Xeon PIII vs. Pentium III

[Guest]

Okay, both the Xeon III and PIII (in the 733MHZ/256K cache varieties) have the "70 unique instructions, called Streaming SIMD Extensions", 256K of full speed cache (L2) and multiprocessing capabilities.

I know the Xeon handles more memory, and their are larger cache versions (750Mhz version is available in 512K,1MB & 2MB). But if both processors will be used with the same components and chipset (440GX): 768MB RAM, G400, Windows 2000, Photoshop 6.0 and RAID controller, what will the performance gain be on the Xeon IIIvs the regular PIII?

What makes the Xeon a better processor? Even Intel's site does not do a fair (they compare PIII dual and Xeon single) SPEC2000 comparaison of these processors.

Your thoughts?

 

[Crashman]

It's all about the cache. I believe the only reason low-cache Xeons cost more is because they fit Xeon motherboards, which are normally reserved for high-end workstations and low-end servers, a market that is normally more willing to support expensive hardware. But additional cache is very usefull, as I experienced a 50% performance gain when I went from a K6-2 400 to a K-6 III 450. What this meant was that I could crash even faster!

 

[Guest]

I think that a single user running Photoshop would be better off with a PIII than a Xeon. The extra cost of the Xeon is not worth it for a 10% increase in speed (maybe more for the really huge cache, but that adds a couple thousand dollars), unless you have just run out of other places to spend money on your system But things like a faster motherboard, faster or more RAM, RAID, etc. would net more of a benefit than the Xeon.

--I'm glad it's not just me that HATES Intel's own benchmark displays. I went there to see how various new processors compared to my PII 450, and of course, there was no record of previous processors. Only 800, 850, 933, and 1000. I love how they don't even compare their PIV to their PIII - what use is a graph comparing a PIV 1400 with a PIV 1500 ?! Now I know that those old benchmarks still exist somewhare on the intel site, but trying to find them you'd think that they don't want you to do an informed comparison

 

[Guest]

Xeons are the only processor Intel will allow a quad or more cpu setup. Using filters on Photoshop would be fun on SMP setup.

 

[Guest]

FYI, I have a dual processor setup and it in no way improves the performance of filter calculations in Photoshop.

Photoshop (and about 99% of all programs) is written to take advantage of only 1 CPU.

Even 3D animation programs such as Maya, Softimage, and 3D Studio Max do not take advantage of multiple CPUs for most of there floating point calculations. The only benefit is felt when rendering a final image with these programs.

If Intel is comparing dual CPU PIII systems (256k cache) performance with single Xeon PIII systems with a larger cache then they are making an unfair comparison since most benchmarks will only use a single processor anyway. So that comparison equals the same thing as a single PIII vs. a single Xeon PIII.

The greatest benefit I have seen with dual CPUs besides rendering 3D scenes is the unmatched stability of such systems, especially running Windows 2000. My system is as close to crash proof as you can ever hope for. The reason for this is when most systems would lock up when a process goes haywire (because it takes over the one and only CPU) a dual system still has a free and unaffected CPU. That leaves the user with enough resources to be able to manually shut down the problem process without ever neading a reboot the system.

<b><i>MK</i></b>

 

[FUGGER]

The 2 previous posts are incorrect.

You can install SBC cards onto a backplane, I have seen up to 16 SBC backplanes. SBC = single board computer = another way to build MP machines typicly rack mounted.

http://www.lannerinc.com/ap-psl19b.htm

ALR (Advenced Logic Research) also experimented with multiple CPU cards, 4 CPU's If I remember correctly, also duaghter cache cards.

Photoshop DOES take advantage of multiple CPU's being a threaded application.

http://www.adobe.com/support/techdocs/12dde.htm

Try this simple test, create a 600dpi 5"x5" image and rotate it 41 degrees.
G4 apple owns this test in the workstation class machines
P4 can perform this test equily but at a much higher Mhz
xeon is very close but will lose
Athlon is dismal along with P3

 

[Guest]

This is true, as I do have a Dual 550 PIII system right now. It's pretty quick, I have a opportunity to upgrade to a dual 550 PIII Xeon (with the same size cache as my current dual PIII setup) and I am wondering if it is worth it.

As for this test, I took a 320 dpi, 24bbp 5.5"x5.5" image and it rotates it 41degrees (clockwise or counter clockwise) in about 1.1 seconds. Certianly not painful. Is there something I am missing in this test?

Thanks

PS I am still not sure that the PIII 866/256K/133MHZ is slower than a PIIIXeon 866/356K/133MHZ. Even using dual processers. Any reason to belive otherwise?

Thanks, everyone!

Armand

 

[Guest]

FUGGER- Thank you for miss-quoting me!

I did not say Photoshop performance on dual processors is bad, I said Photoshop only utilizes one CPU, and that IS correct. This may not be true on a Macintosh, but I have not tested it on a MAC. It is true on a PC though. If you do not believe me run Photoshop on a dual CPU Windows NT4 or Windows 2000 system and view the task manager while doing anything in Photoshop. You will see that no more than 50% of CPU resources is ever in use.

Also you refer to 'rotating' a canvas. Yet I said Photoshop 'Filters', which are a far better test than canvas rotations. Rotating a canvas is hardly any calculation at all even at high resolutions. It is obvious that neither you, nor the last guy knows anything about graphics, because if you did you would never make such a stupid statement.

Maybe you should stick to something you know FUGGER, like talking trash the way you did about Tom a while back.

<b><i>MK</i></b>

 

[Guest]

Also, here is a good question for you heavy Photoshop users out there.

