New hammer benchmarks

[knewt]

I just found these <A HREF="http://www.theinquirer.net/?article=6737" target="_new">benchmarks</A> posted for a prototype hammer. I thought the hammer was demonstrated operating much better than this in the past or am I not understanding the results correctly?

 

[eden]

First of all it's clocked 1.2GHZ against 1 GHZ P4 higher.
And for the results in SySmark, that just may finally allow it to win in that benchmark which the Athlons lost all the time, when it will clock at 2GHZ.
As for FP benches, I think it is rightfully competing well, 1GHZ less yet it almost matches. So the K8 will reclaim the crown for FP once more soon.
As for Integer, same thing as before, it's at 1.2GHZ, not close to the actual clock, so I think it's doing well on its own for now, especially since this is Clawhammer! Per clock it seems to be about 30% better than AthlonXP. This is more than enough for 2.2GHZ ClawHammers to compete 3GHZ+ P4s, and CH will also have SSE2.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 12/14/02 04:33 PM.</EM></FONT></P>

 

[knewt]

I was not referring to how it compared to the P4. I am talking about how it compared with previous benchmarks. Wasn't there a benchmark where it showed higher numbers than the Itanium 2? I may be mistaken. In any case, I agree that the results are interesting.

 

[eden]

Yes those were the Opteron benchmarks and no doubt the Sledge since it used Dual Channel.
They were over 1000 points, indeed strong. But again that was a 2GHZ Opteron. A 2GHZ Claw would have results about 30% lower. I think everything is still accurate, while AMD in fact "preached" that the Opteron results were a bit too selfish and that it probably is higher.
Still, the results are interesting in that if AMD can get 2GHZ CHs out by early Q2, that can be a very good sign of competition from AMD.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>

 

[ritesh_laud]

This is more than enough for 2.2GHZ ClawHammers to compete 3GHZ+ P4s, and CH will also have SSE2.

Well the Clawhammer they tested is stated as having 1 MB L2 cache. I think the regular Athlon 64 is supposed to have 256K of cache (to keep the die size down), so it would perform considerably worse. The best indication we have at the moment is what AMD itself stated earlier this year: that the regular CH would perform about 20-25% faster than the Athlon XP per clock.

Ritesh

<P ID="edit"><FONT SIZE=-1><EM>Edited by ritesh_laud on 12/14/02 06:05 PM.</EM></FONT></P>

 

[imgod2u]

I was not referring to how it compared to the P4. I am talking about how it compared with previous benchmarks. Wasn't there a benchmark where it showed higher numbers than the Itanium 2? I may be mistaken. In any case, I agree that the results are interesting.

The benchmarks released for the Opteron was SpecInt and SpecFP. It outperformed the Itanium 2 at 1 GHz in SpecInt. To bring that into perspective, a 2.8 GHz P4 also outperforms the Itanium 2 at 1 GHz in SpecInt. Although as I recall, the Opteron had a 1300 score in SpecInt, the 3.06 P4 with Intel compiler version 7 scored a little over 1000 I think. Itanium 2 at 1 GHz still was and is the top in SpecFP.

"We are Microsoft, resistance is futile." - Bill Gates, 2015.

 

[vacs]

When analysing these benchmarks don't forget that the P4 only had PC2100 memory (which slows down the CPU noticable) and this Intel system is nothing like the one the Athlon 64 has to compete when available next year: HT, 800MHz FSB, mid 3GHz clockspeed P4s and Dual Channel or DDR II RAM! Prescott P4s will also be available shortly after the Athlon64s hit the stores...

 

[zengeos]

Actually, with a currently projected 6 month difference between release of Prescott and Hammer, my *guess* is Hammer, by the time Prescott hits shelves,, will be clocking higher...maybe near a real 2.5ghz clockspeed...or perhaps even faster. So I think we'll need to see what we see as things transpire. Several people have suggested that it's only a few months betwen Claw and Prescott and Prescott will offer significant performance improvements. I don't doubt it at all. I just think there are too many ifs ands and buts....Claw has to ship. Opteron has to ship...Prescott has to actually ship as well.

Is Hammer shipping in Q2 at all or is it actually being released in Q1/Q2 for actual shipping in Q3. Prescot is supposed to start production in the 2nd half. But if it begins vol production in late Q3 or Q4, doesn't that mean it won't start shipping until Q104?

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>

 

[eden]

Several people indicated that with the memory controller, the cache size is less important or dependant.
So I'd say the extra 768KB would only provide around 8% more performance, as seen here.
The extra cache will probably run the CH a few more bucks as the NW did.

And nobody said these are final cores, AMD did indicate they wanted to improve them, so you never know if they might add something or configure better memory speccs.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>

 

[eden]

At 2.2GHZ, the PC2100 isn't that much of a huge hamper IMO. But still it does reduce.
Nevertheless, even with PC1066 configurations, the CH at 1.2GHZ is still doing well on its own, being 1 GHZ behind. So the margin would reduce to 800MHZ delta in order to be on par if against PC1066.

Dual Channel is already there, but it's DDR400 Canterwood systems that we want to see.
It all depends on the release of CH, its clock speed, its micro-architectural enhancement amount and the price.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>

 

[juin]

Wasn't there a benchmark where it showed higher numbers than the Itanium 2? I may be mistaken. In any case, I agree that the results are interesting.

I guess that was a rendering time so lower it better LOL

Now what to do??