Anyone tried Woodcrest?

[jacross]

I know the woodcrest processors are still pretty new... but it's time to upgrade (funny how that always seems to coincide with major point releases of :lwicon: ). I'm curious if anyone has had a chance to play around with a woodcrest system in either 32 or 64 bit (or better yet... both) and how it might compare to a dual, dual-core AMD machine under LW.

Here's my dilema... Right now I have a pair of Opteron 244's (1.8GHz) that I bought when the Opteron platform was pretty new combined with a pretty new motherboard that will accept up to the fasted dual, dual-core processors (it also has dual 16x PCI-E so I'm good on the video front). I could easily and fairly cost-effectively upgrade the 244's to a pair of dual core 265's, 270's or even 275's (1.8GHz, 2.0GHz and 2.2GHz respectively) without needing to change anything else. The only downside is that the 940 pin opteron platform will be replaced shortly with Socket F that should allow for the upgrade to quad-core when they're available. So if I go this route I know that I'll have to go for a whole new system in another year or so... but it'd bide time til then. The other option is to go with a woodcrest system now because it's available and it too will offer the upgradability to quad cores when available (or so they say). So it basically comes down to: do I bide my time, or buy into a new architecture that's supposed to be upgradable in the future? I'm not looking for an Intel vs. AMD flame war... I've used both and I always go with what's the best solution at the time. I'm just looking for information. Much appreciation to everyone! :thumbsup:

[hstewarth]

I am in the process of building a dual 3Ghz Woodcrest. But I waiting on Supermicro X7DA3/i ( SAS Workstatiion ) motherboard.

I have Lightwave 8.5 and 9.0 will come in the next couple of weeks - already purchase it - got email from Safe Harbor.

The systems specs are as follows

Dual Xeon 3Ghz 5160's
Supermicro X7DA3/i Motherboad
Supermicro SC745TQ-700 Case
Supermicro AOC-LPZCR2 Raid controller with 256M ram and 600Mhz Intel 80321 IO processor.
4 1G 667Mhz FB-DIMM
eVGA 7900GT Signatiure Edition
5 Seagate 320G 7200.10's SATA II drives in Raid 5 ( likely get SAS's drives - but not this year - spent too much already on it ).
Creative X-Fi Elite Pro
NEC 16x DVD burner
Dell 24in 2405 LCD Monitor.

I do planned to make a interesting Benchmark reveiw on the machine which include performance tests include Lightwave 3D 8.5 and 9.0 and also Vue 5 Infinite.

Some of parts already purchase, but primary Motherboard/case/cpu/memory have not been. Waiting primary on the motherboard - probably next month some time for entire system.

[amigo]

I am in the process of building a dual 3Ghz Woodcrest.
....
The systems specs are as follows

Dual Xeon 3Ghz 5160's
.....

Was it a big decision for you to go with the 5160 over lower models? Did you consider cost vs speed against other models or you just wanted the fastest available so that you don't need to upgrade anytime soon?

I'm sort of looking at the 5140 (or 5148 which is the low power version of 5140) and believe might be the best bang for the buck. To me personally, heat is what drives me crazy the most, along with noise, and I figure you'd still have to cool that 5160 pretty well, too, which would require cooling contraptions and what not.

[hstewarth]

Was it a big decision for you to go with the 5160 over lower models? Did you consider cost vs speed against other models or you just wanted the fastest available so that you don't need to upgrade anytime soon?

I'm sort of looking at the 5140 (or 5148 which is the low power version of 5140) and believe might be the best bang for the buck. To me personally, heat is what drives me crazy the most, along with noise, and I figure you'd still have to cool that 5160 pretty well, too, which would require cooling contraptions and what not.

Well I thought about that lately, but I decided I will go for fastest one - some part related to the actual review and the fact that its the fastest cpu out. I would normally go for 5140 because its the best bang for the buck and could save a lot of money. Who knows if the current motherboards will work with quad cores.

Now will cooling, I think this series is a total different story. Dual 5140 or 5150 will use same Watt as some of current Pentium D's . Use the PSU calculator, I determined the system to be around 450W.

[amigo]

....
Who knows if the current motherboards will work with quad cores.
I remember reading somewhere that the motherboard makers for the latest and upcoming (end of July) CPUs were building functionality into their motherboards so they work with quad core processors.
Siince Intel has moved the released date for the quad cores to Q4 2006, I think it would piss off a lot of people if they'd buy motherboards today or in coming months and not be able to put the quad core around end of the year. Especially since quad cores are pin compatible with current CPUs...

[hstewarth]

But there have been some issue with Woodcrest chips not working on some of early 771 boards that supported the Xeon 50xx series ( Netburst ).

I also think the quad core will be much more expensive and also less total power on the cpu. I think I heard that the bus speed will only be 1066 instead of 1333Mhz.

