than likely best it because of the bandwidth –and there are configurations to optimise the overall performance (selecting the right i7 for the right MB). Getting the sandybridge is trading off the rendering volume for highly sequential operation, and general overall better handling ability.

In terms of GENERAL overall value for money, I would recommend the sandybridge. But given VPR, and if you’re considering making a do-it-yourself home movie on the cheap, or some kind of high volume rendering, the x58 asus might be better. it really needs comparative benchmarks... and both systems should be optimised

Looking to the future; Plans are a pairing of the ivybridge with x79... WHAT! The best of both worlds...:lwicon:

Edz
07-01-2011, 09:29 PM
LW 24 threads. an 8 core option is available by special order... but I didn't find any details on it


MR doesn't seem to like hyper threading), Fusion often just 1 as it is rather buggy when it comes to multi threading and multi frame rendering.
VRay i don't know, probably 12 as well. assuming gulftown... probably chipset related.

I was wondering tho, when you render for 40 minutes or more, what raid level do you operate? assuming you have freedom to set it? I'm curious to know if serverboards rely of Raid for guaranteed completion?

Lightwolf
07-02-2011, 04:30 AM
The normal user would definitely struggle to fill even a quarter of the bandwidth.
Let me clarify then: In 3D rendering and 3D processing memory bandwidth is irrelevant.

When I said niche I meant proper niche, such as scientific computing - I.e. multiplying matrices that are GBs in size with each other.

Just about everything else is very effectively hidden by the L3-L1 caches, which is why the difference due to faster memory (either by latency or bandwidth) is marginal at best.
If bandwidth was really that limiting then changing from, say, DDR3-1333 to DDR3-1600 would make a massive difference. While in real life the difference is hard enough to reliably benchmark (2-3% or so, which is almost within the margin of error).

And no, the socket 1366 doesn't best it because of the bandwidth but because it's got two more cores. Last gen, lower CPI cores but still two more. To make it worse, socket 1366 actually shows how little difference to rendering times there is if you cripple the bandwidth by not using a triple channel config.

Mind you, I think that overcklocked 4.7GHz beast here in the thread could get very close to the last gen 6-core i7.

Cheers,
Mike

Red_Oddity
07-02-2011, 07:01 PM
an 8 core option is available by special order... but I didn't find any details on it

assuming gulftown... probably chipset related.

I was wondering tho, when you render for 40 minutes or more, what raid level do you operate? assuming you have freedom to set it? I'm curious to know if serverboards rely of Raid for guaranteed completion?

We are running a RAID 5 server with 8 SATA II drives on an Areca card, we plan on updating to a new Areca with either 12 or 24 drives and 2GB cache on the RAID card (probably going to be RAID 6), so far the RAID performs well (around 400-500 MB/s), but with these huge file sequences performance degrades fast (browsing those folders is dreadful to say the least), but just not enough to impact rendering that heavily.

I'm also asking around what OS i should run the new server at (Windows server has some rather poor file perfomance for what we do i think), we're looking at CentOS or Suse, also the file system is still being debated.

It's really hard to find good info on things like this online, most reviews or forum discussions i find always tend to end the same way Mac vs PC discussion threads end.

The thing is, we are too small a company and have no loans, and are not willing to risk loans to go get some dedicated company to do these things for us and thus pay everything from out own pocket and build the machines mostly ourselves, so we are somewhat tied to what we can find on the market as component for DIY servers sort to speak (our DIY also includes SuperMicro, so not THAT DIY)
So, no DVS-SAN systems with 6 figure price tags (i wish)

Lightwolf
07-02-2011, 07:07 PM
...but with these huge file sequences performance degrades fast (browsing those folders is dreadful to say the least)...
If it's a windows based system running NTFS (what else) there's one or two registry tweaks that can heavily speed up browsing directories that contain a few thousand files.
Essentially, by default, windows writes a last access time per file even if you just browse it (which make sense technically - but slows things down).

Cheers,
Mike

Edz
07-03-2011, 04:14 PM
Let me clarify then: In 3D rendering and 3D processing memory bandwidth is irrelevant.

When I said niche I meant proper niche, such as scientific computing - I.e. multiplying matrices that are GBs in size with each other.

Why would you even say this? Even Intel market their serverboards and workstations for use in graphics, and list graphics as a niche application. Everyone here would rather have a multi CPU motherboards, to handle the heavy workloads.

A large portion of computational mathematics was refined just for the sake of graphics. The large rendering times bare testiment to the calculations.

LW 10 is niches as!

And its incorrect to say bandwidth is irrelevant and I'll explain why.


