Creating a Woodcrest Workstation, advice needed

[Dante_Jose_Cuervo]

Yes woodcrest is running almost a full Gigahertz below Tulsa but the current benchmarks show that a single Xeon at 3GHz is faster than two of the previous generation Xeons at 3.6GHz. The woodcrests don't use the old Netburst architecture and thus now do the same task in less time. It's all about efficiency now. If you have anymore questions I'd be glad to answer them! Hope this helps!

 

[crow_smiling]

Compared to the upcoming Xeon Tulsa, Woodcrest's clock speeds are relatively low. What is the reasoning behind this? Is Tulsa not using the Core2 architecture? Or is Tulsa simply going to be faster?

Tulsa is not a Core 2 processor but a Netburst one, hence the high clock speeds.

Right now i'm running a single-core pentium 4 at 1066MHz FSB and 3.73GHz. Is a dual-core xeon at 1333MHz FSB and only 2.33GHz going to be faster/more efficient than my current CPU? I'm worried about that clock speed being so low. i realize it's dual core, so it's really like having 2 Xeons, combined with the fact that there will be 2 chips, so 4 cores, but i've heard many applications cannot take advantage of dual cpu's, or dual cores for that matter. any thruth to that?

Your P4 3.73 is going to perform roughly the same as one of the cores in a Woodcrest 2.33. You really need to research which of your applications can take advantage of 4 cores, otherwise you might be wasting your money on Woodcrest. As I pointed out previously, you may be better off with a top of the range Conroe instead, which will over-clock nicely. Also, you might have trouble finding a Woodcrest board that over-clocks, so an X6800 @ 4 GHz makes more sense to me than dual Xeon’s @ 2.33. With the X6800 or E6700 for that matter, applications that can only use 1 or 2 cores will run MUCH faster than on the Woodcrest system. Even the few applications that can efficiently use 4 cores won’t run much faster on the Woodcrest i.e. 4 * 2.33 = 9.13 versus 2 * 4 = 8. That’s a less than 15% maximum gain and that assumes that the application scales perfectly for 4 cores which isn’t going to be the case. So you gain very little in the best case scenario, but in the more typical scenarios of applications that can only use 2 cores you end up with
2 * 2.33 = 4.66 versus 2 * 4 = 8. That’s over a 70% gain for Conroe. For single core applications you get a 57% gain for Conroe.
This is only looking at clock speed and not the overall platform performance and there will be cases where Woodcrest will gain at the platform level due to the quad memory architecture. But, this will still not make up for the deficit in clock speed.

Most games still only use 1 core and even those that can use 2 don’t scale well with multiple cores in the way that many Content Creation applications do. X6800 is the obvious choice for games.

Read this article which looks in detail at how application performance scales with up to 4 cores. It’s looking at older Xeon’s and Opterons, but it will still give you an idea on how scaling works.

It would be a shame to spend the extra money on Woodcrest to find that a Conroe X6800 system would typically give you much better performance and at a lower price. There’s something very compelling about the idea of a DP system, but sometimes the reality doesn’t add up.

 

[Logicsequence]

Your P4 3.73 is going to perform roughly the same as one of the cores in a Woodcrest 2.33.

So at the very least it will run applications that can only use 1 core at the same speeds as my current system.

You really need to research which of your applications can take advantage of 4 cores, otherwise you might be wasting your money on Woodcrest.

Applications that i commonly use are:

Load Heavy Apps:
3D MAX 8
Lightwave 8.5
Maya 6
After Effects 6.5
Premiere Pro 1.5
Media Creator 8
Encore DVD 1.5
Photoshop CS
ImageReady CS
Paint Shop Pro X
Nero 7

Everyday Apps:
Internet browsers
Emule
Bittorrent
various video encoders
Microsoft office
various art programs
Windows media player
etc..

