PDA

View Full Version : P4 Vs Xeon



paulk
02-12-2004, 03:19 PM
An article on Tom's Hardware Guide compared the
3.06 GHz Xeon CPU with L3 cache to the 3.06 GHz Xeon CPU without

http://www20.tomshardware.com/cpu/20030811/index.html

The article also included a 3.20 GHz P4 in the test results. For rendering Lightwave 7.5, "The increase in speed of the dual Xeon over the P4 is 52 percent". You'd think 2 CPUs would run twice as fast as one. Other benchmarks showed similar results, with the P4 sometimes beating the Xeons.

Granted, the article didn't cover using the VT. However, a Supermicro X5DA8 w/dual 3.06 GHz Xeons on www.pricewatch.com goes for $1433.99. A Gigabyte GA-8KNXP Ultra with a 3.2 GHhz P4 goes for $641.50.

So, what do I get for an additional $792.49? Is it possible some people have been lured by the Xiren Xong of the Xeon :D into buying more machine than they need?

robewil
02-12-2004, 04:05 PM
When comparing single processor systems vs dual processor systems, the results will vary depending on the application. In this particular example, 52% may be huge. Depending on the size of the project, that 52% may be the difference between making and not making a deadline. So, it's really depends on the purchaser to determine if the extra cost is justified.

Max Steel
02-12-2004, 04:13 PM
You are only examining the 'software' side of the equation.

The 'primary' advatage of going Xeon is not so much pure CPU speed (single or dual)... it's the architecture. ie. the popular SuperMicro motherboard your question in your subsequent message is about, features multiple PCI bus architecture (32, and 64bit), where as the Single P4 achitecture is only 32bit PCI.

On the P4 architecture.. your VT card and storage controller (and all other PCI cards) will share the one finite amount of bandwidth, and that bandwidth runs at a specified PCI bus clock speed.

Xeon architecture on the other hand features the same 32bit PCI bus at the regular bus speed, but also features seperate 64bit bus, which runs at various speeds, (dependent on mobo model) which are at least twice as fast as the standard 32bit PCI bus.

The idea is to place your storage controller and VT card on seperate PCI Bus's ... thus both cards can saturate either bus and obtain maximum benefit.

Bottom line... yeah.. spend the xtra and go dual Xeon with one of the popular Intel 7505 chipset based mobo's.

..steel..

paulk
02-13-2004, 05:03 AM
Max, both boards I cited have a built-in Adaptec U320 SCSI controller. Is this controller on its own separate bus in either (or both) boards? If so, there's no controller-VT[3] conflict. Otherwise, the smart thing to do is buy a "low end" Xeon or 875P mobo and add your own controller. Of course, there's the 875P single bus conflict

I guess the real question is how much is enough. www.videohardware.com says a P4 875P system can handle two uncompressed video streams at once, editing, DV editing and switching and a dual 7505 Xeon system can handle 2+ streams, record and playback while switching, etc. Do you buy more than you think you need "just in case"? I've seen posts from working professionals using older systems with (:eek: ) IDE video arrays and they seem happy.

So, everybody, what were your deciding factors? Have you ever been offered a job you couldn't handle because of your system? Please reply. Thanks.

SBowie
02-13-2004, 10:56 AM
My take - the 800FSB P4's make a very nicely performing system, and are quite adequate for most purposes. And, I agree there's not enough (if any) performance gain between the 512k cache Xeons and the 1 megcache version to make the latter worthwhile.

That all said and assuming the $$ factor isn't a deal-breaker (i.e., you just can handle the extra $$), given a choice I'd go with the dual Xeon in a heartbeat. They just fly. I'm not sure how the 52% figure was arrived at (too busy to read that piece today, sorry) and I don't know if it weighs all relevant factors, nor have I performed side-by-side performance tests, but even if that is entirely accurate, 52% is nothing to be sniffed at in a high perfromance system.

Just my 2 cents :)

mgrusin
02-13-2004, 11:16 AM
It also depends a little bit on your intended VT use. Live production is very demanding on a system especially if you're mixing live and pre-recorded elements. You can't afford pauses or stuttering here. As Paul (or one of his compadres) once said, "get the biggest, baddest system you can". On the other hand, if you're just editing, it's not a serious problem if it needs to pause occasionally to render out the complex sections. I edit DV (not RTV!) quite happily on a low end box with an IDE array. But I am aware of its limitations and would be nervous about taking it live.