The problem is that you don't know until right now.

I think also I need to picky on memory, yes it might be nice to have Curcial - but they are currently not certified for the motherboard by manufactor. Acording to article on 3d professor this important. Here is link to article

http://www.3dprofessor.com/review.asp?id=196

[jacross]

The problem is that you don't know until right now.
:agree:

I am looking at a pair of 5140's as they are the best bang for the buck, especially if the board is upgradable to quad core in Q4 06. I agree though, there's no guarantee until it actually comes out and it's tested. As for memory, I hear that either the Samsung or better yet Kingston memory is the best way to go at the moment. Have you decided on which manufacturer you're going with?

I'm still not 100% sure which direction to take though. The new Socket F AMD Opterons are due to be released sometime around the beginning of August (boy, that came up fast). They too are supposed to be upgradable to quad core when it becomes available in roughly a years time. Also, you can use regular Registered-ECC DDR2 instead of the FB-DIMMS that suffer from higher heat and latencies. Granted, the AMD solution won't beat the Intel solution this year, but when quad core comes out, it could be a whole new ballgame. Intel is just putting two dual core chips on one substrate whereas AMD is going with a true quad core (for round one of the quad core releases anyway). Intel has already stated that their quad core design will be at approx 120W whereas again, AMD is going to try to stay within the 85W - 95W range. Things could flip flop yet again. Granted, the AMD quad core is 6mo. behind Intel... but for me that's still close enough to consider. Seems like predicting the future is getting harder all the time :).

Have you upgraded to LW 9 yet? If so, which vid card are you running? I assume from the vid card you listed in your purchase that you don't feel that LW 9 will make use of pro OpenGL cards, would that be a fair assessment? Thanks again for your feedback!

*EDIT*
I almost forgot, the new Socket F Opterons are also supposed to natively support DDR3. They've designed the dual core in the same way as Intel has so each core can throttle down when not in use, the difference here is that also includes the memory controller since it's built into the die (which can throttle independantly of each core). Although this won't affect maximum consumption I don't believe, it could have a significant impact on typical consumption to lower your heat output and power usage. Again, the Intel solution looks to have the performance crown, they're offering quad cores first (unless they encounter a problem, which I don't foresee) and currently they have good power consumption (much better than the previous XEON's). It's still completely up in the air for me. Can you share your thoughts on why you decided to go with Woodcrest? Much appreciated!

[hstewarth]

I have upgraded to Lightwave with printed manuals from Safe Harbor. I should get mind in 2 weeks from email that I got.

As far as video, I will be running eVGA 7900GT Signature edition on the Woodcrest system and currently run 7800GS AGP on my 3.2Ghz P4.

One thing I notice in reveiws of Woodcrest, SSE Memory trandfer is very impressive. 45,000 compare to 4000 or 5000 with previous generations. I hope that Lightwave developers are researching how to take advantage of SSE Memory trandfer.

I have edited my discussion to keep them on topic.

[jacross]

hstewarth,

No offense, and I really don't want to start something here... but to the comment:

In my opinion, I don't trust AMD comments.. I think they got caught off guard and likely trying to buy time and sway people away from the Woodcrest

Isn't that what Intel did when they were talking about Woodcrest while trying to get people not to buy AMD Opterons? Intel had a whole marketing blitz talking about their new Next Generation Micro Architecture and how great it was before it was available to try to bide time. And this was after nearly two years of being caught 'off guard'. For the most part, the processors are still really hard to find! Then people started blasting AMD for not sharing more information. When AMD did they just said it was a marketing ploy to try to steal Intel's thunder with Woodcrest? Again, no offense intended, but aren't they both guilty then?

I've had both Intel and AMD machines. They've both been great! Before the Athlon I only bought Intel (386, 486, Pentium, Pentium II, Pentium III), and since the Athlon, and now the Opteron I've only purchased AMD. Now that it looks like Intel has something good I'm looking at them again. That wasn't to say that AMD was out for the count, I'm just investigating my options. I don't have a preference for either as a company, I'm after the best products.

[hstewarth]

I have edited my discussion to keep them on topic.

[jwilli3]

But shortly later AMD comes out with statements that there new products are lower power and they also have quad core coming....


Shortly after? No I don't think so, AMD has been targeting lower power chips for years and also has had quad core chips on the roadmap. The K8 core was designed from the start with multicore support in mind, remember? Intel was the one with a stop-gap solution.

I'm glad your crystal ball can show us that AMD systems in the future will run worse with Nvidia video cards, seeing as how that makes no sense and all. :confused: Could you explain that to us? Intel currently makes their own graphics chips, just like AMD now. Are you also suggesting that Nvidia cards work poorly with Intel sysems? Thats what your logic is implying.