Games:
I play a lot of games, i'm a gamer, so this system will need to perform at least as well as my current system in games. I'm not a huge FPS gamer, but 90% of the games i play still make heavy use of 3D.
Games like:

Civilization 4
Galactic Civilizations 2
Homeworld 2
Nexus
SW battlefront2
SW empire at war
Earth 2160
etc...

I'm pretty sure i can benefit from a Dual Dual-Core Processor System, don't you think? I'm a 3D Digital Artist and graphic designer, i'm always rendering and encoding things, daily. It's my job, it's what i do. Not to mention this is going to be a heavy gaming PC too, there has to be a balance. Or more to the point, it has to kill at both. Basically it's needs to be a KILLER DCC machine, KILLER video/audio editing machine, KILLER gaming rig, my home PC, my home stereo, DVD authoring machine, etc.. etc... That's why i think i could benifit the most from Dual Dual-Core CPUs. Correct me if i'm wrong tho.

I made the following Spec sheet last nite of what i was thinking of, it's changed a bit since the first post. Items without a price are parts of my current system i intend to re-use. Again, please offer any advice or help (ANYONE):

[code:1:8dd6abbe04]
CPU1 Xeon Woodcrest 5140 2.33GHz 1333MHzFSB LGA771 $462.50
CPU2 Xeon Woodcrest 5140 2.33GHz 1333MHzFSB LGA771 $462.50
Mobo ASUS DSBF-D 2x DualCore Xeon LGA771 5000p $633.00
RAM Kingston 1GB DDR2 533MHz FB-DIMM $152.98
RAM Kingston 1GB DDR2 533MHz FB-DIMM $152.98
RAM Kingston 1GB DDR2 533MHz FB-DIMM $152.98
RAM Kingston 1GB DDR2 533MHz FB-DIMM $152.98
PSU Thermaltake ToughPower ATX/EPS 12V 650W $141.23
Sound Creative Labs SoundBlaster X-Fi 24-Bit PCI $109.99
HDD WD Raptor SATA 150GB (system) $225.45
HDD Maxtor 250GB SATA
HDD Maxtor 250GB IDE
HDD Maxtor 250GB IDE
HDD Maxtor 250GB IDE
Video ATi FireGL V7100 256MB GDDR3 PCIe
Optical SONY 16x ± R/RW
Optical HP 16x ± R/RW with LightScribe
Monitor Samsung 914v 19" LCD
Monitor Samsung 914v 19" LCD
KB/MOUSE Logitech MX700 Wireless Keyboard/Optical Mouse
Speaker Creative Labs 7.1 Surround Speakers
Printer Epson Stylus CX5400 printer/scanner
Tablet WACOM Graphire4 6x8 Pen Tablet

Total $2,646.59
[/code:1:8dd6abbe04]

Details on the motherboard are here.

 

[Totty]

Mobo ASUS DSBF-D 2x DualCore Xeon LGA771 5000p $633.00

Does that mobo allow OCing?

I'm going to buy two systems virtually identical to yours but I'm waiting to see whether OCable Woodcrest mobos are released first. I want to run 2.33 GHz WCs at 2.66GHz or a bit higher.

 

[Logicsequence]

Well you're not alone, I'm building one also. It looks really good overall but I do have a few suggestions.

1. For the mobo have you looked at anything from IWill? They have a bunch of really good mobos that you should look into for socket 771. They all have 1 PCI slot.
www.iwill.net

2. Yes you will need FB-DIMMS in a quad-channel config. That's if you plan on using both CPUs.

3. For the PCI-X it IS compatible with PCI but I wouldn't recommend it since if the speed of the card and the speed of the bus are mismatched what it will do is drop the speed down to the lowest factor between those two which could be 1MHz!

Other than that good luck with your new rig!

No i hadn't looked at Iwill, but i just did. The DPK66 and DPK66S look very nice actually. Are they priced anywhere yet?? The DPK66S is especially nice...

Have you found any Woodcrests for retail yet? What Motherboard are you planning on using?

