AMD ClawHammer 64-bit 3400+ VS other processors
Look at this graph found on this <a href="http://www.zive.cz/H/PCtuning/AR.asp?ARI=103690&CHID=4&EXPS=&EXPA="; target="_blank">page</a>. These numbers are real-world tests or projective results about raw power of processors in 32-bit integer numbers calculation (SpecInt).
Remember that SpecInt has no relation with floating-point operations (SpecFP) nor SIMD instructions (Motorola Altivec, Intel SSE...). The three are different execution units.
So there is this AMD 64-bit 3400+, which seems very clever in integer operations. It is targeted for early 2003.
But what is this IBM PowerPC "4" @ 1.3 GHz ?
- Is it the Power4 ? I can't remember if the Power4 goes beyond 1 GHz, but I thought that its raw power was beyond all other processors, especially versus Pentium IV. Way beyond...
- Is it a G4 or G4+ ? I thought IBM didn't produce these chips, only Motorola ? And 1.3 GHz ? Today it seems too high, and by early or mid 2003 it seems too low...)
Note that if it is a 1.3 Ghz G4 (Apollo ?) a 2 GHz Pentium IV is less powerful (today), but a 3 GHz Pentium IV is better (early 2003)...
And I saw a 3.6 GHz (!) overclocked PIV yesterday
Hope the G5 will crush all these chips.
[ 01-22-2002: Message edited by: Prim ]
[ 01-22-2002: Message edited by: Prim ]</p>
Remember that SpecInt has no relation with floating-point operations (SpecFP) nor SIMD instructions (Motorola Altivec, Intel SSE...). The three are different execution units.
So there is this AMD 64-bit 3400+, which seems very clever in integer operations. It is targeted for early 2003.
But what is this IBM PowerPC "4" @ 1.3 GHz ?
- Is it the Power4 ? I can't remember if the Power4 goes beyond 1 GHz, but I thought that its raw power was beyond all other processors, especially versus Pentium IV. Way beyond...
- Is it a G4 or G4+ ? I thought IBM didn't produce these chips, only Motorola ? And 1.3 GHz ? Today it seems too high, and by early or mid 2003 it seems too low...)
Note that if it is a 1.3 Ghz G4 (Apollo ?) a 2 GHz Pentium IV is less powerful (today), but a 3 GHz Pentium IV is better (early 2003)...
And I saw a 3.6 GHz (!) overclocked PIV yesterday
Hope the G5 will crush all these chips.
[ 01-22-2002: Message edited by: Prim ]
[ 01-22-2002: Message edited by: Prim ]</p>
Comments
3400 means that this ship is the equivalent in power than an hypothetical Athlon at 3400 mhz.
AMD has not any obligation to keep the secret at the contrary of Motorola.
Probabily Mot is able to show a G5 at 2,6 ghz (but what does it proove if there is only a couple of ships in the market ,)
On this benchmark i see that a P4 3 ghz nortwood is the equivalent of a PPC604 2,5 ghz. Considering that the G3 and G4 are better in integer that means tht a G4 2 ghz is better than the P4 northwood 3 ghz . For FPU where the P4 is not performant at all i bet that a G4 1,5 ghz have better performance.
<strong>where can you download the spec int program?</strong><hr></blockquote>
You don't -- you buy it:
<a href="http://www.spec.org/cgi-bin/order/"; target="_blank">http://www.spec.org/cgi-bin/order/</a>;
They are the first to have ANY benches or specs on the Clawhammer, and the fact that they're czecs (how the hell do you spell that), doesn't exactly add to the credibility. If AMD is developing that taht chip in Europe at Fab 31 or whatever taht thing in dresden is called again, it would more likely be a German page.
I can also make up benchmarks with photoshop that claim a 1MHz G3 si faster than a 1THz P4.
G-News
<strong>Those benches are full of shit. Hammer isn't even sampling yet.</strong><hr></blockquote>
The SPEC scores for the 3GHz P4 and the Clawhammer are based on extrapolated values and information presented by AMD. AMD has said that their initial K8 products will be on the order of 3-3.5 times faster than a 1GHz Athlon.
Extrapolated values never work. Too much tech-info goes down the drain and the assumptions are based on simple maths that can't be applied to such complex things as CPUs.
G-News
That's a pretty weak excuse. Then it should be clearly stated with footnotes. Even then...
industrialy product now, it's only a prototype like the G5. 3400 means
that this ship is the equivalent in power than an hypothetical Athlon at
3400 mhz.
AMD has not any obligation to keep the secret at the contrary of Motorola.
Probabily Mot is able to show a G5 at 2,6 ghz (but what does it proove if
there is only a couple of ships in the market ,)
On this benchmark i see that a P4 3 ghz nortwood is the equivalent of a
PPC604 2,5 ghz. Considering that the G3 and G4 are better in integer that
means tht a G4 2 ghz is better than the P4 northwood 3 ghz . For FPU where
the P4 is not performant at all i bet that a G4 1,5 ghz have better
performance.</strong><hr></blockquote>
Rather, the P4 is a step forward from the P3 in float performance (roughly
equivalent, clock for clock, to an athlon xp) but a step backward in
integer performance (the P3 is roughly comparable to an XP).
As for the 604 performing better, clock for clock, than the P4, it's
certainly possible (probable, in fact) in integer ops.
Regarding AMD's keeping secrets contrary to Motorola's wishes, the two
both were involved in developing the hipermos7 process, so in regard to
specific fabrication details (more? the contract could specify anything)
AMD may be contractually enjoined to keep its mouth shut.