Just about everything else is very effectively hidden by the L3-L1 caches, which is why the difference due to faster memory (either by latency or bandwidth) is marginal at best.
If bandwidth was really that limiting then changing from, say, DDR3-1333 to DDR3-1600 would make a massive difference. While in real life the difference is hard enough to reliably benchmark (2-3% or so, which is almost within the margin of error).

Ok, I'll explain why this is conceptually incorrect. the memory speed is more about the speed to line up. but that line up must then be accelerated by a clock so that it can interface with the CPU. it is at interfacing time that throughput and potential bottlenecks occurs. and the bandwidth is an indication of the throughput capacity.


To make it worse, socket 1366 actually shows how little difference to rendering times there is if you cripple the bandwidth by not using a triple channel config. I can only speculate about this, but I do see instances where configs aren't going to impact too greatly.


Mind you, I think that overcklocked 4.7GHz beast here in the thread could get very close to the last gen 6-core i7.

Cheers,
Mike
OCing works because it accelerates all relevant clocks.

Cheers
Edz

Edz
07-03-2011, 04:33 PM
We are running a RAID 5 server with 8 SATA II drives on an Areca card, we plan on updating to a new Areca with either 12 or 24 drives and 2GB cache on the RAID card (probably going to be RAID 6), so far the RAID performs well (around 400-500 MB/s), but with these huge file sequences performance degrades fast (browsing those folders is dreadful to say the least), but just not enough to impact rendering that heavily.

I'm also asking around what OS i should run the new server at (Windows server has some rather poor file perfomance for what we do i think), we're looking at CentOS or Suse, also the file system is still being debated.

It's really hard to find good info on things like this online, most reviews or forum discussions i find always tend to end the same way Mac vs PC discussion threads end.

The thing is, we are too small a company and have no loans, and are not willing to risk loans to go get some dedicated company to do these things for us and thus pay everything from out own pocket and build the machines mostly ourselves, so we are somewhat tied to what we can find on the market as component for DIY servers sort to speak (our DIY also includes SuperMicro, so not THAT DIY)
So, no DVS-SAN systems with 6 figure price tags (i wish)

Yes, my research over last week included OS and in particular how they handled Raid. There are a tonne of programs out there, but I haven't as yet investigated any in great detail. I can't do anything that specific yet.

A huge list of OS exists @ wikipedia. Good luck with that.

4 now, I'd be happy with Windows and just getting started

Lightwolf
07-03-2011, 04:47 PM
Why would you even say this? Even Intel market their serverboards and workstations for use in graphics, and list graphics as a niche application. Everyone here would rather have a multi CPU motherboards, to handle the heavy workloads.
Of course we all would... just because we get more cores.
However, often it's not efficient in times of price/performance unless you either have very expensive licenses or users severely slowed down by the machine they use.
And obviously, DCC is a niche for them, just as HPC is (which is what I was referring to earlier on). I'd still suspect that theres more DCC users out there than HPC users (but likely more cores used for HPC computing ;) ).


A large portion of computational mathematics was refined just for the sake of graphics. The large rendering times bare testiment to the calculations.
Yup, but core rendering algorithms have already had plenty of cache space to fit into years ago. Which leaves out data access and that isn't nearly as important in terms of raw memory bandwidth. Cache alignment can be, but that's a software problem.


And its incorrect to say bandwidth is irrelevant and I'll explain why.

Ok, I'll explain why this is conceptually incorrect. the memory speed is more about the speed to line up. but that line up must then be accelerated by a clock so that it can interface with the CPU. it is at interfacing time that throughput and potential bottlenecks occurs. and the bandwidth is an indication of the throughput capacity.

I can only speculate about this, but I do see instances where configs aren't going to impact too greatly.

The speed to line up would be latency, no bandwidth. And even that has a negligible impact on current CPUs.
Bottlenecks mainly occur within the multi-staged cores, which is why Intel added Hyperthreading (making more efficient use of a single core by running two threads through it to fill up gaps in the pipeline) and AMD chose the module design for Bulldozer (two integer cores with a shared floating point and vector unit).

I've even managed to pick up a few benchmarks (I'll use Cinebench 11.5, multi-threaded, because it's most relevant to our topic):

Core i7-920 (last gen, X58, [email protected])
single channel: 4.68
dual channel: 4.64
triple channel: 4.65

Six core, Core i7-980X, X58, 3x2GB)
[email protected]: 8.53
[email protected]: 8.63

The last one is quite interesting considering that the slow memory is way below standard memory timings while the fast memory is scraping on the ceiling of what's available now. And still the performance difference is 1.2% at most.