Your opinions seem to be based more on a love of a particular company than reality. Intel has a great chip out now, which is good for everybody. Making stuff up about how AMD systems will run in the future reflects poorly on you and is good for no one.

[hstewarth]

jwilli3,

I have decided to edit my reviews because this was getting out of hand. If you interested in
performance results with Woodcrest system and Lightwave 3d 8.5 and 9.0, please look forward to
my review.

[amigo]

I am also looking at 5140 or 5148 (low power) for my new system. For the motherboard I'm eye-ing out Intel S5000XVN, but specs are pretty scarce right now for it. I called Intel today and the tech did not know whether the board supported EFI or BIOS, nor if it had TPM chip. I asked about support for quad-core CPUs but he said he did know, nor could he comment on upcoming processors that have not been released yet, figures.

On the topic of quad core processors, just a bit of clarification from above (jacross I believe said) that Intel is putting two dual core processors on the single substrate. That is true for the Clovertown but not for the Tigerton, which will be true quad core on the single dye.

As for the FB-DIMMs producing more heat than DDR, where did this information come from? I thought that because FB-DIMMs are not as dense as DDR ones, heat produced should be significantly lower. Am I wrong in thinking this? (I guess I'll just Google for it later...)

Either way I consider investing in a single 5140 now and wait till New Year for the Clovertown. A single dual core CPU should last a few moths and if prices go down another one can be added. Then if previews of quad core prove to be good, go for that, else stick with what I got. :)

[hstewarth]

Only thing I heard is that the FB-Dimm use another 5 watts - not any heat issue. I expect the the quad cores to be quite expensive.

[jacross]

hstewarth,

Good catch on the SSE memory transfer, I didn't even notice that. Hopefully that will turn into an actual performance increase for LW since it appears to use SSE heavily. My only concern there is that Woodcrests overall memory bandwidth rates (from the benchies I've seen so far) are still lagging because of the higher latency of the FB-DIMMS. How this affects performance (although possibly not at all) with Intels 'memory disambiguation' remains to be seen. All I can say is that I'm very much looking forward to your review! One can never have too much information :).

Also I wanted to apologize if I came across inappropriately. I wasn't trying to pick on you, I was just confused by your answer. I also wanted to try to take out arguements of either manufacturer over the other from a political perspective. Always point and counter-point. I'm trying to remain as neutral as possible between vendors and I'll be the first to admit that AMD wants our money just as bad as Intel, and as such is not entirely trustworthy either. They're both the same to me in that regard. Remember when Intel first demo'd their NGMA everyone said that the benchmarks had to be trumpted up somehow. Now low and behold, they've stood up! The same thing will happen to AMD (if history repeats). The real challenge is trying to separate the signal from the noise :). Again, sorry if I came across poorly.

Regarding the FB-DIMM's using more juice. With the extra buffer chip on the DIMM it adds an additional 5Watts of power per DIMM. Although this doesn't sound like much, if you have 4 in there that's an extra 20Watts, if you have 8 that's an extra 40Watts! Watts has a close relationship with heat, so the more power you add the more heat you have to get rid of. Think of it this way... if you have 2 x 65Watt Woodcrests with an additional 40Watts of power, that's just like having 2 x 85Watt Woodcrests. If you have two of the 3GHz bad boys that are already at 85Watts each then that would be like having two at 105Watts. Now of course the situation isn't that dire since we're not trying to remove all of that heat from the small space on one processor (like the previous XEON's). It has more surface area to escape from, thus it should be easier to cool. The main point still stands though that more power will equal more heat. I'll admit I'm a bit anal about this since I like my workstations to be as quiet as possible while I'm working, so it may not be a big deal to most people.

The benefit of course is the serial connection between the memory controller and the DIMM's which allow for simpler motherboard traces and higher memory densities. Unfortunately the buffer chip also increases the memory access latencies. There's always a good and bad to everything :).

To amigo

On the topic of quad core processors, just a bit of clarification from above (jacross I believe said) that Intel is putting two dual core processors on the single substrate. That is true for the Clovertown but not for the Tigerton, which will be true quad core on the single dye.

You are absolutely correct that Intel will eventually release a true quad core processor. However the first interation, and the one that's supposed to work with existing Woodcrest motherboards (according to the info I've seen so far) is Clovertown. Since this is the one with two dual core processors basically attached to the same substrate I must admit this isn't the most appealing solution. Will it be bad? I don't believe so, in fact I'm sure it will be very successful. But again, it's point and counter-point. What is the best bet? I'm not sure... I'm still on the fence. In any respect, I apologize for not providing better clarification in my post.

I'd just like to say again that I appreciate everyone's comments so far. It is definately helping me think through things. It's very much appreciated!