<strong>Thus they mean SHIT.
Extrapolated values never work. Too much tech-info goes down the drain and the assumptions are based on simple maths that can't be applied to such complex things as CPUs.
G-News</strong><hr></blockquote>
Of course they mean shit, this information comes from a rumor site. I would think all of you have been around rumor sites to know that 95% of what they have to say is shit, and the other 5% is usually a lucky guess. Hammer will probally perform 3x faster on some applications, but not SPEC.
<strong>The P4 is not equivalent in FP at the same MHZ as an Athlon. Intel has pushed it too 2.2GHZ while Athlons are 1.667GHZ, and from the benches I've seen the extra 500 or so MHZ brings it about even.</strong><hr></blockquote>
i have seen benchamrks in FP operations : the athlon are much more performant than the P4 about the twice of performance . Incredible not : the P4 as only one fp unit the athlon 3.
some prefer a sleek and clever approach (such as the G4), some prefer a clever and scalable approach that relies heavily on MHz (P4) and some consider the brute force approach the best, stuffing as much execution units into a chip as possible (Athlon family)
G-News
<strong>
i have seen benchamrks in FP operations : the athlon are much more performant than the P4 about the twice of performance . Incredible not : the P4 as only one fp unit the athlon 3.</strong><hr></blockquote>
I'm speaking from SPECfp 2000 results, but if you have some other ones, by all means present them.
<strong>
I'm speaking from SPECfp 2000 results, but if you have some other ones, by all means present them.</strong><hr></blockquote>
OK :
sandra FPU
pentium 4 A 2200 mhz 256 MB rambus : 1306
Athlon XP 1900 + 256 MB DDR ram : 2220
Perhaps your benchmark take in account the SSE 2 instruction of the P4.
<strong>
OK :
sandra FPU
pentium 4 A 2200 mhz 256 MB rambus : 1306
Athlon XP 1900 + 256 MB DDR ram : 2220
Perhaps your benchmark take in account the SSE 2 instruction of the P4.</strong><hr></blockquote>
Interesting. But no, spec doesn't have any SIMD suite, which is probably a
good idea, although I have heard of nothing of the sort.
Judging from the hits I get on google, Sandra is primarily used by
tomshardware and clones; I wouldn't put it in the same league as a
cooperatively developed benchmark like SPEC. Also SPEC should be compiled
on the target box, afaik sandra isn't (correct me if i am wrong, I don't
know much about sandra). The fact that I can't find everything I need to
know in 10s on google isn't helping it, either
Technical Information
The AMD AthlonÂ? XP processor with QuantiSpeedÂ? architecture is the latest member of the AMD Athlon family of processors designed to meet the computation-intensive requirements of advanced software applications running on high-performance desktop systems.
This section includes technical information that explains the key technologies involved in making the AMD AthlonÂ? XP processor the high performance processor to meet your needs. You'll find a Competitive Comparison, Technical White Papers, and links to Additional Resources.
White Papers
Click on a topic to view a description and to download the white paper.
*\tQuantiSpeedÂ? Architecture
*\tUnderstanding Processor Performance
*\tAMD AthlonÂ? XP Processor Benchmarking and Model Numbering Methodology
More Information
Additional Technical Information is available at <a href="http://www.amd.com,"; target="_blank">www.amd.com,</a> or at the following links:
*\tSupport
*\tChipset Information
*\tOther Technical Documentation
Competitive Comparison
Feature\tAMD Athlon� XP\tPentium® 4
QuantiSpeedÂ? Architecture\tYes\tNo
Operations per clock cycle\t9\t6
Integer pipelines\t3\t4
Floating point pipelines\t3\t2
Full x86 decoders\t3\t1
L1 cache size\t128K\t12k µop (Trace Cache) + 8KB (Data Cache)
L2 cache size\t256KB
(on-chip)\t256KB
(on-chip)
Total on-chip full-speed cache\t384KB\t64KB + 12k µop
Total effective on-chip full-speed cache\t384KB
(exclusive)\t256KB - 12k µop
(inclusive)
System bus speed\t266MHz\t400MHz
3D Enhancement instructions\t3DNow!Â? Professional\tSSE2
Cache/prefetch controls\tYes\tYes
Streaming controls\tYes\tYes
DSP/comm extensions\tYes\tYes
The athlon has 3 fpu unit and the P 4 only 2, that's why it sounds logical that the Athlon is faster (he has a smaller pipeline either)
Concerning Spec fpu (correct me also if i am wrong) these benchmarks where made to compare different chips without the interference of the mobo and the system. Sandra is a benchmark that depends also of the mobo. In my example the athlon was on a KT 266 mobo and the P4 on a i850 mobo.
I think in real application, sandra is better test. However , the fpu benchmarks are not very important. AMD himself in his site dont take that benchmark in account, he prefers to show different families of applications : media, video , busisness , games.
Thats make more sense.
[ 01-25-2002: Message edited by: powerdoc ]</p>
<strong>
Interesting. But no, spec doesn't have any SIMD suite, which is probably a
good idea, although I have heard of nothing of the sort.
</strong><hr></blockquote>
Actually you are incorrect. While SPEC doesn't have any suite specifically designed to test a SIMD the compiler team that Intel pours millions of dollars each year into has developed compilers for SPEC (and by SPEC bylaws available to the public) which vectorize many of the SPEC applications to take advantage of SSE2. AMD's inclusion of SSE in their AthlonXP is one reason their SPEC scores saw a noticable improvement.
[ 01-25-2002: Message edited by: RazzFazz ]</p>