There are synthetic benchmarks where the difference is more noticeable. But even something like compiling a linux kernel (which is quite memory intensive) only shows a single digit difference in percentages at most.
(I'd link to the article but it's on a print magazine and pay per view on-line, as well as in German ;) ).


OCing works because it accelerates all relevant clocks.

It doesn't anymore. Actually one of the trickiest bits is keeping most clocks in spec while pushing the CPU as far as it can go.
Sandy Bridge makes this easy as the only overclocking that is usually done just affects the core multipliers (which uses a base frequency of 100Mhz). This is completely decoupled from the rest of the system and incidentally also used by Intel itself for automatic overclocking (->Turbo Boost(tm))
To the contrary, you wouldn't want to mess with the base clock at all since everything is derived from it. As little as a deviation of 5-7% can result in an unstable system.

Cheers,
Mike

Qexit
07-03-2011, 05:03 PM
Sandy Bridge makes this easy as the only overclocking that is usually done just affects the core multipliers (which uses a base frequency of 100Mhz). This is completely decoupled from the rest of the system and incidentally also used by Intel itself for automatic overclocking (->Turbo Boost(tm))
Absolutely, overclocking the 2600K to 4.5GHz that I mentioned in my earlier post was embarassingly simple. Go into BIOS, change multiplier to 45, save and reboot. That was it ! The system rebooted and was immediately running at the new speed. It even had a stock Intel h/s and fan :D I remember the good old days when every reboot could mean a system lockup or no restart, followed by a few worrying minutes spent clearing the CMOS RAM to reset everything back to default. I suspect this particular little black box could go up to 5.0GHz with a better cooler and a little extra tweaking :thumbsup:

To Edz, you started this thread asking for advice on the best entry level system. You've received a lot of practical advice which you now appear to trying to reject or disprove using the theoretical information you have found and read up on in your research. You've even 'recommended' that people not use the 2600K cpu but to go with the older system you were originally considering even though actual bench tests using LW scenes suggest otherwise. Have you decided which system you are going to purchase yet or are you going to conduct more research ?

gristle
07-04-2011, 03:01 AM
Dont spend too much time researching, or you will loose any time you have gained from the extra bandwidth!

Hieron
07-04-2011, 02:21 PM
Dont spend too much time researching, or you will loose any time you have gained from the extra bandwidth!

+1
My god what a thread.

I really think you (Edz) are pulling our leg here, this can not possibly be serious. If it is, goodluck.

Red_Oddity
07-04-2011, 02:25 PM
If it's a windows based system running NTFS (what else) there's one or two registry tweaks that can heavily speed up browsing directories that contain a few thousand files.
Essentially, by default, windows writes a last access time per file even if you just browse it (which make sense technically - but slows things down).

Cheers,
Mike
Yeah we have that disabled, 8.3 file name writes as well (why in 2011 that still is turned on by default, who knows?).
Just for a fun test turned last access time stamps on again...Browsing those massive folders is like watching grass grow.

Also, no registry tweaking necessary these days Mike, fsutil is your friend.

Edz
07-04-2011, 06:48 PM
Dont spend too much time researching, or you will loose any time you have gained from the extra bandwidth!

finished research last week.

Lightwolf
07-05-2011, 01:40 AM
Also, no registry tweaking necessary these days Mike, fsutil is your friend.
Yay, progress! (it seem to do the same thing anyhow) :D

Cheers,
Mike

Red_Oddity
07-05-2011, 04:48 AM
Yay, progress! (it seem to do the same thing anyhow) :D

Cheers,
Mike

But command line utils make wrecking my registry so much more fun :D

(Just wish all tools had the -whatif switch Powershell 2 has, half the time i don't know what the f i'm doing anyway :o )

erikals
07-14-2011, 01:58 PM
5.5    \;o
http://www.youtube.com/watch?v=KmMJFQOFKus

  

JustForFun
07-14-2011, 02:38 PM
Nice. I should replace my garglebite mobo with that puppy. :D That is after I get its DOA replacement. :(

GregMalick
07-14-2011, 06:39 PM
Aloha,

My boss just asked me for specs for some new PC for our developers.
We're software developers so we don't need hardCore video cards.

But Windows7-64bit and 8GB memory and good CPU speed seems essential for a good development workflow. (Right now I'm waiting 3-5 minutes for my app to publish to the local server on my box - which puts a kink in my code/testing flow).

I'm kind looking at http://www.newegg.com/Product/ComboBundleDetails.aspx?ItemList=Combo.691189

Since I understand the i7-2600K can be overclocked a tad.


