• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

AMDs Hammer - 40 % Faster Than Athlon XP?

A

AGodspeed

Diamond Member
That's the title of this German article. If anyone speaks German maybe they can translate this, since Altavista wasn't all that much of a help.

Despite what the title may imply, here's a little bit of what I was able to translate from this German article:

With 32-Bit-Applikationen the ClawHammer thought for fastidious final users is to be faster about 20 to 25 per cent than the Athlon XP. If 64-Bit-Software (x86-64) is used, the projection/lead would rise around further 15 per cent to 35 to 40 per cent, since then the extended instruction sets can be used fully.
 
Looks like that 30 to 40% number is coming only from a 64bit prospective. In wich case I would certainly hope a 64bit processor is that much faster than a 32bit proc in a 64bit app.
 


<< Looks like that 30 to 40% number is coming only from a 64bit prospective. In wich case I would certainly hope a 64bit processor is that much faster than a 32bit proc in a 64bit app. >>



Most reports have claimed a 5-20% performance if the app is recompiled for X86-64, AMD has previously indicated the number to be approximately 15% and Ace's Hardware seems to agree with that estimation.... that's quite a stretch from all of the 30-40% you've quoted as being due to X86-64. I can't see how the increased number of registers from 8 to 16 could give a boost of 40%, and that's where the vast majority of the gain in X86-64 performance is bound to come from.

Beyond that I have little to say, because this exact topic has already been covered thoroughly in the earlier thread that AGodSpeed linked to above.

I find it interestig that AMD denies multi-core versions of the SledgeHammer are in development.
Very interesting indeed given how well suited the core seems to be to such an implementation and the rather significant number of patents AMD has issued regarding such a potential implementation in the future.
 
I dont think that the 40% number comes from x86-64 alone... the 40% includes the 20-25% boost in performance v. Athlon XP @ 32bits and @ equal clock speeds. They say that above the 20-25% performance boost you would see another 15% boost by going to 64bit. So, 20-25% + 15% ~40%....

anyway, thats how I understand it..
 
I find it interestig that AMD denies multi-core versions of the SledgeHammer are in development.

Yes I found that quite odd as well. I wonder what that really means....
 


<< Maybe it means their not making them? >>

Why would the Hammer's integrated NorthBridge support dual core logic then? What about the numerous patents AMD has filed for?
 
AGodSpeed,

AMD has already said Sledgehammer won't be dual core; it was in the c't article on IDC; it was said again in Planet3Dnow.de's interview with AMD. As you know, the primary benefits of Sledgehammer are a second memory channel (dual vs single on Clawhammer) and increased cache. On Planet 3DNow, AMD confirms that Sledgehammer will have up to 1Mb cache (compared to 256Kb L2 for Clawhammer?). As an aside, AMD confirmed in the same interview, also listed under the news, that Thoroughbred would not increase the cache or FSB--of course, most of us already knew that.

As for future incarnations....if it was dual core, it probably wouldn't be called Sledgehammer, but rather something else. I would definitely expect dual cores from AMD with the move to 0.05....in what was it...2005? It's possible we'll see dual cores with 0.09, but I seriously doubt it. With all the extra cache you'd need to make good use of a second core...you're looking at somewhere between 150 and 200 million transistors. I think 100 million is about as much as anyone wants to mass produce on 0.09um (and that's Intel, AMD will be doing substantially less with Clawhammer and Sledgehammer).

For comparison, the "Northwood" P4 contains 55 million transistors (an extra 13million just to move from 256Kb to 512Kb L2 cache) and the Athlon/Thoroughbred contains 37 million. The Clawhammer is rumored to contain ~45 million, and Sledgehammer about ~80-85 million with 1Mb L2 cache. Intel indicated the next-generation "Prescott" P4 based on .09um would have 100 million transistors (sufficient for the current P4 core with 1.3Mb L2 cache, or a 1Mb L2 cache plus additional execution units and other architectural improvements). AMD might increase Clawhammer's cache from 256Kb to 512Kb L2 with the 0.09 shrink, although their current roadmap implies a direct shrink with no changes (62.5% from 104mm²->64mm² is just about max theoretical with no changes).
 
Reading some things a while back, I came to a conclusion that AMD has multi-core technolgy... i might have been wrong, but i doubt it...

anyway, all I have read is that there are NO plans for a multicore version of the sledgehammer... nowhere have i read that there are no plans for the hammer line... so sledgehammer's replacement in a few years or months may well have multicore. smt would rule =)...
 
Well, if i remember correctly an alpha 21164 500 would kick the behind of any PIII/athlon at equal speed, it was almost
2/2.5 times faster...the alpha 21264 was even faster so I think that you should expect a decent preformance increase!!

Provided they do things right...
But since Intel got the alpha division from Compaq I think that in no time the Intel/alpha processors will say fair well to
everyone else..
 


<< With 32-Bit-Applications the ClawHammer thought for fastidious final users is to be faster about 20 to 25 per cent than the Athlon XP. >>


Ouch. If this estimation is true, then I certainly will reconsider my plans to buy a ClawHammer this year. I was really hoping for the 40% estimate from the other thread. That is a huge difference in speed.

