Sign in with
Sign up | Sign in
Your question

800Mhz Clawhammer Benchmarks up - looks good.

Tags:
Last response: in CPUs
Share
June 6, 2002 10:22:01 PM

<A HREF="http://www.tecchannel.de/hardware/937/index.html" target="_new">Looky here</A>.

tested at 800Mhz.
the (only) Quake 3 Arena bech looks pretty good - 40% faster then Athlon MP on the same clockspeeds, on par with the a 1.6 Ghz willmate.

though we have to remeber that Quake 3 Arena is pretty memory dependent, Id expect Hammer to be litle less then 40% faster then Athlon Xp on most tasks, unless, Hammer lunches with more then the tested 256kb Cahce version...

This post is best viewed with common sense enabled
June 6, 2002 10:33:50 PM

The Quake 3 benchmark shows about a 33% higher IPC than Athlon XP. No other benchmarks though, which is strange.

<font color=blue>Hi mom!</font color=blue>
June 6, 2002 10:37:25 PM

well, it seems as they were short on time...
anyway they did other benchmarks - Quake 3 is the only gameing benchmark though...

This post is best viewed with common sense enabled
Related resources
June 6, 2002 10:50:38 PM

its a bit more then 33%
135Fps (fastest mp score) * 1.33 = 179.55
Hammer score = 183.

its ~35% faster then the fastes athlon MP score.


This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 06/07/02 01:51 AM.</EM></FONT></P>
June 6, 2002 10:56:40 PM

sweet!

Jesus saves, but Mario scores!!!
June 6, 2002 11:48:48 PM

bump.

now im a tank, armoured against bs@dum.
<i>if you know you don't know, the way could be more easy
June 6, 2002 11:48:52 PM

bump.

now im a tank, armoured against bs@dum.
<i>if you know you don't know, the way could be more easy
June 7, 2002 1:38:47 AM

What impresses me most, is that this thing isn't even complete, is at A0 stepping, AND is only 256KB cache. Imagine IF it will use 1MB, this thing at 800MHZ will then be able to compete the NWBs ratio-wise the same as it did with Willy 1.6, as in 2 less speed, more performance. It is about 64% stronger than a Willy 1.6GHZ, and with the future cache, and improved steppings, at 800MHZ it will continue to compete if compared to a 1.6B 533MHZ.

--
Meow
June 7, 2002 2:50:17 AM

external clock is 200Mhz..... people say the FSB will be quad pumped, equivalent to a 800Mhz FSB...interesting?


or have i got something wrong here :) 
June 7, 2002 3:00:30 AM

How can it be 200MHZ? It has no FSB from what I hear. This must be the mem, which would translate to 400MHZ DDR, but that still does not make sense too.

--
Meow
June 7, 2002 3:02:23 AM

I'd like to see Mat's reaction on this one!

--
Meow
June 7, 2002 3:07:43 AM

i agree, I'm a little confused too, but there is a northbridge still, although there's not much for it to do?

perhaps the "external clock" reading of 200Mhz was a suprious reading, much like the programs thinking it was a Duron processor....

oh and the memory they used on the benchmark was DDR 333
June 7, 2002 3:47:37 AM

"Pentium 4 Willamette 800 MHz: Intel's CPU pumps about as much performance from the fast RDRAM as the Clawhammer from DDR333-SDRAM"

I missed where they came up with 800Mhz P4, did they take figures for a 1600Mhz wilma and divide by 2? its listed a few times as 800Mhz P4.

You are limited to what your mind can perceive.
June 7, 2002 4:57:31 AM

the role of the northbridge has changed.

it is no longer involved with cpu <--> mem interactions.
the NB is for agp and the SB

<font color=blue>Pants Down! Turn Around! Bend Over! You're about to Experience Telstra broadband! :lol: 
June 7, 2002 5:11:47 AM

Quote:
How can it be 200MHZ? It has no FSB from what I hear. This must be the mem, which would translate to 400MHZ DDR, but that still does not make sense too.


That makes sense to me. Or maybe it's reading the effective clock and it's PC1600? It's possible.