Any comments?

OH! I need a recommendation for a video card (I like NVidia).
I don't do Games - but having outputs for a dual monitor setup would be sweet when developing/debugging.

BTW, I'm hoping that with OS & video card the system will be about $1200.

JustForFun
07-15-2011, 07:37 AM
OH! I need a recommendation for a video card (I like NVidia).
I don't do Games - but having outputs for a dual monitor setup would be sweet when developing/debugging.[/I]

Earlier in this thread, Lightwolf recommended nvidia 560Ti vcards.

Lightwolf
07-15-2011, 07:43 AM
Earlier in this thread, Lightwolf recommended nvidia 560Ti vcards.
They might be a bit over-powered for a pure developer rig (unless you develop gfx) - but they also have a fairly low power use when they idel (20W or so for a single display). And when coding that's usually the default state.

Anything below that should do just as well for a coding rig... heck, even the GPU on the Intel chip would be sufficient.

Cheers,
Mike

GregMalick
07-15-2011, 09:53 AM
They might be a bit over-powered for a pure developer rig (unless you develop gfx) - but they also have a fairly low power use when they idel (20W or so for a single display). And when coding that's usually the default state.

Anything below that should do just as well for a coding rig... heck, even the GPU on the Intel chip would be sufficient.

Cheers,
Mike

Aloha Mike,

That 560Ti is a bit much on both power & price. Looking for something more in the $100-150 range that would support dual monitors and be sufficient for win7-64.

oh, and any comments on the base rig?
http://www.newegg.com/Product/ComboB...t=Combo.691189

mahalo

Lightwolf
07-15-2011, 10:02 AM
Aloha Mike,

That 560Ti is a bit much on both power & price. Looking for something more in the $100-150 range that would support dual monitors and be sufficient for win7-64.
You could always stick with the on-board GPU and then see if you need more.


- Supports HDMI with max. resolution 1920 x 1200 @ 60 Hz
- Supports DVI with max. resolution 1920 x 1200 @ 60 Hz

And afaik it allows for dual monitors as well.


oh, and any comments on the base rig?

There's nothing wrong with even though I think both the PSU and the RAM are overpowered.

I'd also drop the HDD for a "green" one and instead get a 120GB SSD as a boot drive (and for the compiler of course :) ).

Cheers,
Mike

GregMalick
07-15-2011, 12:51 PM
You could always stick with the on-board GPU and then see if you need more.

And afaik it allows for dual monitors as well.

There's nothing wrong with even though I think both the PSU and the RAM are overpowered.

I'd also drop the HDD for a "green" one and instead get a 120GB SSD as a boot drive (and for the compiler of course :) ).

Cheers,
Mike

Awesomely Great ideas about the SDD and Green drive. I changed the request to my boss.

I see the ASUS P8Z68 has D-sub+DVI Video ports and 1 HDMI.
But it also says Onboard Video: None

Doesn't that mean I need a video card?

Lightwolf
07-15-2011, 02:25 PM
I see the ASUS P8Z68 has D-sub+DVI Video ports and 1 HDMI.
But it also says Onboard Video: None

Doesn't that mean I need a video card?
Well, technically it's correct... the on-board video is on the CPU. The motherboard only provides the connections.
The Core i7-2600K (and pretty much every CPU in the core i#-2#### series) are technically a CPU+GPU (using a part of the RAM as shared memory for the GPU part of it).

The Z68 chipset allows you to use the GPU (thus the video connectors), P67 based motherboards don't.

Cheers,
Mike

GregMalick
07-15-2011, 03:41 PM
Thanks Mike,

That helps a lot.

JustForFun
07-18-2011, 07:43 AM
The Z68 chipset allows you to use the GPU (thus the video connectors), P67 based motherboards don't.

So if the vid card has two monitor outputs, one would be able to use all three for three monitors?

Lightwolf
07-18-2011, 07:53 AM
So if the vid card has two monitor outputs, one would be able to use all three for three monitors?
At the moment I don't think so. Well, I don't remember reading anything that said it's possible.
You can use either the on-CPU GPU or a discrete GPU to access the idle GPU with a bit of software. But I don't think you can drive monitors from each at the same time.

The only reason to access the Intel-GPU (if you use a normal GPU anyhow) is because of the fast and decent video encoder. If that's not a priority then you can ignore it.

Cheers,
Mike

Quiet1onTheSet
07-18-2011, 10:40 AM
At the moment I don't think so. Well, I don't remember reading anything that said it's possible.
You can use either the on-CPU GPU or a discrete GPU to access the idle GPU with a bit of software. But I don't think you can drive monitors from each at the same time...