[hstewarth]

jacross,

First of all, with all the AMD-ATI stuff - its really getting crazy, but I feel this is not the place for such disucssions. I got too carely away.

On the Watts, right now using PSU Calculator, I believe that my watts will be 450W and I will have 700Watt. I have no idea what the actually noise level will be in the new machine, but right now my 3.2Ghz P4 is quite loud itself.

One thing I am not sure about is if I could use a PASSIVE heat sink which would definely lower the noise level. Maybe not with 5160's.

Anyway I did find a review of 5160 today, it uses the Supermicro X7DAE. My system should be similar to this but even has a better Supermicro motherboard - it is the X7DA3/i with 256M raid 5 caching card on it. Anyway here is the link. By the way my review probable will be first one that has games benchmark on such a machine. Which would be interesting because results could be compare to desktop. I think the memory on the Woodcrest is what makes them different than the Conroes.

http://www.hexus.net/content/item.php?item=6256%20target=

[jacross]

First of all, with all the AMD-ATI stuff - its really getting crazy, but I feel this is not the place for such disucssions. :agree:

hstewarth,

It's funny cause I just read that woodcrest review today. I thought it was good to pit the best AMD Opteron to the best Woodcrest XEON. Sure, there will be those that say that the AMD was down by 400MHz per chip so it wasn't fair... but I don't want to compare products that aren't available yet. That's AMD's bad for not being there yet.

I did notice that in the review they used passive heatsinks on the XEON's in a Supermicro case. I was hoping that this would be a good and quiet setup unfortunately Hexus said that it is perhaps the loudest workstation they've had in a while.

From HEXUS.net>> Indeed, it's very loud when the fan array is spinning at good speed, giving us cause to suspect the Magnetar X2 of being the loudest PC or workstation system to hit HEXUS for testing in quite some time. The Opteron box it goes up against is much quieter, although we suspect better heatsinks for the Woodcrest CPUs, that aren't passive, would do the same job for less noise.

Admittedly my Dual Opteron system I have now isn't the quietest, but it definately isn't loud either. I hear my power-supply over nearly everything else. Granted, I only have 2 HDD in a RAID 1... nothing like what you're putting together. I know that talking about noise isn't performance related, but to me if I'm sitting next to a loud box, it's hard to get anything done :). I would love to hear how your woodcrest system sounds with the standard coolers to see if it's quieter. If it's too much, I might be looking at the low power 5148's too to help out.

Lastly, I'll be the first to admit that Woodcrest's performance in that review was pretty darn good. The only point is that performance really seems to skip around depending on the application (for example the Opteron lost the Maya rendering benchmark but won the MentalRay one). How will the new rendering capabilities in LW 9 do on Woodcrest? I'm not sure, but I look forward to finding out from your review!

[hstewarth]

jacross,

You really can't compare AMD vs Intel in Ghz/Mhz rating anymore - it funny how things have change - with the Core 2 level procesor the speed of CPU is a lot more than other generations. For example at work I have both 1.66Ghz Core Solo notebook and 3Ghz Pentium 4, the Core solo is faster compiling code than the 3Ghz Pentium 4.

I think the Woodcrest will be signifcantly better than my current system which is a 3.2Ghz P4 for Lightwave. I would not doubt it it would be a factor 10x performance. Not really sure. It actually depends on portions that Lightwave can handle multi-threads. I hope to discovered this is my review - specific what applications take advantage of the multi-core and cpu and by what factor. I know that at least with Lightwave 8.5, Surface baking must be single threaded - so it likely only be 2 to 3 maybe more better performance.

What supermicro case was in the review, was it a TQ case - which I believe stands for quite. I also have a supermicro case ( old ) for my 2.26 at home on my farm - it one of my oldest boxes but its also never had any problems.

The following is the case I planned to get

http://www.supermicro.com/products/chassis/4U/745/SC745TQ-700.cfm

and the 256M raid controller - I am not 100% sure on this one.

http://www.supermicro.com/products/accessories/addon/AOC-LPZCR2.cfm

[jacross]

hstewarth,

I just did a quick skim of the review at Hexus and I don't see that they list which Supermicro case is used. Regarding comparing MHz between CPU's, for the most part your are 100% correct. This was especially true for the Pentium 4's because of their incredibly long pipeline which had a detrimental affect on their IPC. The new Core processors however are more like the old P3, and thus the Athlon than the Pentium 4's used to be. Both the Intel Core's and AMD Athlons have short pipelines that have a high IPC rate (although Intel's is higher at the moment). Of course they still have plenty of technologies to differentiate from eachother like on-die memory controllers for AMD and 'memory disambiguation' (out-of-order loads, better prefetching, etc) for Intel. From the reviews I've read so far, it appears that the Intel Core2 based processors have approximately 9 - 15% higher IPC than AMD at the moment. This means for AMD to beat a 3GHz woodcrest (using fuzzy math of course :) ), it would have to run at approximately 3.45GHz. I just made the point of the review not having at least 3GHz processors for each system as stating that I thought it was fair. You can't compare a 3GHz xeon to a 3GHz Opteron because AMD doesn't have a 3GHz widely available yet.

