Spec VPR Results for various processors...

Ok, so the AMD 64 bit machines seem to outrun their 32 bit machines and even outrun the fastest Intel 32 bit machines, but this is while running

32 bit programs. The same AMD 64 bit machines run slower when running the same program in 64 bit mode.

Hmmm... I'm not sure what this means in terms of the future of AMD 64 bit processing. I guess place and route is not the only engineering application, so the level of adoption of the AMD 64 bit architecture depends on a lot of other apps as well.

What I do know is that the Itanium has been a big flop. It will be interesting to see what Intel's new 64 bit (32 bit compatible) machine will be like. But I'm not sure I want a CPU that clocks at frequencies higher than my cordless phone uses ;)

--

Rick "rickman" Collins

rick.collins@XYarius.com
Ignore the reply address. To email me use the above address with the XY
removed.

Arius - A Signal Processing Solutions Company
Specializing in DSP and FPGA design      URL http://www.arius.com
4 King Ave                               301-682-7772 Voice
Frederick, MD 21701-3110                 301-682-7666 FAX

WILD A** guess says this probably comes down to the faster memory heirarchy in the AMD 64 bit processors. Since the Athlon 64 has the memory controller on the CPU, directly connecting to the DIMMs, memory latency is something like 15 processor cycles + DRAM access, which is amazingly low, saving 2 chip crossings and a boatload of chipset overhead.

--
Nicholas C. Weaver                                 nweaver@cs.berkeley.edu

Hi Kenneth,

Looking through the SPEC results again, I think I may have misread them the first time I tried to mine through for this data. It's possible I compared

32-bit and 64-bit results on two machines that had different speed processors, which would have been pretty silly. Here's what I found this time:

Under Spec CINT2000 results, you'll find 8 results listed as AMD "A4800". The third entry in the list appears to be a 1.8 Ghz Opteron 144 w/ 2GB RAM running SuSE 8 and is compiled using gcc targeting 64-bit instruction set (I believe that's what SPEC_CPU2000_LP64 means but am not positive). The fourth entry appears to be the same machine and OS except with code targeting 32-bit instructions. This is the only example of 32-bit vs.

64-bit I could find in this listing.

Overall SPEC Results (higher better)

32-bit 1045 64-bit 980 (-6%)

175.VPR results (higher better)

32-bit 865 64-bit 957 (+10%)

300.Twolf results (another p&r tool):

32-bit 1133 64-bit 997 (-12%)

So what do I think now? I think results on one computer on a couple tools can't tell us the whole story and that we need more data for a meaningful conclusion. But I'll stick to my original conclusion that 64-bit won't be a silver bullet for speed -- it may give a small boost or may not.

Regards,

Paul

The difference here would be one of cache and memory system.

Forget clock rating - unless its the same processor, it doesn't really mean anything.

--
	Sander

+++ Out of cheese error +++

It is not just that. The new AMD K8 core (used in AThlon 64 and Opterons) has a number of advantages over the old K7 core (used in Athlon XP, etc). you can read more on this on say

--
	Sander

+++ Out of cheese error +++

On Paul Leventis' additional info that the TWOLF Benchmark was also a P&R app I went back and nabbed those results as well for just the 64bit vs.

32bit AMD cases:

2.0 GHz Opeteron time 186, score 1611

2.2 GHz Athlon XP (32bit) time 304, score 929 2.0 GHz Athlon 64 time 204, score 1468

Same story as VPR. New 64 bit processors trounce 32 bit processor even with higher clock.

Ken