Actually, Lightwolf and others -- there's been much thinking, designing and subsequent discussion on this in varied places around the web -- with some encouraging outcomes.

For starters, I simply did a Yahoo search on: <QuickSync, simultaneous integrated and discrete graphics> and found some compelling info for those with Sandybridge setups, desiring to explore this.

Firstly, at tomshardware site, dating back to April 2011:
http://www.tomshardware.com/forum/289385-30-discrete-graphics-card-integrated-graphics-simultaneously

I'd imagine one should perform due diligence to investigate whether or not the Sandybridge socketed motherboard/BIOS they're considering for purchase, supports this "adapter switching" functionality.

After all, as one pundit at superuser.com put it, according to his understanding, this should work if...

1. The motherboard itself allows you to keep the integrated graphics enabled with a discrete GPU attached to the system, AND

2. the drivers for the two (integrated and discrete GPUs) coexist peacefully in your OS of choice, AND

3. you have at least one monitor cabled to the integrated graphics display connector.

And more exciting still, is a pertinent AnandTech article with discussions following, that dates a bit earlier (Feb 2011), wherein a piece of software no less -- called Virtu provides the magic to make the simultaneous use of both integrated and discrete graphics card a reality with the new chip architecture Sandybridge brings to the table, that gets around some previous troublespots.

http://www.anandtech.com/show/4199/lucids-virtu-enables-simultaneous-integrateddiscrete-gpu-on-sandy-bridge-platforms

Kind regards to all,
Q1:thumbsup:

Lightwolf
07-20-2011, 01:19 AM
Well, I followed my own advice, meet Eastwood:
96850
i7-2600K (no OC yet)
Scyte Grand Kama Cross (huge but quiet and it cools the area around the CPU as well).
16GB DDR3-1333
ASUS P6P67-M (I don't really need any of the Z68 features nor does SLI make sense)
nVidia 2GB 560Ti (Gainward Phantom, which is fairly quiet even under load)
OCZ Vertex 3, 120GB (no problems with the current firmware, knock on wood)
2GB green Seagate HDD for data
Cooler Master CM 960 II case ( still a little tacky for my taste but it has great cooling and an extremely neat SATA docking station built in).

Sorry, no benchmarks yet, I'm still moving things over.

Cheers,
Mike

GregMalick
07-20-2011, 01:43 AM
Well, I followed my own advice, meet Eastwood:
96850
i7-2600K (no OC yet)
Scyte Grand Kama Cross (huge but quiet and it cools the area around the CPU as well).
16GB DDR3-1333
ASUS P6P67-M (I don't really need any of the Z68 features nor does SLI make sense)
nVidia 2GB 560Ti (Gainward Phantom, which is fairly quiet even under load)
OCZ Vertex 3, 120GB (no problems with the current firmware, knock on wood)
2GB green Seagate HDD for data
Cooler Master CM 960 II case ( still a little tacky for my taste but it has great cooling and an extremely neat SATA docking station built in).

Sorry, no benchmarks yet, I'm still moving things over.

Cheers,
Mike

what's the total cost?

Lightwolf
07-20-2011, 02:22 AM
what's the total cost?
Around 1300 incl. 19% VAT (roughly US$1550 without VAT).

That's including the OS (Win7 Ultimate, because I want to switch languages) and the 650W PSU, no optical drive.

You can probably shave off that price a little (different graphics board and SSD, Win7 Pro or Home).

Cheers,
Mike

GregMalick
07-22-2011, 11:50 AM
Just sent in my request for 5 PCs to be built from:

Intel Core-i7-2600K (http://www.newegg.com/Product/Product.aspx?Item=N82E16819115070)

MB (http://www.newegg.com/Product/Product.aspx?Item=N82E16813131729)

PS (http://www.newegg.com/Product/Product.aspx?Item=N82E16817153136)

Win7-64 (http://www.newegg.com/Product/Product.aspx?Item=N82E16832116992&cm_re=windows_64_bit-_-32-116-992-_-Product)

8GB DDR3 (http://www.newegg.com/Product/Product.aspx?Item=N82E16820233170)

2.5 TB Green Drive (http://www.newegg.com/Product/Product.aspx?Item=N82E16822136765&cm_re=Green_drive-_-22-136-765-_-Product)

120GB SDD for OS & dev-IDE (http://www.newegg.com/Product/Product.aspx?Item=N82E16820233181)

Tower (http://www.newegg.com/Product/Product.aspx?Item=N82E16811133181)

2-DVD drives (http://www.newegg.com/Product/Product.aspx?Item=N82E16827151233)

Comes in at about $1270 (not including shipping/tax)

These machines need to last us 2-3 years and will only be used for development. I'm really hoping for approval.