<font color=blue>Hi mom!</font color=blue>
June 7, 2002 5:27:40 AM

I got the impression that they either divided by 2, or had an engineering sample that they knocked the multiplier in half. I seriously doubt that they cut the FSB in half, though if it's pro-AMD that much, it's possible.

Also, it is possible that they took a range of results from the whole Willy line and extrapolated.

Edit: They stated time and again that the P4 Willy is at 400Mhz FSB, so the cutting the FSB in half to get 800Mhz is not possible.

-SammyBoy

<P ID="edit"><FONT SIZE=-1><EM>Edited by SammyBoy on 06/07/02 00:37 AM.</EM></FONT></P>
June 7, 2002 5:35:06 AM

Hmm... the WCPUID seems to indicate that there is no FSB or clock multiplier, as those areas are ghosted. The only thing shown is the Internal Clock... plays into the horrible side-step that I posted earlier in AMDs FAQ about the FSB of the Hammer. Said something about how it would be unlike anything seen before, or something like that.
June 7, 2002 6:55:28 AM

~ AND is only 256KB cache. Imagine IF it will use 1MB ~

With the memory controller on-die there is not as much value in having a large L2 cache provided that the memory is actually physically close (as in path length) to the CPU.

It is actually better to have a moderate L2 cache (therefore reducing CPU heat) so you can crank up the Mhz or more importantly increase the memory access speed.

<font color=blue> The Opteration was a success... I'm now a full-wit</font color=blue> :eek: 
June 7, 2002 9:44:22 AM

MMMM nummy, exactly what was stated before give or take, 35% faster IPC on a process which will be able to scale even higher than the axp, looks good to me. A hammer@1.8ghz will be equal to an axp@2.435ghz, which is over pr rating 3000 I believe, by quite a bit, and the hammer will be able to scale to 2ghz+!

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
June 7, 2002 2:25:13 PM

Is it possible (or likely) that quake 3 gives misleading results for the hammer? i hope it doesnt, but amd said that the hammer is 15% faster than athlon for 32bit applications (and a further 10% faster for 64bit).

I need a 1.5 Ghz Athlon + 512mb ddr ram to write emails......honestly
June 7, 2002 4:00:16 PM

It's very possible, and it's odd that there was only one benchmark.

<font color=blue>Hi mom!</font color=blue>
June 7, 2002 5:25:24 PM

I think you are right. Quake 3 is very memory intensive application and that's where willy P4 was beating the Athlon even when all the other benches shoed that Athlon was ahead. And I remember then that everyone just ignored the Q3 benchmark since it didn't prove anything.

Anyways, no matter what it's still a big preformance improvement for Hammer since 800Mhz can come on par with 1600Mhz willy.

KG

"Artificial intelligence is no match for natural stupidity." - Sarah Chambers
June 8, 2002 11:12:08 AM

Actually. I believe AMD said Hammer performance improvement per clock would be around 25-35% for 32 bit and 10-15% better than that for 64 bit.

So, the Quake performance bench is at the high end of the 25-35% improvement. That makessense too since that was one of Athlon's weak benches.


Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>
June 8, 2002 2:29:25 PM

Well just keep in mind all, per clock Quake 3 is still running better on Athlon. All the P4 is doing is beating it by big clock speeds, that is all really. Hammer just makes this move easier to compete now. Consider the 512-1MB cache in the future, and you got a killer NW too.

--
Meow
June 8, 2002 2:32:12 PM

Quote:
Consider the 512-1MB cache in the future, and you got a killer NW too.


Its a northwood killer now, 1.6ghz clawhammer=3.2ghz p4(from what weve seen here), add to that the fact it should scale well past 2ghz(it has more stages than the axp thus should be able to clock higher).

Also add the possible 1 mB cache and you get a bad ass mofo of a processor.

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
June 8, 2002 2:39:16 PM

and the SOI process... the extra piplining could be just to KEEP Hammer as scaleable then the Athlon XP - as you know X86-64 is a more complex ISA. thus the Hammer might need some extra piplining to handle x86-64 and stay as scalable as the Athon.

with 512Kb/1Mb of cache and a better memory latncy as result of higher-clockspeed (the memory controler speed scales with CPU speed) I'd expect AMD to have a PR ration of 1/2 meaning a 2Ghz ClawHammer markets as 4000+.



