Man walked up to Clawhammer, benchmarked it. Here's the result!!

[CheapTOFU]

Text

Anyone tried this bench?
I don't know what it is exactly ~
but as I see it, hammer @800mhz is about 4X faster than celeron @400mhz..

 

[Wingznut]

So, let me get this straight...

An 800mhz Hammer is 4x as fast as a chip that's almost four years old, runs at half the speed, and uses a fraction of the effective FSB and memory bandwidth?

No wonder AMD doesn't want people to benchmark it. Those desperate for a story, come up with ridiculous results such as this one.

(Yet another stupid article from theinquirer.)

 

[nagger]

Wingnut Pez,

You'll be more surprised when you look at the results the Itanic got at the same benchmark

Itanic and Hammer contrasted

 

[SSXeon5]

Originally posted by: nagger
Wingnut Pez,

You'll be more surprised when you look at the results the Itanic got at the same benchmark

Itanic and Hammer contrasted

Yeah go get an itanium 2, then benchmark it, you can see how much it smashes that slow hammer, still at 800Mhz 256K L2 huh

SSXeon

 

[nagger]

Originally posted by: SSXeon5


Yeah go get an itanium 2, then benchmark it, you can see how much it smashes that slow hammer, still at 800Mhz 256K L2 huh

SSXeon

I'm not saying that the Clawhammer beats the Itanic to the ground.

I don't even think that these benchmarks are at all believable since I don't think that AMD would allow any person in the Linuxworld to benchmark engineering samples.

I was only agreing with Wingnut Pez about this stupid article in The Inquirer.

 

[mechBgon]

Well, I sure hope Itanic 2 would at least merit a comparison with Sledgehammer, instead of a desktop Clawhammer with 256kb of cache.

 

[OneOfTheseDays]

dude this crap is bogus. And even if, BIG IF, it is somewhat credible this is only an 800mhz Hammer, early alpha stages, not indicative of any real world performance that we will be seeing.

 

[dexvx]

To get the full power of the Itanium, you need to modify the benchmark program with IA-64 in mind. Having worked with several Itanium (Merced) machines with fully optimized software, they are incredibly fast. But when you get them to do IA32 emulation, they can be incredibly slow.

And I highly doubt AMD would let some guy walk up to one of their computers and run a benchmark application.

 

[mechBgon]

The benchmark the man ran was the RSA crypto benchmark included in openssl (version 0.9.6d).

Sounds like this is built into OpenSSL and is run by a command, just like he could have typed "ping www.anandtech.com". Make any sense...?

 

[rIpTOr]

A Hammer with 256K cache yea right

 

[Duvie]

A Hammer with 256K cache yea right

What are you talking about!! It has ben said many times it was only going to be 256kb in some fashion of the chip at debut. It is likley the on-die memory controller makes more sense instead of adding more costly cache. The p4 likely with the prescott will have to add on die cache to an excess of 1mb to help compete against the amd memory controller. The more cache will help the p4 from having to dump down to physical memory and thus slowly the performance down. The on die memory controller if I understand it correctly alleviates this since it is right there with little to no hit in performance or time. cache should not be as much important in thoery.

 

[shathal]

The thing that "worries" me for AMD is that those are still the 800 MHz samples - I'd assume the same ones as seen on Comdex. I'd have thought they'd have faster ones by now. After all, if they do intend to hit 2 GHz, they should be around at least 1.5 GHz or so now, right?

Curious whether they're having problems that they keep their mouths shut about (which is fair enough), or whether there's just too few "new" samples to go around, and the 800 MHz parts are the only ones "shown" to the public, as it were.

Anyway - my thoughts on that article. Taking this sort of thing seriously is quite questionable. I'll rather wait 'til Hammers come out and see how they fare against Xeon/Itanium 2. THEN we'll see how things go, eh?

- Shathal.

 

[shathal]

What are you talking about!! It has ben said many times it was only going to be 256kb in some fashion of the chip at debut. It is likley the on-die memory controller makes more sense instead of adding more costly cache. The p4 likely with the prescott will have to add on die cache to an excess of 1mb to help compete against the amd memory controller. The more cache will help the p4 from having to dump down to physical memory and thus slowly the performance down. The on die memory controller if I understand it correctly alleviates this since it is right there with little to no hit in performance or time. cache should not be as much important in thoery.

Errr - no. Having a Memory Controller on-die is supposed to get you (round about) 5-10% better performance. Not because you need to access the Cache of the CPU less (complete nonsense) - but rather because it's all on-die and the speed-benefits that come from that.

You will still have to access main memory just as much with 256 KB cache - with or without on-die memory controller. The difference is merely in "how quick" you get there. An on-die memory controller doesn't all of a sudden give you better caching algorithms.