JustForFun
07-24-2011, 07:50 AM
Well, I followed my own advice, meet Eastwood

Nice. How well does that cooler work (Idle/load temps)? I got my i7 2600k box running and I thought the cooler that came with it would be sufficient for now. But the temps I saw were scarey (~38C/~78C). I used AS5 and had good contact.

GA-Z68MA-D2H-B3
GX-750W
DDR3 16GB
GeForce GTX 560Ti
Intel 320 120GB (apps)
Green 1.5TB (projects)
CM 912 (Fans-o-plenty)

96909

Lightwolf
07-24-2011, 07:58 AM
Nice. How well does that cooler work (Idle/load temps)? I got my i7 2600k box running and I thought the cooler that came with it would be sufficient for now.
On idle ~29C, on load ~60C (I'm just rendering a quick VPR preview to test).
And the CPU fan doesn't even break a sweat (there's no increase in the noise level).
I suppose it could take a lot more once it actually starts to rev up ;)

Cheers,
Mike

realgray
07-24-2011, 03:59 PM
When the new 6 core sandybridge I7's hopefully come out in the Fall, will I be able to use it with a P67 or Z68 motherboard? I would love to build a new machine now (2600K) and change out the CPU when the new ones arrive. Maybe just make a stripped render node with the remaining quad.

Lightwolf
07-24-2011, 04:03 PM
When the new 6 core sandybridge I7's hopefully come out in the Fall, will I be able to use it with a P67 or Z68 motherboard?
No, they have a different socket and chipset.

Cheers,
Mike

JustForFun
07-25-2011, 08:05 AM
On idle ~29C, on load ~60C

Do you know your ambient temp at that time? The Scyte Grand Kama Cross looks like a good choice for mine (don't want liquid), but I can't find any spec diagrams to see if it will clear these RAM heat fins.

Lightwolf
07-25-2011, 08:25 AM
Do you know your ambient temp at that time? The Scyte Grand Kama Cross looks like a good choice for mine (don't want liquid), but I can't find any spec diagrams to see if it will clear these RAM heat fins.
Around 20-30C or so. I don't quite remember (and also don't trust the MB sensor in that regard).

I didn't get RAM with heat fins in the first place (they're mainly cosmetic unless you really push it, which doesn't make sense with that CPU anyhow).

It seems to look o.k. though: http://www.xbitlabs.com/articles/coolers/display/scythe-grand-kama-cross_3.html#sect0

It is tricky to reach some parts of the board between the heatsink/fan and the case after you mount it though (especially toward the top).

Cheers,
Mike

JML
07-25-2011, 09:34 AM
I used to get asus motherboard all the time, but all of mine died after about 2 years, and I was not even overclocking.

I hear good thing about gigabyte but I'll buy intel motherboards now.

Lightwolf
07-25-2011, 09:43 AM
I used to get asus motherboard all the time, but all of mine died after about 2 years, and I was not even overclocking.

Asus and MSI here (some of both brands in their 5th year now).

Gigabyte is a bit tricky atm as they're switching over to EFI and their current BIOSes are a bit, erm, iffy at the moment. The did improve their fan control though (Intel is apparently still king of the hill in that regard, closely followed by Asus).

Cheers,
Mike

JML
07-25-2011, 10:01 AM
Just sent in my request for 5 PCs to be built from:
Intel Core-i7-2600K (http://www.newegg.com/Product/Product.aspx?Item=N82E16819115070)
MB (http://www.newegg.com/Product/Product.aspx?Item=N82E16813131729)
PS (http://www.newegg.com/Product/Product.aspx?Item=N82E16817153136)
Win7-64 (http://www.newegg.com/Product/Product.aspx?Item=N82E16832116992&cm_re=windows_64_bit-_-32-116-992-_-Product)
8GB DDR3 (http://www.newegg.com/Product/Product.aspx?Item=N82E16820233170)
2.5 TB Green Drive (http://www.newegg.com/Product/Product.aspx?Item=N82E16822136765&cm_re=Green_drive-_-22-136-765-_-Product)
120GB SDD for OS & dev-IDE (http://www.newegg.com/Product/Product.aspx?Item=N82E16820233181)
Tower (http://www.newegg.com/Product/Product.aspx?Item=N82E16811133181)
2-DVD drives (http://www.newegg.com/Product/Product.aspx?Item=N82E16827151233)
Comes in at about $1270 (not including shipping/tax)
These machines need to last us 2-3 years and will only be used for development. I'm really hoping for approval.