This post is best viewed with common sense enabled
June 8, 2002 2:49:31 PM

Quote:

Well just keep in mind all, per clock Quake 3 is still running better on Athlon. All the P4 is doing is beating it by big clock speeds, that is all really.

Well just keep in mind all, the Pentium 4 still atains the Highest clockspeeds of any CPU do date. the only way the Athlon doesn't stay far far behind (only litle behind) is just by having big big IPC, thats all really.

now plese stop with this disrespect for clockspeed. it is JUST and as importent and meaningful as IPC.



This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 06/08/02 05:51 PM.</EM></FONT></P>
June 8, 2002 3:41:43 PM

Im gonna disagree iib...

clockspeed is worthless, I wanna know how much work its doing. Your boss doesnt care how fast your working, he cares about how much your actually getting done. The P4 definitely gets a lot done... but i could care less how fast it has to work to do it. I just want it done (ie.. IPC)

Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.
June 8, 2002 5:07:40 PM

IPC is just as worthless as Mega Hertz...
or just as meaningful as Mega Hertz.

your overall preformance is Mhz * IPC
or the "actully getting done" or "how much work its doing" thing, however you'd like to call it, is not IPC alone.

for instance if you want to mesure your work for each second it will be composed of how many cycles you do in a second and the avrage instructions per such a cycle you do.
thats what your "actully getting done" in a given second.


This post is best viewed with common sense enabled<P ID="edit"><FONT SIZE=-1><EM>Edited by iib on 06/08/02 08:15 PM.</EM></FONT></P>
June 8, 2002 6:07:16 PM

of course, IPC and clock speed are both important because at the end of the day all we're interested in is work done per unit time.
June 8, 2002 6:38:21 PM

exectly.


This post is best viewed with common sense enabled
June 8, 2002 8:10:23 PM

Oh geez there we go with this again. I did not start the old clockspeed vs IPC argument here. I stated that Kemche isn't right to say that the P4 Willy would win just because of the bandwidth, while if we went technically wise, at its clock speed, it is NOT superior than Athlon even bandwidth wise. It does compete, but it does not do better. Just add the clock speeds up to 2GHZ and above, and it explains easily why the P4 in overall is better at Q3. But stating technically that P4's bandwidth or whatever makes it better than Athlon in Q3, is not entirely true at all, it's just clock speed in the end that wins.
I would assume that the on-die mem controller is what right now has made it THIS much competitive against a processor that is 2 times the clock speed than CH.

That is also IMO why the Internet Content Creation was Athlon's bad thing, it's mainly a clock speed dependant bench. Proof lies in the OCed 1.9GHZ AXP vs P4 1.9GHZ or 2GHZA, where it was competing the 2GHZA, and should be equal to a Willy 1.9GHZ's performance, even with a higher FSB due to OC.

--
Meow<P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 06/08/02 04:15 PM.</EM></FONT></P>
June 8, 2002 8:11:40 PM

Texas techie's sources stated that the more cache the CH has, the more it helps. IIRC he said that 1MB makes a huge difference. But I can't confirm, nor can he really?

--
Meow
June 8, 2002 8:12:47 PM

I agree it doesn't prove anything, but the point is that it has become like a 3d Mark bench, not for FPS but to see the average % increase. It's must easier to use it for that purpose, than to expect to play at 400FPS and be excited!

--
Meow
June 9, 2002 4:51:10 AM

What is the effective speed of HyperTransport?

That would be the MHz speed and what is the effective bit-size of HT?

Since the MCH is on-chip and the interface is HyperTransport to the memory modules, how can it have a FSB for the memory be anything other than a percentage of core clock speed?

Back to you...

<b>"Sometimes you can't hear me because I'm talking in parenthesis" - Steven Wright</b> :lol: 
June 9, 2002 12:06:33 PM

Quote:

than to expect to play at 400FPS and be excited!