The big "problems" as I see it with an on-board Memorycontrollers are a few, and I think that some of those might have been reasons why Intel had dropped that project (Timna processor, if I remember rightly) a few years back (in case you wondered AMD came up with this first).

1: Heat.
======
More sillicon == more heat. Memory controllers DO run hot, and seeing as Athlons aren't the coolest blocks of sillicon going around, I have doubts that they'll be a lot cooler with a memory controller stuck on the die as well. Well - it'd be nice to see a "cool" AMD CPU at some point again ...

2: Memory support:
==============
An integrated Memory controller is not the "be all and end all". Remember, that the chip needs to be Rev'ed for EVERY new memory technology that's coming out, which I think is going to be a serious pain. So, taking your CPU and just "upgrading" to a new mobo which (for instance) supports DDR II would mean that you'd have to use the mobo's memory controller again. There goes your performance gain, unless you buy a NEW CPU.

Now me - being a little budget-minded - I prefer getting more out of my CPU's and I actually prefer them to be "independant" - that is generally for all things IT. I'd be happy if "all OS's worked on all hardware", and thus you'd really be able to slug together what you want, rather than have hardware-specific OS's and suchlike. I am an idealist .

Sure - you can TURN the thing OFF, but then you lose the benefit, when you return to using the memory controller on the motherboard. And if mobo vendors have to include a memory controller on their product STILL, then where's the reduction of cost? Not present, I think. And if you use the mobo's memory controller, AMD is wasting Sillicon/die-space by having a none-too-insignificant part on there that's not getting used. It's a bit of a gamble on AMD's part I think, and I am not certain whether they'll get much out of it. But I tend to be weary anyway, so feel free to ignore my words of caution .


****

So, it's all a bit of a marketing game. Sure, the integrated memory controller HAS got benefits (if you can use it). Intel's engineers aren't stupid though (despite what readers might want to believe), and I am quite certain that they had good reasons to can their version of the "CPU with integrated memory controller". Whether any of the reasons I stated above have anything to do with their decision is as good a guess as any.

Hope this brings an interesting debate up - I am happy to learn if I should be mistaken somewhere .

- Shathal.

 

[CheapTOFU]

I think ppl don't know the difference between P4 and Hammer..
Hammer was designed for 64bit softwares and some~ 32bit softwares... P4 may have more cache and more speed, but because it cannot use 64bit software, it's almost useless in the future.. Athlon XP also..
I know that Intel is also making 64bit CPU, but it's not out yet and no one had seen it.....
But @800mhz, at least you can bench Hammer..
Who knows... Hammer may be really good or really bad... We will know in the future..
But I'm certainly gonna buy Hammer..
Oh~ also I don't care much about speed... I think 1.2 ghz is enough to do anything.. and if I can use 64bit OS, that's better

 

[shathal]

I think ppl don't know the difference between P4 and Hammer..
Hammer was designed for 64bit softwares and some~ 32bit softwares... P4 may have more cache and more speed, but because it cannot use 64bit software, it's almost useless in the future.. Athlon XP also..
I know that Intel is also making 64bit CPU, but it's not out yet and no one had seen it.....
But @800mhz, at least you can bench Hammer..
Who knows... Hammer may be really good or really bad... We will know in the future..
But I'm certainly gonna buy Hammer..
Oh~ also I don't care much about speed... I think 1.2 ghz is enough to do anything.. and if I can use 64bit OS, that's better

Right - let's go through this, shall I?

Hammer was designed for 64bit softwares and some~ 32bit softwares

Nope - not true. Hammer is a 32-bit CPU with 64-bit extensions slapped on top. From what I've seen on it (which are things you're not going to find on the inq or so), this is quite clear. Hammer is not a real 64-bit CPU. Designing a REAL 64-bit CPU takes ages - Intel have done this with the Itanium and if I remember right, they worked on that since 1995 or so. Hammer *is* a 32-bit CPU, with some 64-bit extensions. Not the other way round - don't believe the marketing stuff.

Hammer was designed for 64bit softwares and some~ 32bit softwares... P4 may have more cache and more speed, but because it cannot use 64bit software, it's almost useless in the future.. Athlon XP also..

Also not the case. What exactly do you NEED 64-bit for on Desktops? Come on - give me reasons here.

Servers - YES - but even here, only in SOME instances. Pushing 64-bitness out is not a small effort, and I don't see desktops needing 64-bitiness for a few years yet. In what way exactly would you say that the current XP / P4P (chose your own poison ) are "not good enough"?

Furthermore, the cost of memory for 64-bit OS's would ruin most people. You can quite happily get away with 128 - 256 MB of memory, so what exactly do you need 64-bit addressability for, if you can still go that much higher with 32-bits?