I'm not sure though you can expect a 10x performance increase for the dual XEON system over your current P4. That's pretty tough to come by even with Intel's new NGMA. I think at best 8x performance (and this is just for rendering, not actually working in LW) will be possible. Since most of LW is single threaded, I'm not sure how much of a boost you get actually working within LW. I could be wrong though (boy, I'd love to be wrong on this point). I am very much looking forward to hearing your reports once you get everything put together.

[pixym]

Hello jacross,

Last week I ordered a Work station to replace my "old" DELL Dimension 8250 (P4 3,06ghz FSB 533 - 1,5GB ram - radeon 9700 Pro)
This is the config:
DELL Precison 490 (mini tower)
Dual 5160 (Woodcrest 3ghz fsb 1333)
4GB ram
80GB HD
Quadro FX 3500 256MB Graphic card

I am about to receive it, once done, I will give you some replies (real gain from my old system etc.)

Edit: Bad news from my reseller :(
I should receive my workstation on august 12...

[hstewarth]

Edit: Bad news from my reseller
I should receive my workstation on august 12...


Understand complete, I am building my system by components - with the prime components from one place. Its also going to 3Ghz 5160's - The motherboard I desired is not yet ready.

Our systems should be quite similar - except that looks like I have a lot more storage and Raid controller while you have 3500 while I have 7900GT. For Lightwave they should be quite similar.

I hoping to build it mid july - looks like atleast mid august.

[Radamanthys]

Hello jacross,

Last week I ordered a Work station to replace my "old" DELL Dimension 8250 (P4 3,06ghz FSB 533 - 1,5GB ram - radeon 9700 Pro)
This is the config:
DELL Precison 490 (mini tower)
Dual 5160 (Woodcrest 3ghz fsb 1333)
4GB ram
80GB HD
Quadro FX 3500 256MB Graphic card

I am about to receive it, once done, I will give you some replies (real gain from my old system etc.)

Edit: Bad news from my reseller :(
I should receive my workstation on august 12...

I ordered about the same system, precision 690 (1kw chassi with 16 memory slot)

- dual 5160
- 8 GB RAM
- 250 GB HD
- video card will be my Asus 7900 GT Top (7900 GT at higher frequency)
- 30" Monitor

[pixym]

Radamanthys,

Have you ordered your workstation with Windows 64?

[Radamanthys]

Radamanthys,

Have you ordered your workstation with Windows 64?

yes, its a must to use 8 GB memory,

32 bit operating system can't use more than 4 GB (theoreticaly, because in practice memory adress reservation can make this, from my own experience, go as low as 2.75 GB memory usable, depending on your system, and especialy the amount of video memory on you graphic board)

[Radamanthys]

My screen should arrive today, and my workstation just ship, I should get it on wensday :D
it will be time for some benchmark against my PowerMac Quad and my Athlon X2 4600+

--> pixym: don't worry too much about the date they indicate, my station was supposed to be shipped on august 28th

[hstewarth]

hstewarth,

I just did a quick skim of the review at Hexus and I don't see that they list which Supermicro case is used. Regarding comparing MHz between CPU's, for the most part your are 100% correct. This was especially true for the Pentium 4's because of their incredibly long pipeline which had a detrimental affect on their IPC. The new Core processors however are more like the old P3, and thus the Athlon than the Pentium 4's used to be. Both the Intel Core's and AMD Athlons have short pipelines that have a high IPC rate (although Intel's is higher at the moment). Of course they still have plenty of technologies to differentiate from eachother like on-die memory controllers for AMD and 'memory disambiguation' (out-of-order loads, better prefetching, etc) for Intel. From the reviews I've read so far, it appears that the Intel Core2 based processors have approximately 9 - 15% higher IPC than AMD at the moment. This means for AMD to beat a 3GHz woodcrest (using fuzzy math of course :) ), it would have to run at approximately 3.45GHz. I just made the point of the review not having at least 3GHz processors for each system as stating that I thought it was fair. You can't compare a 3GHz xeon to a 3GHz Opteron because AMD doesn't have a 3GHz widely available yet.

I'm not sure though you can expect a 10x performance increase for the dual XEON system over your current P4. That's pretty tough to come by even with Intel's new NGMA. I think at best 8x performance (and this is just for rendering, not actually working in LW) will be possible. Since most of LW is single threaded, I'm not sure how much of a boost you get actually working within LW. I could be wrong though (boy, I'd love to be wrong on this point). I am very much looking forward to hearing your reports once you get everything put together.