Athlon XP is roughly 25% faster clock for clock than a 400 MHz fsb P4. Thus a 2.0 GHz Hammer will be roughly the speed of a 3.1 GHz 400 MHz fsb P4 (2.0*1.25*1.25=3.1). However if the 533 MHz fsb P4 gets a 10% or more speed boost over the 400 MHz fsb P4 - then the Intel 2.8 GHz processor will tie or beat the Hammer. Plus there will likely be no price reason to go for the Hammer with AMD's current price increases.


<< Mr. G&uuml;tter also stated that ... ClawHammer will initially sell at a premium over Athlon XP. >>


If the 2100+ Athlon XP sells at $420 in lots of 1000 from AMD, I wonder what this "premium" will add. Of course street prices will be lower since companies can buy more than 1000 and get an even better deal.
 
Dullard the AthlonXP 2100+ can be had for 276$ and that is about the same as the AthlonXP 2000+ was when it first came out. 420$ that is Intel pricing.

I also dont think you took in account when comparing p4 and hammer in benchmarks SSE2 which Hammer will have as well. I dont think that the hammer will debut at 2ghz! I think it is suppose to debut at 2.2ghz or higher since that is where the Athlon is suppose to croke!
 


<< Dullard the AthlonXP 2100+ can be had for 276$ and that is about the same as the AthlonXP 2000+ was when it first came out. 420$ that is Intel pricing. >>


Here is AMD's official price list. See the $420 in lots of 1000? Price watch currently has that part for $277+$7 = $284 as the best price shipped. That is the reason I stated that street prices are lower.

Here is Intel's official price list. See that the $562 also doesn't match pricewatches $483. This is a few weeks before Intel's scheduled April price drop though (about a $125 drop in the 2.2 GHz price is expected which would result in roughly $437 official and roughly $358 street).

I don't know what AMD meant when it said Hammer will initially have a price premium over the Athlon XP. To me "premium" means at least $100 and more likely $200. If I assume that the premium is $150, then the Hammer's official price will be roughly $570 and Hammers street price will be roughly $434. However AMD might mean something quite different when they said that Hammer's price will carry a "premium"...


<< I also dont think you took in account when comparing p4 and hammer in benchmarks SSE2 which Hammer will have as well. I dont think that the hammer will debut at 2ghz! I think it is suppose to debut at 2.2ghz or higher since that is where the Athlon is suppose to croke! >>


I assume that the 20% to 25% boost quoted above took into account all of Hammers parts, SSE2 included. Since I don't have a Hammer, all I can do is to go with the best estimates of others. It would be stupid of them to give a performance estimate and not include all of the features in that estimate - but you do have a point maybe they forgot SSE2 in those estimates.
 
Ouch. If this estimation is true, then I certainly will reconsider my plans to buy a ClawHammer this year. I was really hoping for the 40% estimate from the other thread. That is a huge difference in speed.

There were no 40% estimates from any other threads that I read (at least any sane threads). Maybe you mean Jerry Sander's statement that "Hammer" (not revealing if it's Claw or Sledge that he's talking about) will be 30-50% faster at the same clock speed compared to current Athlon XP's. You must make the distinction between integer performance and floating point performance though. Paul DeMone estimated about a 30% increase in integer performance and a 15% increase in fp performance for ClawHammer; he also estimated a 35% increase in integer performance and a 50% increase in fp performance for SledgeHammer. Of course those are only estimates, and the real thing could end up being worse or could end up being better, we won?t know until final silicon is available later this year.

Point is, there?s no point in taking this ?20-25%? figure more seriously than most of the other figures discussed by AMD and hardware web sites on the Internet. We'll just have to wait. 🙁

But anyway, AMD is still scheduled to release a ClawHammer 3400+ processor this year so...
 
Id also like to point out that AMD is usually conservative when giving performance ratings! We all know that the model numbers on the AthlonXP are meant to debunk the p4 despite what they say! Point being AthlonXP's do much better than their model numbers suggest against the p4 with the same clock speed as the model number. If Jerry Sanders says 30 or so percent better than the AthlonXP at the same clock speed then I am thinking that actually there could be as much as 40 or 45 percent difference! AMD has always played it conservatively when quoting performance!
 
I cant remember, but what will a model 3400+ translate to in mhz? Remeber that is compared to the AthlonXP and not the p4 nor the Thunderbird?
 


<< Id also like to point out that AMD is usually conservative when giving performance ratings! Point being AthlonXP's do much better than their model numbers suggest against the p4 with the same clock speed as the model number...AMD has always played it conservatively when quoting performance! [W]hat will a model 3400+ translate to in mhz? Remeber that is compared to the AthlonXP and not the p4 nor the Thunderbird >>



We were talking about actual performance not some made up PR rating. The Hammer could have a rating of 340+, 3400+, or 34000+ and it won't affect the performance or speed of the Hammer one bit. If AMD is conservative or flat out misleading doesn't matter in this thread since we are talking about performance and not names.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top