 

[Logicsequence]

Mobo ASUS DSBF-D 2x DualCore Xeon LGA771 5000p $633.00

Does that mobo allow OCing?

I'm going to buy two systems virtually identical to yours but I'm waiting to see whether OCable Woodcrest mobos are released first. I want to run 2.33 GHz WCs at 2.66GHz or a bit higher.

I don't know, ASUS's website has sparse information. I plan on emailing them tomorrow and asking them if it supports overclocking. I would *imagine* it does, most all ASUS products do, BUT I'M NOT SURE. So that's why i'll email them.



Dante_Jose_Cuervo: Do the Iwill boards support overclocking??

 

[crow_smiling]

Your P4 3.73 is going to perform roughly the same as one of the cores in a Woodcrest 2.33.

So at the very least it will run applications that can only use 1 core at the same speeds as my current system..I think I underestimated the performance, the CPU itself should be faster for single core applications, but I don’t know how the FB-DIMM’s latency penalty affects overall performance.

You quoted a price for FB-DIMM 533 rather than 667; the 1333 FSB of Woodcrest can utilise the extra speed of the 667 part.

Applications that i commonly use are:

Load Heavy Apps:
3D MAX 8
Lightwave 8.5
Maya 6
After Effects 6.5
Premiere Pro 1.5
Media Creator 8
Encore DVD 1.5
Photoshop CS
ImageReady CS
Paint Shop Pro X
Nero 7

Did you read the XBitLabs review of DP systems that I linked to above? It shows how many of the above applications scale on DP dual core systems. Some of them scale better with two fast cores than 4 slower ones, which is what I’ve been getting at.

One definite advantage with quad cores is that you can assign two of them to transcode video whilst using the other two to create media at the same time.
But, if you look at the price differential when building a Woodcrest system over a high end Conroe system, you could actually build two Conroe systems for the same price. This assumes that the 2nd system would be a stripped down system and would be used as a render box so wouldn’t need a DVD drive, sound-card, VGA card. Some heavy duty rendering applications can even work on multiple machines as you probably know. This has it’s pros and cons; one pro is that you can leave the render box to do it’s thing and still have full access to your main machine for other uses.

 

[crow_smiling]

Mobo ASUS DSBF-D 2x DualCore Xeon LGA771 5000p $633.00

Does that mobo allow OCing?
I don't know, ASUS's website has sparse information. I plan on emailing them tomorrow and asking them if it supports overclocking. I would *imagine* it does, most all ASUS products do, BUT I'M NOT SURE. So that's why i'll email them.Try downloading the Manual for it from Asus’s website, that should give you all the details that you need.
Even if the board doesn’t support over-clocking in the BIOS, you may well be able to use Clockgen or CPUFSB to overclock it in software. You just need to determine what clock generator it uses and see whether either of these programs supports it. This will of course only be feasible if it has PCI/PCIe locks and suitable memory dividers.

 

[Totty]

Mobo ASUS DSBF-D 2x DualCore Xeon LGA771 5000p $633.00

Does that mobo allow OCing?
I don't know, ASUS's website has sparse information. I plan on emailing them tomorrow and asking them if it supports overclocking. I would *imagine* it does, most all ASUS products do, BUT I'M NOT SURE. So that's why i'll email them.Try downloading the Manual for it from Asus’s website, that should give you all the details that you need.
Even if the board doesn’t support over-clocking in the BIOS, you may well be able to use Clockgen or CPUFSB to overclock it in software. You just need to determine what clock generator it uses and see whether either of these programs supports it. This will of course only be feasible if it has PCI/PCIe locks and suitable memory dividers.

There's nothing in the bios manual about changing the FSB, only the divider. But then again bios manuals are notoriously poorly written

 

[BaronMatrix]

Compared to the upcoming Xeon Tulsa, Woodcrest's clockspeeds are relatively low. What is the reasoning behind this? Is Tulsa not using the Core2 architecture? Or is Tulsa simply going to be faster?