What do you do more often rotate your canvas to odd angles (NOT using the crop marquee), or use filters?

<b><i>MK</i></b>

 

[Guest]

I said Photoshop only utilizes one CPU, and that IS correct

Photoshop is a poor multiprocessor application, but some functions do use both processors. I just put this to the test on a dual PII 450 running NT4sp5, and both the Resample and the default Unsharp filter do use both processors. It does not use both nearly as fully as it should though

 

[FUGGER]

Microwave King... WTF is wrong with you...
Photoshop is a threaded application, and does take advantage of both CPU's

Rotating canvas is pure CPU intensive untill memory is filled then disk swapping takes place. This function is not FSB or bandwidth intensive. Rotate canvas is not a filter.

Microwave King, your surely are king dumbass in this area.
Maybe you should learn what a "threaded application" is and how it relates to SMP. until then STFU. try reading link I posted that might help some.

If you want to test my knowledge in the graphics arena, bring it on. I don't mind setting the record straight.

 

[Guest]

There you go FUGGER that's more like it! Now you're talking trash like we're all used to. : )

<b><i>MK</i></b>

 

[alphateam]

I don't want to get in the middle of your pis*ing match here, but photoshop is multitreaded. I know it has been since at least version 3. Here is a quote for FUGGERS link.

"All Photoshop features are faster on a multiprocessor system, although some can take greater advantage of the multiprocessor system's capabilities. "

 

[alphateam]

I don't want to get in the middle of your pis*ing match here, but photoshop is multitreaded. I know it has been since at least version 3. Here is a quote for FUGGERS link.

"All Photoshop features are faster on a multiprocessor system, although some can take greater advantage of the multiprocessor system's capabilities. "

 

[Guest]

I understand what you are saying, and I respect your opinion. One thing you need to realize though is that it is easy to say that a certain thing works in theory, but for it to actually work it must stand up to benchmark scrutiny.

I have not just used some obscure tests like FUGGER has, I have used the program extensively for print work. In some cases Photoshop will use more than 50% of system resources on a dual CPU system, but in most cases it does not.

So Adobe, or anyone else may claim the software is written to take advantage of multiple CPUs, and it may even be true, but the fact remains that if it was written that way it was not done very well. Even FUGGER's test at high resolutions will not utilize 100% of system resources.

You know it's kind of like saying that a Pentium 4 is all this and all that and it's faster than any other processor, but when the benchmarks show different it kinda makes you wonder. Even though we know most software has yet to be revised to properly take advantage of the SSE2 instruction set we also know that most of it will not be for a long time to come.

So please don't tell me how the software is written, because if that is a viable multithreaded application, it is a crippled one at best.

<b><i>MK</i></b>

 

[FUGGER]

the arbitrary rotate is not a benchmark nor a stress test. It just forces a load on the CPU.

I can't really justify using xeon for photoshop even though is performs decently. If your gonna spend xeon kinda bucks your better off with a O2 R12k-A or the QED RM5200 CPU
The UMA archetecture is better suited for these tasks.

Xeon is better as a database server, back office server, web server, handling lots of requests.

 

[jg38141]

Wow, A FUGGER post with no flame!!! I wonder if someone else used his computer when his back was turned. Is this a new FUGGER? Stay Tuned!!!!

"Are you saying that I can dodge bullets?"

 

[FUGGER]

Ducar, increase the res on the image or image size to increase the load on the CPU. run the same rotate on both machines. try not to create file too big or it will swap to disk.
Xeon will pull ahead as file size increases.

Xeon can also access larger amount of RAM on system board. so if your working with huge files 1GB + then xeon might be better choice for you.

Or you can just test filters =P but those do require more FPU than CPU calculations.

 

[Guest]

Wow, what a heated disscussion.
The file I was useing for the rotate test was only a little over 8MB, I will try it again with one of my typical files (~30MB for 24bit, 56MB for 36bit) tonight.

Here is what I am doing: Photo Restoration/Digital Photography (http://www.re-co.com).

Currently I use a Dual PIII 550Mhz Supermicro P6DGU, with the adaptec RAIDPORT 1130 card. My main drive is four 9.1GB drives configured for RAID 5 and I have two 2.1GB drives as a stripe set for my both my Win2K page file and the photoshop swapfile. I also have 768MB of PC100 registered SDRAM. Three monitors are driven by a(2x) AGP Matrox G400 Dual Head and a single PCI MGA G200.

The system is pretty quick right now for my primary needs (Photoshop) (Average 8x10 photo is 28.8MB @ 320dpi,24bpp)

I have been using dual processors on my Photoshop systems for years now, and it does make more of a difference when using filters, but not so much the regular (save,open,cut,clone, etc) commands. To be honest I don't use filters much in my work (it's mostly cloning and a lot of color/channel corrections.) Photoshop does support dual processors, but I will leave it up to debate about how well it does so! I can only say it does it well enough that once I switched from a single P120 to Dual P120's years ago, I never wanted a single processor photoshop system again.

I can get a Dual Xeon version of my current motherboard for ~$500 and the Xeon processors are free (removed from server upgrades) So the only thing that will change in my system will be from PIII to PIII Xeon... (of the same speed/cache/FSB) Is it worth the $500?

Right now I am leaning toward getting the board just to compare and get some good data! Hopefully I wont be dissapointed, but it just seems that the Xeon has nothing over the PIII of the same speed/bus/cache, other than using 4-8 processors on a server. Which makes me wonder why Intel expects to sell a dual Xeon vs. a dual PIII?

Is there more than a marketing reason for the Xeon in a Dual CPU (workstation)?

Thanks

-Armand