The config looks great, but thermaltake power supply?
I would think corsair or antec might be a better choice.. and roughly
same price..
that 200gb ssd is at a great price.. I got mine 2 years ago (128gb corsair)
amazing drive.

realgray
07-26-2011, 01:12 PM
No, they have a different socket and chipset.

Cheers,
Mike

How often do things get messed up in 1st gen Intel? Thinking about waiting for the Sandybridge-e but worried about the problems that came last year repeating themselves. If it's a gamble then I'll probably buy a couple 2600K systems for a render garden.

GregMalick
07-26-2011, 07:59 PM
The config looks great, but thermaltake power supply?
I would think corsair or antec might be a better choice.. and roughly
same price..
that 200gb ssd is at a great price.. I got mine 2 years ago (128gb corsair)
amazing drive.

Because this had to go through our IT dept, the SSD was not approved. Their head said that SSD's are only useful when shock resistance is needed. I said they are much faster - but he disagreed saying it was extravagant. So I dropped that in order to try and get it approved. It's still being batted between my boss & IT. I do have approval to personally purchase the SSD and use it in the box though. :D

I'm still waiting for final approval.
If this works out, with the experience of building them at work, I'll be all set to build one for myself. :bday:

Lightwolf
07-27-2011, 12:43 AM
I said they are much faster - but he disagreed saying it was extravagant.
Haha... well, speed is always extravagant.

It makes me wonder why you don't still work on i386s ;)

Cheers,
Mike

Rayek
07-27-2011, 01:42 PM
Because this had to go through our IT dept, the SSD was not approved. Their head said that SSD's are only useful when shock resistance is needed. I said they are much faster - but he disagreed saying it was extravagant. :

Extravagant?!! Even an old xps laptop of mine now works great again since I installed a ssd - before I had to wait about 4 minutes for boot-up. Now 30 seconds. System feels incredibly responsive.

My experience with IT deps: they lag behind generally. In more ways than one. ;-)

JustForFun
07-29-2011, 06:32 PM
It is tricky to reach some parts of the board between the heatsink/fan and the case after you mount it though (especially toward the top).

That sir, is an understatement. :eek:

I got that thing today and it's huge. Had to pull the board out to get at the two top push pins. It clears the RAM fins easily though. CPU temps dropped 20*C @100% load. :)

kopperdrake
08-02-2011, 06:49 AM
Guys - thanks for this thread - on the back of your advice (and some reading of course), I've just put an order in for:

Asus P8P67 R3 P67 MoBo
16GB Corsair Vengeance Low Profile DD3 1600
120GB Corsair Force Series 3 SSD - nice read/write speeds (550MB/s, 510MB/s) - only for OS and apps as all work sits on an external RAID
2GB Point of View GeForce 560 Ti
Akasa Venom Ultra CPU cooler - hopefully with that and the low profile memory I'll have enough clearance
And all wrapped up in the Fractal Design Arc Midi tower, which I've fallen in love with.

This'll be the first PC I've built myself in about 7 years - it's all changed a bit, but I'm looking forward to the O/C option now that you don't have to wear heat-resistent clothing to try it out.

erikals
08-02-2011, 10:43 AM
just bought this, quite satisfied... the i7 runs on 4.5GHz...

http://www.newtek.com/forums/showpost.php?p=1169611&postcount=114

 

kopperdrake
08-02-2011, 11:17 AM
Ooh - I hovered over the watercoolers but thought I'd see how a mid-range air cooler would handle the heat first, I've heard you can clock to over 4GHz on the stock cooler with the i7's running so cool anyway. Can't wait to get the rig together, I'll carry out some test runs as well and post them - would be great to see everyone's times :)

AmigaNewTek
08-03-2011, 03:11 AM
This processors would be nice:

http://www.youtube.com/watch?v=G0i3DB21rcg&feature=related

Lightwolf
08-03-2011, 03:20 AM
This processors would be nice:

http://www.youtube.com/watch?v=G0i3DB21rcg&feature=related
Maybe... or maybe not. I expect them to be slower on single-threaded tasks and roughly on par with the i7-2600K when multi-threading.

I'm certainly looking forward to see benchmarks (I suppose September is a realistic date).