Right now i'm running a single-core pentium 4 at 1066MHz FSB and 3.73GHz. Is a dual-core xeon at 1333MHz FSB and only 2.33GHz going to be faster/more efficient than my current CPU? I'm worried about that clock speed being so low. i realize it's dual core, so it's really like having 2 Xeons, combined with the fact that there will be 2 chips, so 4 cores, but i've heard many applications cannot take advantage of dual cpu's, or dual cores for that matter. any thruth to that?

No Tulsa isn't Core 2. It's the last of the NetBust chips with 16MB of cache.

 

[Dante_Jose_Cuervo]

The mobo that I'm looking at is the DKP66SCSI because I still find that SCSI's are the kings of I/O.

About the overclocking... last time I checked they don't. I'm not sure about these though, but with the performance on these Xeons you might not even need to OC.

About the multi-threading. Most of those programs you're using do support multi-threading and should benefit from the dual dual-core system.

 

[crow_smiling]

About the multi-threading. Most of those programs you're using do support multi-threading and should benefit from the dual dual-core system.

Read this article as it shows that whilst most content creation type applications will benefit from dual cores, far fewer of them benefit from the move to 4 cores. That’s why at the moment I suggest than an over-clocked X6800 will perform better than dual Woodcrest 2.33 at stock speed for most applications. This situation will probably change as more of these applications are optimized for more than 2 cores.
Personally, I would save myself some money and buy an X6800 or even an E6700 now, knowing that by the time Woodcrest is released on 45nm there will be better motherboard support (hopefully with over-clocking), 64 bit applications will be released for Vista and hopefully with more support for 4 cores and quad cores will also be cheaper then.

 

[Logicsequence]

well i don't have the money to go upgrading my computer ever 6 months, this will have to last me at least 2 years. U seem to have a mean aversion to Dual Dual-Core systems, whats your beef?

 

[MadHacker]

...
2. PCI-X is a 64-bit-wide slot operating at 66, 100, or 133 Mhz that looks nothing like and is not compatible in the least with standard PCI 32-bit 33 MHz slots.
...

Completly untrue...
PCI-X is PCI 32 bit backwards compatible.
any PCI card will plug into a PCI-X slot as well as any PCI-X card will plug into a PCI slot.

this doesn't mean you will get it running in optimal configuration... but it will work.

PCI-X will slow down to 33Mhz to support PCI 2.1 spec.

 

[Dante_Jose_Cuervo]

He's probably saying that because it's true that a lot of programs won't support that many CPUs. However if I were you I'd totally go for it. If you like the idea of the 5148 at a LV config then more power to you.

 

[Logicsequence]

He's probably saying that because it's true that a lot of programs won't support that many CPUs. However if I were you I'd totally go for it. If you like the idea of the 5148 at a LV config then more power to you.

It's not as if i'm saying gung ho! i'm gonna go for it b/c i can! The applications i use commonly, esp 3d rendering software and video editing software, will really see a boost from a multi core system, just as they would have a year and a half ago be4 dual cores, on a dual processor system. not to mention that, and correct me if i'm wrong, this system should go toe to toe with most any gaming rig out there (short of spending $10,000 on a quad SLI machine) and KICK @SS.

And i don't have the money to upgrade again and again every few months. this will have to last at least 2 years. And with vista upcoming, and 64bit apps upcoming, and more and more games and/or programs beginning to be able to use dual or multi cores, shouldn't i get my money in for a multi core system now? rather than have to wait another 2 years? I mean am i missing something here? is this system going to suck at gaming? or at DCC? or at everyday computing tasks?

Even at the worst case, that a game or program can only use one core on one processor, won't a single 2.33GHz Woodcrest core running at 1333MHz fsb with the new core2 architecture perform AT LEAST as well as a single core of a pentium 4 660, or a pentium D 960? And i've got to beleive the woodcrest is AT LEAST as good as it's conroe counterpart of the same clock speed. After all, the woodcrest is a suped up version of the conroe. And lets be honest here, with the advent of hyperthreading, which is basically a logical second core, many programs began to be able to use 2 cores.