First of all my 10x estimate is basically an estimated - 8x is closer to what it caluclated - but NGMA also has enchancements over the technology which I base this assumption on. I do have something else that is serious different with the machine, faster video and much faster hard disk. On the other parts like modelling and such, it depends on how well Lightwave 9 is multithreaded and also how well it uses OpenGL I think. I am going from 2G to 4G of memory also.

I actually planned to do a review on the machine with actually performance percentage difference for different activities. With dual core enable and off - not sure if I can disable 2nd processor. The idea is should the effectiveness of dual core/ dual cpu in different applications.

On another topic, there has been dicussions about the Cloverton ( Quad core ), but my guess is it is going to be quite expensive.

[pixym]

Radamanthys,
My resseler has just told me my workstation will be delivered on the end of next week...

[Radamanthys]

Radamanthys,
My resseler has just told me my workstation will be delivered on the end of next week...

just got mine :D :D :D

what an incredible machine, very fast and very quiet

my first test:

cinebench scores

Dell Precision M90 (Core duo T2600 @ 2.16 GHz)
1 core: 319
2 cores: 581

Apple Power Mac Quad @ 2.5 GHz
1 core: 382
4 cores: 1179

Dell Precision 690 (Dual Xeon 5160 @ 3 GHz)
1 core: 503
4 cores: 1618

I'll do some lightwave test tonight

[pixym]

Thank you for sharing these benches,
I stay tuned for your tonight results ;)

[hstewarth]

I ran the Cinibench 9.5 on my 3.2Ghz P4..

1 core 266
x core 220 ( I assume hyperthreading slows it down in this benchmark )

So combine with you #'s, I estimate that 3Ghz 5160 system that I planned to build will have the following incerease

1 Core: 1.8x
4 core: 6.1x

To me it sounds like Cinibench is not taking advantage of all cores - maybe lightwave will be a lot better. To me it looks like it only using 3 of 4 cores - but maybe that is just internal overhead.

[Captain Obvious]

To me it sounds like Cinibench is not taking advantage of all cores - maybe lightwave will be a lot better. To me it looks like it only using 3 of 4 cores - but maybe that is just internal overhead.
Cinebench's scaling is not as good as some of the other renderers out there. It uses all cores, but I don't think it achieves more than 85-90% efficiency or so per extra thread.

[pixym]

Radamanthys,

Where are your lightwave tests?
Thank in advance.

[pixym]

Talking about benches, I have made a test on modo 202, in order to test the speed of my new laptop (MacBook Pro Intel dual core 2ghz).
This is the results:

MacBook Pro dual core 2ghz: 2mn 38sec
PowerMac G5 dual 2ghz: 4mn 34sec
Dell P4 3,06 ghz: 5mn 28sec
Old laptop Powermac G4: 13mn 49sec

I was very stunned to see it outperforming my powermac dual 2ghz for 70%!
I am waiting for my Dell new Precision 490 minitower to run this test!

Edit: jpeg test removal

[hstewarth]

I was very stunned to see it outperforming my powermac dual 2ghz for 70%!
I am waiting for my Dell new Precision 490 minitower to run this test!


Just think the 3Ghz 5160 has 50% more clock frequency, dual cpu and the Core 2 based cpus are more efficient. I would doubt it will have that time close to minute.

[Radamanthys]

lightroom scene (found in lighting exemples that come with lw9)

Athlon 64 4600+ (dual core 2.4 Ghz): 499.3 s
dual Xeon 5160: 194.7 s

more to come

[pixym]

Thanks for sharing rada ;)

[hstewarth]

Thanks for test - I am curious if there a list machines for current Lightwave 9 benchmark. Was this in 64 bit or 32 bit and did it make any difference.

[Radamanthys]

Thanks for test - I am curious if there a list machines for current Lightwave 9 benchmark. Was this in 64 bit or 32 bit and did it make any difference.

64 bit version on the xeon (machine has 8 Gb RAM) and 32 bit version on the 4600+

I can also try some benchmark on a powermac quad (but the lightroom project dont open, there is one plugin missing) and on a Core Duo Laptop.

i tried some other exemples from the content that come with lw 9, but the one i tried where ways too fast to be revelant, i'll try harder this week end ;-)

[hstewarth]

Well it looks like my work is going to get a Dell 2900 Woodcrest system for SQL Server - massive amount of storage is need - for 6 billion records.

Probably before I get my system - like mind it will support SAS - but will actually be all SAS drives - over a Terabyte of data would be required.

One thing I like about the raid controller is the 256M cache is battery backup - one on my planned system is not.


