woodcrest 5150 vs opteron 285

[duploxxx]

on my desk

screenshots of cpuid
http://users.telenet.be/duploxxx/toy...es/285cpuz.jpg
http://users.telenet.be/duploxxx/toy...wood32cpuz.jpg

opteron is a 280 oc'ed to 2624 with nforce4 ultra mobo and 4gb ram cas3 ecc and 250gb/7200tr hd

woodcrest is a 5150 with intel chipset and 4gb fbdimm 667mhz cas5, a 76gb sas 15k hd and for 64bit a 76gb sata 10k hd for 32bit


cine 2003 32 bit opteron 285
http://users.telenet.be/duploxxx/toy...5cine32bit.jpg
373 single cpu
1202 multi cpu

cine 2003 32bit woodcrest 5150
http://users.telenet.be/duploxxx/toy...woodcine32.jpg
431 single cpu
1305 multi cpu

cine 2003 64 bit opteron 285
http://users.telenet.be/duploxxx/toy.../285cine64.jpg
438 single cpu
1415 multi cpu

cine 2003 64bit woodcrest 5150
http://users.telenet.be/duploxxx/toy...dcine64bit.jpg
459 single cpu
1472 multi cpu

povray 32bit single thread benchmark image opteron 285
http://users.telenet.be/duploxxx/toy...bit2thread.jpg
11min04sec

povray 32bit single thread benchmark image woodcrest 5150
http://users.telenet.be/duploxxx/toy...ovsingle32.jpg
7min21sec

povray 32bit 4thread benchmark image opteron 285 average
http://users.telenet.be/duploxxx/toy...it_4thread.jpg
11min13sec

povray 32bit 4thread benchmark image woodcrest 5150 average
http://users.telenet.be/duploxxx/toy...pov4thread.jpg
7min14sec

povray 64bit single thread benchmark image opteron 285
http://users.telenet.be/duploxxx/toy...ov64single.jpg
8min32sec

povray 64bit single thread benchmark image woodcrest 5150
http://users.telenet.be/duploxxx/toy...bit_single.jpg
9min35sec

povray 64bit 4thread benchmark image opteron 285 average
http://users.telenet.be/duploxxx/toy...bit4thread.jpg
9min16sec

povray 64bit 4thread benchmark image woodcrest 5150 average
http://users.telenet.be/duploxxx/toy...bit4thread.jpg
9min28sec

woodcrest is faster then equal clocked socket 940 opteron. perfromance difference get smaller when using 64bit. povray 64bit seems to have an issue with the woodcrest, its not a driver issue since the cine is working fine

layout of the woodcrest system:
woodcrest has an active cooling and is rather cold, the speedsteps comes in directly "agressieve" from 2660 to 19xx, unable to tern off) The fbdimm is active cooled and you are not able to touch it for a few seconds

more test will follow when i have some spare time

duploxxx

[Andrewv]

woodcrest looks good ... very good

what about the temp ?

[duploxxx]

update on coretemps and 3dmark06 score

wood idle and load with actieve cooling!
http://users.telenet.be/duploxxx/toy...odidletemp.jpg
http://users.telenet.be/duploxxx/toy...odloadtemp.jpg

3dmark run for opteron 285 (3563) and woodcrest 5150 (3788)
http://users.telenet.be/duploxxx/toy...rk06_32bit.jpg
http://users.telenet.be/duploxxx/toy...rk06_32bit.jpg[/QUOTE]

[M1nDH4cK]

looking good. Can you do some spi benches ?

[Silver Bullet]

Results look pretty good, but 64bit is going to be more and more important .. intel should really make sure they don't fall down further.

[couppi]

A lot closer than I would have though, especially in the 64 bit tests. Intel should watch out...

[Carfax]

I think the discrepancies in 64-bit are probably due to compiler inefficiencies when generating code for Conroe.

Intel's EM64T compiler was made using the P4, and we all know that the P4 is an ugly bird to code for.

[mesyn191]

I think the discrepancies in 64-bit are probably due to compiler inefficiencies when generating code for Conroe.

Intel's EM64T compiler was made using the P4, and we all know that the P4 is an ugly bird to code for.

If it was that bad then the A64 would have bad performance too with apps compiled by it, seems to do pretty well to me...

As for the test results they look pretty close for a single socket system, with duals I bet it'd be a dead tie or favor the Opteron some, in a quad socket Opteron will probably win hands down. Good news for AMD I guess, there is at least 1 market segment they won't lose in (they may in fact continue to gain...) so financially they should be better off than last time around vs. a new Intel arch. (ie. AXP vs P4C days...).

[uOpt]

Yeah, this is probably a case of running code overoptimized for Netburst.

I see no trace of Conroe being any less good in 64 bit mode using generic compiler options not targetted at a specific CPU.

[Carfax]

If it was that bad then the A64 would have bad performance too with apps compiled by it, seems to do pretty well to me.....).

Why is that? The A64 has it's own specially optimized compilers, while Conroe as of yet, has none.

When the compilers are optimized for Conroe, we'll see a large increase in performance.

[uOpt]

Why is that? The A64 has it's own specially optimized compilers, while Conroe as of yet, has none.

When the compilers are optimized for Conroe, we'll see a large increase in performance.

The whole point you are missing is that AMD64 is not that picky about the code optimization. It runs pretty generic code with no CPU-specific optimizations very well. Also, there don't seem to be many compiler which even claim to have specific AMD64 targets.

From what I have seen so far Core2 is similar to AMD64 in that you don't have to go a long way of CPU-specific compiler and optimization modes.

Only Netburst was a to deal with in that respect, I am glad this is over.

[alucasa]

Yep, I agree. I've seen many programs *optimized for Intel CPU*, but haven't really seen "optimized for AMD".

AMD never really stood on adventage and has come this far. Gotta admit that they've done fine jobs so far, that was until Conroe hits the public.

The whole point you are missing is that AMD64 is not that picky about the code optimization. It runs pretty generic code with no CPU-specific optimizations very well. Also, there don't seem to be many compiler which even claim to have specific AMD64 targets.

From what I have seen so far Core2 is similar to AMD64 in that you don't have to go a long way of CPU-specific compiler and optimization modes.

Only Netburst was a to deal with in that respect, I am glad this is over.

[mesyn191]

Why is that? The A64 has it's own specially optimized compilers, while Conroe as of yet, has none.

When the compilers are optimized for Conroe, we'll see a large increase in performance.

Architectually Conroe and A64 are very similar.

There is only 1 compiler that I know of that was designed with AMD64 in mind and almost no one uses it, in fact AMD recommends people to use Intel's compiler instead...

Compilers have limited performance benefits outside of corner cases anyways, thier main purpose is to make it easier to develop programs, not to extract as much performance as is possible out of a given processor. If you're expecting miracles or even anything more than some very minor performance gains then dream on...

[Lightman]

duploxxx could you re-run cinebench tests for opties on x32 windows, because cine 2003 have different (smaller) preformance that cine 9.5 used for conroe!
THX!