Personally, I'm sick and tired of people using Quake 3 as a benchmark! Who cares about a 3-year-old game that any modern PC can run perfectly?

:wink: <b><i>"A penny saved is a penny earned!"</i></b> :wink:
June 9, 2002 12:14:22 PM

Well my main issue with looking at Quake 3 benches and saying "This is the improvement over this OTHER processor." It's ONE benchmark test ONLY. OVERALL performance will vary depending on what the CPU is doing and which benchmark tests are performed. SO, overall performance improvement will likely differ quite a bit from the Quake 3 test.

Mark-

<font color=blue>When all else fails, throw your computer out the window!!!</font color=blue>
June 10, 2002 7:18:13 PM

Quote:
<i>Eden doesn't say:</i>
Well just keep in mind all, overall Quake 3 is still running better on the P4. All the Hammer is doing is beating it by high IPC, that is all really.


Reread that, and tell my why one is right and the other is wrong :smile:

Quote:
<i>iib says:</i>
now plese stop with this disrespect for clockspeed. it is JUST and as importent and meaningful as IPC.


I agree, except for the fact that a higher clock speed tends to produce more heat than a higher IPC. (When dealing with similar processors, i.e. not talking about a 486 and an Itanium)

Quote:
<i>texas_techie says:</i>
clockspeed is worthless, I wanna know how much work its doing.


So your point is that even if the P4 is 50% faster in all benchmarks, it's magically slower because it has a lower IPC?
(Obviously the P4 isn't 50% faster, I'm just exaggerating to make my point)

Quote:
<i>Eden says:</i>
Texas techie's sources stated that the more cache the CH has, the more it helps. IIRC he said that 1MB makes a huge difference.


I thought he said the opposite, that more cache wasn't worthwhile because of the onboard MCH? Or was that someone else?

Quote:
<i>AMD_Man says:</i>
Personally, I'm sick and tired of people using Quake 3 as a benchmark!


I agree, so do lots of people. I can understand including games based on the Q3 engine (Jedi Knight II, MOHAA, etc), but not the original game.

<font color=blue>Hi mom!</font color=blue>
June 10, 2002 8:49:07 PM

I don't get what you want me to reaffirm?

The P4, disregarding the 533MHZ FSB one (since nobody tried to downclock them to 1.8GHZ at least, to see if per clock it has finally become over an Athlon in Q3), is not better per clock than the Athlon. Why does the P4 win here? Clock speeds. A 1.6A is not enough for example to compete the 1.6GHZ XP, however when it reaches speeds of 2.4GHZ, the PR rating can no longer keep up, as it is not a clock speed thing anymore here.

So the only way Hammer can possibly compete Q3, is by dramatically increasing its IPC, which it has and according to the tests, a 256K ClawHammer needs twice less clock speed to beat a 1.6 Willy. I can only deduct that the CH at 512K L2 will easily compete the NW A. As for NWB, who knows really, that's still an early CH sample.

So techniclly speaking, a 2GHZ CH, should be able to compete a 4GHZ NW in Q3. They should have tried the NV 15 demo however, it is much more stressing and would show how much the CPU and bandwidth is squeezed, there shows the 533MHZ prowess.

I am sorry if some consider this the clock speed vs IPC argument. IF I said something like IPC is more important than clock speed, then feel free to butt in and shut me up. But some people have become real asses when someone mentions the word "per-clock". How else can you know which chip consumes more power?
Fat can you provide me a board, where I can see a 1.8 Willy vs a 1.8GHZ XP so I can really see if those things consume more power and heat?

And yeah Phelk said more cache may not do better, but texas said the opposite.

--
Meow
June 10, 2002 10:28:29 PM

Quote:
I am sorry if some consider this the clock speed vs IPC argument. IF I said something like IPC is more important than clock speed, then feel free to butt in and shut me up. But some people have become real asses when someone mentions the word "per-clock". How else can you know which chip consumes more power?


Without clockspeed there is no such thing as I.P.C.LOCK/Cycle.

Therefore both are equally valid.

:wink: The Cash Left In My Pocket,The BEST Benchmark :wink:
!