Athlon64 Preview at xbitlabs

[imgod2u]

After looking around, I'm starting to think it's the p4 2.8ghz's hyper-threading that makes its score so high in SSE.

This is very doubtful as even one SSE instruction takes up the entire FP calculation resource on the P4 and Athlon for a single clock. I don't see how allowing multiple threads to run would possibly speed this up because the instruction to operation ratio is 1:4 (i.e. for one instruction fetched and decoded, 4 operations need to be carried out). SMT is ultimately to circumvent the bottlenecks for decoding instructions and other things such as data dependencies and memory latency. With such a streaming nature, I doubt SIMD benefits much, if any, from SMT.

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

 

[Schmide]

It may seem doubtful but the benchmarks speak for themselves. Most likely the SiSoft multi-media benchmarks are comprised of many separate tasks that when executed in order produce dependences that are elevated by hyper-threading.

Integer
14221 (<A HREF="http://www.hardwareanalysis.com/content/image/10594" target="_new">HT 3.06</A>) - 12292 (<A HREF="http://www.hardwareanalysis.com/content/image/10597" target="_new">NHT 3.06</A>) = 1929 / 12292 = 15% improvement
Floating Point
21916 (<A HREF="http://www.hardwareanalysis.com/content/image/10594" target="_new">HT 3.06</A>) - 15618 (<A HREF="http://www.hardwareanalysis.com/content/image/10597" target="_new">NHT 3.06</A>) = 6298 / 15618 = 40% improvement

Although not the same benchmark results (or for that matter not sure if it is the same benchmark), the analogy should still hold.

Dichromatic for your viewing plesure...

 

[imgod2u]

I dug up <A HREF="http://www6.tomshardware.com/cpu/20021114/p4_306ht-16.html" target="_new">Tomshardware's</A> benchmark using Sandra 2002 (one of the few sites that even bother to use Sandra as a benchmark anymore). The improvement seen from HT was 16.75% for SSE/SSE2 FP. I'm not sure exactly why this would be. But it is a substancial improvement nonetheless.

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

 

[vacs]

AMD said the A64 would be 15% faster in 64 bit mode so this pr rating is probably the final one it is just that we need a 64bit o.s. to see the A64 is all its glory.

Ha, AMD got you there. How can 64-bit faster when there are no 64-bit programs/games available? And even if you have the 64-bit version of UT2003 and compare the fps to the 32-bit version on the same Athlon64, the difference in speed will hardly be more than 5%, if at all...

The Athlon64 needs initially to be strong at 32-bit, a year later when more programs and games are available as 32/64-bit ports, then the Athlon64 can show it's 64-bit potential if it has one. But 64-bit is not going to be important this year...

If the Athlon64 can't keep up with the x86 performance of a 3.4 Prescott than the Athlon64 will fail miserably!

 

[mr_gobbledegook]

Hammer clock speed issues, cause for concern or will AMD able to scale it higher ??

<A HREF="http://www.theinquirer.net/?article=9045" target="_new">Link</A>

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple>

 

[bikeman]

Is it just an impression I have, or is that actually a reliably-looking article from the Inquirer?

Seriously, though ... If this thing is going to happen to Opteron, what would it mean for the A64? I guess that for the server market the lower clockspeed isn't that much of a problem. Maybe in server-applications the Hammer core takes bigger advantage of it's beeter branch-prediction and on-die memory-controller. But for the desktop a higher clockspeed is necessary, as seen in the first article discussed in this thread Though if we suppose the A64 does come out at 2.0 GHz, and we extrapolate the result from the benchies done bij XBitLabs, it should be able to compete with the fastest P4 at that moment. On the other hand, if there are true problems scaling the Hammer core, then the future doesn't look very bright for AMD ...

Greetz,
Bikeman

<i>Then again, that's just my opinion</i>

 

[martinrichards23]

"But 64-bit is not going to be important this year..."

good point, i suppose 64bit is pointless thing to mention until software is available.


"the difference in speed will hardly be more than 5%, if at all"

AMD said it would be 15% faster in 64bit, mind you Nvidia said gf fx would be fastest thing ever and Intel said willy p4 would be fast, but i trust AMD more.

also check this out; http://www.theinquirer.net/?article=9046

it raises some good points.

 

[mr_gobbledegook]

This is more to do with Opteron but could also mean something for the Athlon64.

Came across this website and found some heatsinks for the Opteron. If you click on the link and look at products J023C and J023H it states they are certified for AMD Opteron processor up to 2.0~3.0 GHz !!!

Does this mean AMD reckon they can get Opteron to 3Ghz ? (which is quite probable) or even more significant already have samples running >2Ghz...I dont know if this is a typo or just shoddy info.

<A HREF="http://www.genesysrack.com.tw/1u_accessories.htm" target="_new">http://www.genesysrack.com.tw/1u_accessories.htm</A>

<font color=purple>Ladies and Gentlemen, its...Hammer Time !</font color=purple>

 

[eden]

Just check out Thermaltake, they already have a WHOLE range of A64 and Opteron HSFs, with all info in there. They include a retention mechanism similar to Pentium 4s as well!

--
This post is brought to you by Eden, on a Via Eden, in the garden of Eden.

 

[reever2]

http://www.theinquirer.net/?article=9046

incase you wanted to know what the inq thought of the review

 

[juin]

The best is Windows.net is so much faster on a P4 due to Highly optimization of tread sheduling and SMP support and a big improvement for NUMA systemes like I2.Ms made is choice INTEL like they have allwyase do wich make them rich also.Some Madison bencmark are out please add 50% more power in Itanium ligne last stop before Dual core SMP Itanium.

According to intel the clock speed will go around 2.4 relate to speed that were 90NM start.Not the 2.5 give by most or still belief in this figure.(do i have say 2.3GHZ or less) good yiled will be reach only at 1.8GHZ to 2.1GHZ over this speed AMD will have to deal with low price.

I have see some support for PAE 36 bit addresing.



[-peep-] french