-MG.

Max Steel
02-14-2004, 07:39 PM
>Max, both boards I cited have a built-in Adaptec U320 SCSI >controller. Is this controller on its own separate bus in either (or >both) boards?

The Gigabyte mobo you mention features standard 32bit PCI bus, so the onboard SCSI will share bandwidth with the Toaster card, and speedwise will be basically 'crippled'.

>If so, there's no controller-VT[3] conflict. Otherwise, the smart >thing to do is buy a "low end" Xeon or 875P mobo and add your >own controller. Of course, there's the 875P single bus conflict

Well. you could of course also use the SuperMicro mobo, but only populate it with 1 Xeon. (thus gain 64 bit capability)
However.... 'Get Dual CPU!' (grin)

>I guess the real question is how much is enough. >www.videohardware.com says a P4 875P system can handle >two uncompressed video streams at once, editing, DV editing >and switching and a dual 7505 Xeon system can handle 2+ >streams, record and playback while switching, etc. Do you buy >more than you think you need "just in case"? I've seen posts >from working professionals using older systems with ( ) IDE >video arrays and they seem happy.

A single P4 1.7Ghz, 400Mghz FSB can also handle two uncompressed streams when using Newteks recommended SCSI specs. I have several VT[2] and VT[3] systems.. off all kinds of Specs... including 875, 800Mghz FSB. Quite simply.. the Dual Xeon systems blow them all away. Of course I'm speaking from a pure performance perspective using the VT's complete featureset.

That is the nice thing about VT... in that it is very backwards compatible with slower computers.. it's just that the featureset and realtime capabilities also reduce.

Bottom line.. that SuperMicro X5DA8 is just simply better technology that the Gigabyte model you mention.
The X5DA8 is a server based product.... where as the Gigabyte mobo is just pretty boring really. SuperMicro also have otehr 7505 models to choose from, so the most expensive DA8 isn't your only choice. (tho it is a lovely unit)

The Gigabyte 875P will certainly run Toaster.. but if you want the best Toastering experience possible.. then I honestly suggest you opt for the SuperMicro range of 7505 mobo's.

Then.. you will have the technology to quite adequately run either 64bit UltraSCSI or 64bit IDE, and I would suggest that you look closely at the storage subsystem to make any dollar savings you require.

Well.. that my 2 cents worth... or 4 cents if I count both postings.
(heh heh heh)

..Steel..

mdwainwright
04-01-2004, 12:31 PM
Chip for chip a P4 with HT (800mhz FSB) will be faster than a Xeon with HT because Xeons are only 533mhz FSB.

But having said that the new P4s with the PRESCOTT architecture is not as fast as its predecessor NORTHWOOD core.

Best bet is to either get one of the P4c (northwood cores) or go dual XEON, which is when they really work best. "2 heads are better than 1"

Bobt
04-01-2004, 12:45 PM
This link is the answer to the PIV 64bit bus issue.
BUT no one knows if it works or not..
CANT wait. If it does I am throwing my curent
motherboard into a different machine and using this one!!
Check it out

http://www.supermicro.com/products/motherboard/P4/875/P4SCT+II.cfm

PIZAZZ
04-01-2004, 01:22 PM
That looks Yummy!!

I hope to have one in by middle of next week to test out. Could be the solution to solve world hunger AND solve the mystery of the "Magic Bullet".

I will keep everyone posted.

Yeah!

paulk
04-08-2004, 02:47 PM
I hope to have one in by middle of next week to test out. Could be the solution to solve world hunger AND solve the mystery of the "Magic Bullet".

Anything to report, Pizazz?

PIZAZZ
04-08-2004, 03:19 PM
Still waiting on delivery. Our distributors have had a hard time getting them in stock just yet.

Will report details as soon as we get a chance to test everything out.