Cheers,
Mike

kopperdrake
08-03-2011, 06:28 PM
Ok - on the new rig, non OC'd - HDRI_Spheres: 121s

OC'd by just pressing the 'go fast' button in the snazzy bios drops that render to 98.9s

JustForFun
08-04-2011, 04:41 PM
HDRI_Spheres: 121s

HDRI_Spheres? Where are those modeler/layout files?

kopperdrake
08-05-2011, 08:02 AM
HDRI_Spheres? Where are those modeler/layout files?

Assuming you've installed the Contents CD with LW10 then they're in Content/Scenes/Benchmark - all good scenes for us to check system speeds out :)

I have to say, having come from OS X via XP (Vista just scared me), I am absolutely loving Windows 7 - feels very stable, MS are even getting some of that 'this works as I expect' that Apple are good at. Hats off to MS, it's a pleasure to use this PC :)

JustForFun
08-05-2011, 03:01 PM
Assuming you've installed the Contents CD with LW10

Still on 9.6. Not sure when I'll upgrade.

GregMalick
09-07-2011, 02:45 PM
Believe it or not, I'm still trying to get this purchase through...but I need some help.

Lenova has forwarded a quote using Intel Core E3 - 1270(3.4 Ghz) and says they are better/faster than the i7-2600K.

Can anyone shed some light on this so I can push this purchase ahead.

erikals
09-07-2011, 03:12 PM
not so sure, i overclock my i7 2600k to 4.5 GHz, no problem,...
Xeon can't overclock afaik.

i'd go with the 2600k.

Lightwolf
09-07-2011, 03:42 PM
Believe it or not, I'm still trying to get this purchase through...but I need some help.

Lenova has forwarded a quote using Intel Core E3 - 1270(3.4 Ghz) and says they are better/faster than the i7-2600K.

Can anyone shed some light on this so I can push this purchase ahead.
Better maybe, as it's a server grade CPU that allows for ECC memory. Certainly not necessary though.
But hey, ask them for the E3-1290 (2.6GHz base clock) ;)

Cheers,
Mike

GregMalick
09-07-2011, 05:46 PM
Unfortunately they aren't interested in the staff building them from parts, so it looks like price rules the day.

five Lenovo E3-1270 PCs for $6,481 vs five Dell i7-2600K for $10,157.06

At least i'll finally get a new machine.

Qexit
09-07-2011, 05:50 PM
Unfortunately they aren't interested in the staff building them from parts, so it looks like price rules the day.

five Lenovo E3-1270 PCs for $6,481 vs five Dell i7-2600K for $10,157.06

At least i'll finally get a new machine.Since you probably can't overclock the Dells, the Lenovos are probably a better deal as they'll run at pretty much the same speed but use less power and run a bit cooler.

GregMalick
11-01-2011, 07:37 PM
Believe it or not, I'm still trying to get this purchase through...but I need some help.

Lenova has forwarded a quote using Intel Core E3 - 1270(3.4 Ghz) and says they are better/faster than the i7-2600K.

Can anyone shed some light on this so I can push this purchase ahead.



Believe it or not ... our machines arrived today (finally).

erikals
11-01-2011, 07:49 PM
unsure, the think is, the 2600K is really easy to overclock, not sure about the E3...

Rayek
11-01-2011, 07:58 PM
the i3 1270 is essentially a server version of the 2600k - performance seems more or less identical. Differences: xeon got the intel gpu ripped out, which saves some wattage. And support for EEC ram was added.

Check out this article:
http://www.servethehome.com/intel-xeon-e31270-sandy-bridge-benchmarks-review/

erikals
11-01-2011, 08:58 PM
oh, it's a xeon, then forget about overclocking,... i think i would go for the 2600K...

edit: (hm, think i misread here, looks like they went for the E3 :]

Qexit
11-02-2011, 05:14 AM
unsure, the think is, the 2600K is really easy to overclock, not sure about the E3...The 2600K is indeed easy to overclock if you have access to the required settings in the motherboard BIOS. However, Dell bioses do not give you access to this facility (unless they have changed in recent months) so it is not an option/consideration with the systems mentioned by GregMalick. I believe Lenovo systems also lack this facility.

GregMalick
02-20-2012, 02:42 PM
Believe it or not, I'm still trying to get this purchase through...but I need some help.

Lenova has forwarded a quote using Intel Core E3 - 1270(3.4 Ghz) and says they are better/faster than the i7-2600K.

Can anyone shed some light on this so I can push this purchase ahead.


Believe it or not ... our machines arrived today (finally).



Sorry I haven't mentioned this, but the time to clean(compile) & publish went down from 5-10 minutes to a fantastic 1-minute or less (mostly less).

I am one happy programmer.