I am planning to go with 4 Gig of memory and move 12Gig ( with 4 2G's ) next year. I notice that a lot of Woodcrest mention here have 8g - is there much benifit above 4 Gig with Lightwave.

Also did the benchmark use more than 2Gig of memory, if not than I am curious what 32bit Version of Lightwave results would be. If we had the figured than we could see the overhead of going to 64 bit - yes with larger scenes, 64 bit version has signficant advantage - but for smaller one I am just curious.

[Radamanthys]

dell 2900 is a great machien, we got one at work 2 weeks ago, with dual 2.33 woodcrest and 8 Go RAM (4x2). the only drawback is its incredibly noisy, and i was happy to move it to the computer room when i finished installing it. in our case the machien is there to host virtual servers.

[hstewarth]

dell 2900 is a great machien, we got one at work 2 weeks ago, with dual 2.33 woodcrest and 8 Go RAM (4x2). the only drawback is its incredibly noisy, and i was happy to move it to the computer room when i finished installing it. in our case the machien is there to host virtual servers.


How about the rack mounted version, it is quiter - I think it much more expensive.

[StereoMike]

btw, when tesing LW on your multicore machines:
sequences will render faster with severall lw instances running.
e.g. you have 4 cores, then the fastest results will come with 4 singlethreaded lw instances rendering the same scene (but different frames, use frame stepping).
Renders _much_ faster than using only one instance with 4 threads.
Why?
Easy answer: While rendering is multithreaded, many other things in LW aren't. Everytime your 4 threaded LW encounters such a task 3 of your 4 cores take a nap. Using 4 singlethreaded instances will do _everything_ 4-threaded (though at the cost of RAM).

Mike

[pixym]

Good notice StereoMike...
But for big scenes you need to have huge amount of ram, and windows 64, I think ;)

[hstewarth]

StereoMike,

This is same technique I found with Hyperthreading P4 with Mojoworld, I found that I could run 6 test in dual sessions as fast as 4 session serially. 50% increase in rendering.

I would assume the only way to take advantage of this is setup a network rendering for your frames. I am also curious if LW 9.0 is better at taking advnatage of multi-threads - software that does take advantage of multi-threading should not have to this.

On another thing, I was thinking maybe I should have gone the Dell workstation route instead of what I am doing, but one big advantage that I will have its number of big dirves - 8 of them and hot swappable. But still waiting for motherboard to be released.

[StereoMike]

Of course there's much place for improvement regarding multithreading in LW, but I doubt everything could be multithreaded. For that reason my guess is, it will always be faster to run severall instances. I made a feature request to add a new dropdown allowing severall rendering instances without the hassle of lwsn or severall opened layouts.
http://www.newtek.com/forums/showthread.php?t=53125
(I've upoaded even a menu mockup to that thread, but it isn't there anymore. Can it be, this forum looses images?)

Mike

[hstewarth]

lightroom scene (found in lighting exemples that come with lw9)

Athlon 64 4600+ (dual core 2.4 Ghz): 499.3 s
dual Xeon 5160: 194.7 s

more to come

I assume that you are talking about the QuickRoom68 Example that comes with Lightwave 9. I just recently did it on my 3.2Ghz P4 and it took 1399.5s

Thats slight above 7x improvement. I just found the X7DA3 motherboard is delayed a month or so.. but it has dual 16x slots.

[pixym]

I have received my DELL 490 today and unfornatly a processor has been forgotten!
One Xeon 5160 instead of two :(
My reseller will install the second one in the next 10 days...
I have made some benches on some of my LW scenes:
Some comparisons have been made between the 490 (with only one processor) and an "old nocona" dual xeon 3ghz HT on
FPrime: 1,66 times average faster
LW8.3: 1.65 times average faster
I have a scene 4,8 times faster and I don't know why...
I will do the sames benches when I receive my second 5160 xeon processor
Edit: after 14 days, I still have one processor instead of two...

[mgraveen]

I have posted some LightWave benchmark results on http://www.blanos.com
The system is a Dell PE2950, 2 processors, dual core (Woodcrest), 3.0GHz, 1333FSB. It's running 4GB of 533MHz memory. I ran Windows Server 2003 Enterprise (32bit). I didn't see any speed gains going to a 64bit OS with 64bit LightWave.

I used Autobench to submit the results so it's kind of screwy on how it reads the system specs. It reads it as 4 processors with 1 core each, instead of 2 x 2. It also reads the core speed as 1995MHz which is confusing because they are listed as 3GHz processors. It also didn't report any of the memory.

The rendering times on this system just scream! This two-way dual core box is keeping up and in many cases surpassing, 4 processor, dual core Opteron boxes. Of course, this is a server so I wouldn't suggest using a box like this to actually model objects (weak video).

Hope this helps.

Mike

[pixym]

Thanks,

Is there any way to fix your results on the blanos web site?