Dan Hong
04-14-2004, 09:01 AM
We've been building systems with P4 800 fsb boards for a while now. The current fave is P4 3.2 Extreme ($$$) on the the Gigabyte 8KNXP Mobo. Built with the onboard SATA. We're getting 145 mb/sec out of a set of four 250gig WD discs. This is my current traveling road system, I love it!

paulk
04-14-2004, 11:28 AM
Dan,

What about the number of video streams you can process with your P4EE systems? I'm thinking if it's good enough to take on location, it's got to be able to handle a decent amount of video at one time.

paulk
04-28-2004, 02:36 PM
Still waiting on delivery. Our distributors have had a hard time getting them in stock just yet.

Pizazz, is the P4SCT+II motherboard still backordered?

Kevin
04-29-2004, 06:30 PM
Pizazz, if you have not actually ordered (and paid for) the board yet, try Newegg.com They have the P4SCT+II in stock.

edhorton
07-01-2004, 04:11 PM
Just wondering if anyone has tried the P4SCT+II motherboard yet? Seems to be available. Also, has anyone tried the Intel 7505VB2 motherboard as well?

Thanks,
Ed Horton

PIZAZZ
07-01-2004, 04:23 PM
Originally posted by edhorton
Just wondering if anyone has tried the P4SCT+II motherboard yet? Seems to be available. Also, has anyone tried the Intel 7505VB2 motherboard as well?

Thanks,
Ed Horton

Yes Ed,

We are giving it a rundown right now. Nothing spectacular to report as of yet. It went together with no problems and seems happy with the VT card in it.

I will be able to give more details after the weekend.

RuiFeliciano
07-09-2004, 06:04 AM
Any news on this Jef? :D

PIZAZZ
07-09-2004, 10:23 AM
We have been trying some things and no definite results to report. The onboard SATA controller with 2 drives is doing the expected 110mb/s. (2x250gb Seagate Drives) We can do 2 Uncompressed SD streams pretty reliably. We just put a 3ware 8 port IDE card in the system and are waiting to populate it with a few more drives to see what it will do.

I also will be trying out a 39320 to see what 4 U320 drives will do performance wise. Waiting on a 39320 from my distributor right now.

I will update everyone when I have more info.

Paul_Newman
11-19-2004, 01:54 AM
We have been trying some things and no definite results to report. The onboard SATA controller with 2 drives is doing the expected 110mb/s. (2x250gb Seagate Drives) We can do 2 Uncompressed SD streams pretty reliably. We just put a 3ware 8 port IDE card in the system and are waiting to populate it with a few more drives to see what it will do.

I will update everyone when I have more info.

Any results from this test with the 3Ware card on this mobo?

Are the PCI-X slots on this mobo (Supermicro P4SCT+II) on an seperate bus? Or would they clash with a Toaster sitting in one of the a 32bit PCI slots?

Paul_Newman
11-20-2004, 11:49 PM
Pizazz, you out there?

PIZAZZ
11-21-2004, 12:34 PM
Pizazz, you out there?


You Rang???

Sorry been on the road a bit lately.

It does seem like the 64bit bus and the 32bit slots are separate. We still do not see any real advantage of going with this motherboard over one of the new dual Xeon systems with PCIExpress. I am putting the 64bit IDE 3ware controller up on the shelf unless someone wants to buy another one. We just have no use for it. SATA is much better for our applications.

Paul_Newman
11-21-2004, 01:52 PM
It does seem like the 64bit bus and the 32bit slots are separate. We still do not see any real advantage of going with this motherboard over one of the new dual Xeon systems with PCIExpress. I am putting the 64bit IDE 3ware controller up on the shelf unless someone wants to buy another one. We just have no use for it. SATA is much better for our applications.

Thanks for the update. The reasons I'm interested in this mobo is (1) that I do not want to invest into a new RAID controller right now - I feel I can still get lots of mileage out of the one I have and (2) that a powerful single P4 setup will be more affordable for me right now than a dual Xeon or Opteron and (3) give me an ATX board to fit my very ample chassis without requiring custom mods to accomodate special server boards.

My application is post production editing, not live switching. So my setup does not have to be cutting edge to serve my client's needs.

The other option I have is finding a good single Opteron board with a PCI-X slot on board. Not sure if there's anything like that. Or getting a dual Opteron mobo with only 1 CPU initially.