I know that Intel is also making 64bit CPU, but it's not out yet and no one had seen it..... But @800mhz, at least you can bench Hammer..

Eh? And what exactly do you think Itanium and Itanium II are? They're "true" 64-bit CPU's, insofar that they're 64-bit only, rather than 32/64-bit combined / mixed-together. And they've been around for a while. They're server-CPU's - YES - hence not a lot of people know about them that don' t deal with servers. But what you need 64-bit Desktops/Workstations for I've yet to see exactly...?

But I'm certainly gonna buy Hammer..
Oh~ also I don't care much about speed... I think 1.2 ghz is enough to do anything.. and if I can use 64bit OS, that's better

And how exactly is a 64-bit OS going to help you, as an end-user (which - let's face it - you would be). 64-bits will NOT be "for free", you will need OS-support, and the IHV's will have to write 64-bit drivers, which I doubt AMD will quickly gather the muscle for to do. So really, you'll be stuck with a 32-bit CPU tha's got 64-bit sillicone. And although 64-bit sounds "nice", as it's twice as much as 32-bit, in actual terms, it'll do you no good, unless you use Linux.

So - your line of argument seems rather shaky and I'd not buy a 64-bit platform with such loose footing. I'd recommend that you wait until the Hammer's come out and let things cool off for a bit - see how "real world" reacts to them. Don't buy processors (regardless of make) purely on marketing hype - which, at the moment, you certainly seem inclined to do.

Right.

Hope this helps a little.

Note: Don't try & flame this as an Anti-AMD post. It isn't. It's trying to point out that 64-bitness doesn't really bring anything for Desktops that's my point.

- Shathal.

 

[dexvx]

Hammer was designed for 64bit softwares and some~ 32bit softwares... P4 may have more cache and more speed, but because it cannot use 64bit software, it's almost useless in the future.. Athlon XP also..

The p4 has SSE2 instructions which extends to 64bit. Since the hammer is a 32bit hybrid, it'd be interesting to see how customers react. I know right now that some dont consider the hammer as a true 64bit solution (thus some hesitation), but it is 64bit nevertheless.

Also in a pure 64bit environment, it is slower than a 32bit environment since the memory pointers should all be 64bit. Having used 64bit Windows it looks and operates exactly the same as a 32bit Windows. As for the Itanium, the McKinley I believe is the most powerful single processor to date, floating point wise, under an IA-64 environment.

Also not the case. What exactly do you NEED 64-bit for on Desktops? Come on - give me reasons here.

I see a general trend towards 64bit in the future. I mean back then, who would've thought you needed more than 16bits? (Pentium Pro vs Pentium).

 

[shathal]

dexv:

Oh yeah, I agree fully towards there being a tendancy to moving to 64-bit. Eventually, I am assuming that we'll have Itaniums (whicher ever generation) in our desktops, instead of 32-bit CPU's.

I am making the point that 64-bitness is a complete waste of time to even be considered on desktops, otherwise (I'm quite sure) Intel would be trying to be ahead of AMD in that arena. It is far too early imho to even consider 64-bitness for desktops (really - no need from a CPU point of view) at the moment.

Therefor, my oppinion is that AMD will have a number of 32-bit CPU's out there, whose 64-bit parts WON'T get used (Sledgehammer), as you'd not only need OS support (which as far as I recall, M$ might give), but you also need the IHV's to give you drivers => THAT is going to be a problem. Also - the apps would need to make use of 64-bitness, and - I dare say - that is not the case by a LONG shot.

Clawhammer is an of sorts hybrid, I have a feeling that it's sort of "jack-of-all-trades, master of none". It has got 32-bit stuff, it has got 64-bit stuff, but I can't see it really being fantabulous on either side. Admittedly, I can see the reason - AMD *HAVE* to play in the 64-bit arena, and they don't have Intel's budget to go and develop a true 64-bit architechture of their own.

Insofar, I'm not blaming them for the choice they make and have made - I can see the reasons why. I have my doubts, however, about whether this will bring a desired result or success.

- Shathal.

 

[CheapTOFU]

Therefor, my oppinion is that AMD will have a number of 32-bit CPU's out there, whose 64-bit parts WON'T get used (Sledgehammer), as you'd not only need OS support (which as far as I recall, M$ might give), but you also need the IHV's to give you drivers => THAT is going to be a problem. Also - the apps would need to make use of 64-bitness, and - I dare say - that is not the case by a LONG shot.
______________________________________________________________________________________________________________________________________

I believe this is what ppl said when first pentium1 came out!!
My first computer was 486 66mhz..
I was told by ppl that pentium 60mhz was not that faster than 486 66mhz (I didn't know much about comps then)..
of course, when you use dos or old window, 486 is fast..