[mgraveen]

Thanks,

Is there any way to fix your results on the blanos web site?

I suppose I could send Chris Blanos an email with the request. I just wanted to let people who go to the web site looking for the benchmarks about the anomalies with the reporting on this system.

Next week I hope to benchmark the "Tulsa" processors with Lightwave.
(4) dual core 3.4GHz processors in a Dell PE6850. I'll also post the results on Chris's site.

Mike

[Captain Obvious]

The Tulsa is not an interesting CPU. It's just the same old Netburst crap as before. Chances are the Woodcrest Xeons are vastly superior at just about everything.

[mgraveen]

The Tulsa is not an interesting CPU. It's just the same old Netburst crap as before. Chances are the Woodcrest Xeons are vastly superior at just about everything.

You're probably right. But since I have the chance to benchmark the system, I figure what the heck. The results will speak for themselves.

Mike

[Captain Obvious]

You're probably right. But since I have the chance to benchmark the system, I figure what the heck. The results will speak for themselves.

Mike
Of course. :)

[pixym]

At last I receive today my second proc:
DELL dual INTEL woodcrest 3.0 ghz/4GB ram/quadro FX 3500

Raytrace: 21.8sec
Radiosity reflective things: 7.0 sec
Tracer radiosity: 1mn 22sec

[hstewarth]

lightroom scene (found in lighting exemples that come with lw9)

Athlon 64 4600+ (dual core 2.4 Ghz): 499.3 s
dual Xeon 5160: 194.7 s

more to come

I got my Xeon 5160 system yesterday and did the above tests with it and my P4.

Xeon 5160: 253.6s 4 Cores - 4 threads
3.2Ghz P4: 2167.8s HT - 2 Threads

This comes out to be 8.5 time increase in performance. I was running on Windows XP 64 bit with 64 bit version of Lightwave 9.0.

My machine specs

Supermicro Superworkstation 7045A-3
Supermicro X7DA3 mother board on 743T-645 low noise case
2 Xeon 5160's
4 2G 667Mhz FB-Dimm's
Supermicro AOC-LPZCR2 Raid Controler ( its fast ).
2 SAS 15k rpm in Raid 0
5 Seagate 7200.10 in Raid 5
eVGA 7900GT Signature Edition
Other misc stuff..

Use default settings after installed of Lightwave on Xeon. I am not sure why my time is slower than yours - I can only think is your have a quadro card and that might make the difference. But 8 time increase is not bad. What I expected.

[pixym]

...
Xeon 5160: 253.6s 4 Cores - 4 threads
3.2Ghz P4: 2167.8s HT - 2 Threads
... 8 time increase is not bad. What I expected.

It is a HUGE speed increase!

[hstewarth]

At last I receive today my second proc:
DELL dual INTEL woodcrest 3.0 ghz/4GB ram/quadro FX 3500

Raytrace: 21.8sec
Radiosity reflective things: 7.0 sec
Tracer radiosity: 1mn 22sec

What content are these scenes on. I didn't notice them on Lightwave 9.0 content directly.

Also for ones with 64 bit, 64 bit and 32 bit can be installed on same OS. Just put it in seperate directory.

[pixym]

they are in the v7-v8 content (I am not sure)

[hstewarth]

they are in the v7-v8 content (I am not sure)

Ok I will check them - I doubt I going put early versions on Xeon. I have them installed on 3.2Ghz.

[hstewarth]

just got mine :D :D :D

what an incredible machine, very fast and very quiet

my first test:

cinebench scores

Dell Precision M90 (Core duo T2600 @ 2.16 GHz)
1 core: 319
2 cores: 581

Apple Power Mac Quad @ 2.5 GHz
1 core: 382
4 cores: 1179

Dell Precision 690 (Dual Xeon 5160 @ 3 GHz)
1 core: 503
4 cores: 1618

I'll do some lightwave test tonight

Just for reference, I did it on my dual 5160. I think the difference here is that you probably have a Quadro and I have a eVGA 7900GT Signature Edition.

CINEBENCH 9.5
************************************************** **

Tester : hstewarth

Processor : Dual Xeon 5160
MHz : 3000
Number of CPUs : 4
Operating System : Windows XP 64

Graphics Card : eVGA 7900GT
Resolution : 1280x960
Color Depth : 32

************************************************** **

Rendering (Single CPU): 495 CB-CPU
Rendering (Multiple CPU): 1558 CB-CPU

Multiprocessor Speedup: 3.15

Shading (CINEMA 4D) : 575 CB-GFX
Shading (OpenGL Software Lighting) : 2036 CB-GFX
Shading (OpenGL Hardware Lighting) : 3878 CB-GFX

OpenGL Speedup: 6.74

************************************************** **