In the same vein, if i wait 6 months for a clovertown, that's a total of 8 cores. If a program can't use 4 it certainly can't use 8. And a clovertown will probably cost more than a woodcrest at launch. Not to mention clovertown clock speeds will be even lower than woodcrests. And also, after conroe's release, and then clovertown, the next processor will be tigerton, and that's going to use CSI bus, which means a whole host of new more expensive motherboards and maybe RAM...

I mean please, by all means, someone correct me here if i'm off in left field and totally off base.. Am i?

 

[crow_smiling]

A Woodcrest system will get better over time for you as more of your applications get optimized for more than 2 cores.
I wouldn’t buy Woodcrest now because I don’t think its good value. An O/C Conroe will beat your Woodcrest’s @ 2.33 by a decent margin in most cases. It’s not just that’s its cheaper, it’s the fact that if you buy Woodcrest now you are paying the hefty early adopter premiums for the motherboards and the RAM. By next year there should be a better range of motherboards for Woodcrest at better prices and with O/C’ing and with more chance of them supporting 45nm Xeons. FB-DIMMS should be cheaper and hopefully more software will be optimized for 4+ cores. I’m going to wait until more 64 bit applications are released and see how they run under Vista on AMD and Intel platforms before looking at DP systems again.

I don’t think that you’ll lose money if you buy Conroe now and Woodcrest next year. You just sell your Conroe CPU/motherboard/RAM and the money that you lose on those will be offset by savings that you make on the Woodcrest parts. Of course, this is theoretical and could end up going more for or against you.

I’m not saying that buying Woodcrest now is a bad idea; I just don’t think it’s a black and white decision in your particular case. If I were a Media Creation Professional then I would just go and buy dual Xeon 3.0s as for those people the cost premium is not really relevant as time is money to them. But money is an issue for you, which is why I suggested that it’s worth at least CONSIDERING Conroe as an alternative.

Enough already. 8O

 

[Logicsequence]

anyone else like to offer advice or comment?

 

[yourmothersanastronaut]

I agree with Crow. Go with a Core 2 Duo until Woodcrest platforms are more mature, at which point you can upgrade to a Woodcrest system. I feel that you'll find that four cores versus two cores won't justify the price premium at this point in time, and you'll be so busy "workstationing" it up that you won't notice you "only" have two cores.

 

[Logicsequence]

I agree with Crow. Go with a Core 2 Duo until Woodcrest platforms are more mature, at which point you can upgrade to a Woodcrest system. I feel that you'll find that four cores versus two cores won't justify the price premium at this point in time, and you'll be so busy "workstationing" it up that you won't notice you "only" have two cores.

and i say again i don't have the money to upgrade again in 6 months or a year.

 

[yourmothersanastronaut]

cough...cough...sell old parts...cough...

 

[Logicsequence]

cough...cough...sell old parts...cough...

and i say again i don't have the money to upgrade again in 6 months or a year.

 

[yourmothersanastronaut]

Who said you have to build again in six months? The beauty of the DIY system is that you upgrade when you can upgrade. Go ahead and say that you can't upgrade in 6 months. You don't have to. A little time for the platform to mature wouldn't be a bad thing either, considering the fact that your work will reside on this machine.

 

[allhell]

I am 99.99% sure that PSU will die on you soon. Just not big enough to handle all that & I'm sure you gonna add more stuff to it.

 

[Logicsequence]

following that line of reasoning, if i were to upgrade in 6 months or a year (which i wouldn't be able to get the money to do), WHY would i get a woodcrest then??? Why wouldn't i get a Clovertown, which is a quad core woodcrest, since they'll be out by then? It would make no sense for me to get a woodcrest when clovertown would be available.