Also, ppl said pentium 60mhz had problems with dos and windows and I should wait or just buy 486.. but as soon as everyone started buying pentium PCs, there came win95.. and 486 was too slow to do anything...

Same thing will happen again... Microsoft is already working on 64bit OS.. when it comes out, P4 and athlon xp will become like old 486 that no one cares about

 

[dexvx]

I think more or less its going to be a PR game. Intel will have clock speed while AMD will have 64bit. It will be interesting to see what consumers think will be better.

Your analogy with the 486 doesnt necessarily hold true. Pentium4 has SSE2 instructions which can bring it to 64bit floating point precision if used. The pentium 60-66's were a waste of money when they first came out. They generally performed slower and were more expensive than 486 counterparts until they got software optimized for them. But by that time, Pentium133s were dominant (if that holds true, than we'd be seeing Hammer @ 6800+ before 64bit is accepted), and the Pentium 60-66's were fading in memory and performance. The only people who bought Pentium 60/66's were usually companies who had proprietary software ready/being built for them to take full advantage.

Also its far more worthwhile to upgrade from 16 bit to 32bit precision (REALLY worthwhile going from 1 to 2 bits ). But upgrading from 32bit to 64 bit is "eh?". Then 64bit to 128bit precision is iffy.

 

[THUGSROOK]

thats assuming anyone will want M$ next OS at all.

from what ive been reading its gonna be like installing a tollgate with a cop inside it.
you thought WinXP is taking your rights away? the next OS is gonna take the PC out of your computer completely.

YOUVE BEEN WARNED.

 

[LastRide]

Like having the wiggies (cop) inside your computer:disgust:I'll go back to ME or 98SE thanks

 

[shathal]

Originally posted by: CheapTOFU


I believe this is what ppl said when first pentium1 came out!!
My first computer was 486 66mhz..
I was told by ppl that pentium 60mhz was not that faster than 486 66mhz (I didn't know much about comps then)..
of course, when you use dos or old window, 486 is fast..

Also, ppl said pentium 60mhz had problems with dos and windows and I should wait or just buy 486.. but as soon as everyone started buying pentium PCs, there came win95.. and 486 was too slow to do anything...

Same thing will happen again... Microsoft is already working on 64bit OS.. when it comes out, P4 and athlon xp will become like old 486 that no one cares about

Well - there's a bit of a difference with Pentium. And yes, I was around when the whole "My 486/100 beats your P-60 / P-75..."-thing went round. The difference then was Intel was pushing things - as they are doing with Itanium. AMD just does NOT have the same muscle (read: funds/influence) that Intel has. This is not a "Oh, Intel is better than AMD"-type comment, it's merely stating a fact.

Also, it's one thing comparing the "general route" of things to come at the time, which back then was still Intel-only really, as opposed to now, where Intel and AMD *seem* to be trying to go different ways. SEEM because I doubt that they will.

My prognosis/theory is this:

* Intel HAS got the funds/capabilities to have developed and keep on developing a 32-bit CPU and a 64-bit CPU.
* AMD hasn't. Therefor, AMD has to look for other solutions. A less-than-ideal way (but a way nonetheless) is to have its Hammer Solution - the 32/64-bit crossover.
* AMD has to stretch itself - it's got to make sure it has it's fingers in the 64-bit pie for future strategic reasons. It'd be stupid not to.
* The 64-bitness for Desktops is a complete waste of time (see my comment above) - I have extreme doubts that AMD will really be able to push it to any significant degree. Look at the "speed" that the market is picking up Itanium. And that's with Intel's muscle/funds behind it. See now where I am goin?

I worry a little about AMD's future - spreading itself too thinly and trying to do too many things at once. Desktops are DEFINATELY not ready for 64-bitness yet and don't need it. Most servers - don't need it, 32-bit is quite enough for MOST (not all). I am not saying "Thou shall stay on 32 bits", no no - just that the time for "64 bits" has not come yet - though I expect it to be ripe in a few years.

A similar thing has happened with Pentium/Pentium Pro. Yes, at the time it seemed like an iffy move on Intel's part, but I doubt many of you regret it in the long run, as PPro was father/mother to PII and PIII as well - and that was a definite improvement over even a 486/100, eh? .

One point is, you need to have muscle to push 64-bitness out. Software needs to make use of it, the OS needs to be able to support it, and a whole raft of other things that need doing. AMD itself doesn't have the muscle for it (That's my oppinion - correct me if you know better). Intel is pushing towards 64-bit, and look at how far it's got. It's not because of half-hearted attempts, rather just resistances on various fronts. Pushing out a new architechture like that is not an easy or a simple thing.

Hope this helps a little.

- Shathal. .