AMD 45nm Deneb Pictures, CPU-Z and Super Pi Results

Printable View

Show 100 post(s) from this thread on one page

07-11-2008, 07:35 AM
GoThr3k

Quote:

Originally Posted by BrowncoatGR

These should be DVT samples at this point. The results look very good but i'd like to see some real benchmarks. I disagree that prefetch is a problem for AMD though.

someone willing to explain the differences between the prefetchers of AMD and Intel?
07-11-2008, 07:35 AM
demonkevy666

Quote:

Originally Posted by Calmatory

Well, isn't P3 based on P2 anyway? And P2 is based on Pentium Pro which is based on original Pentium? Larrabee having multiple smaller cores which are very alike to original Pentium cores?

So whats the problem with AMD, if not the architecture? Too low yields? Too high costs? Too bad fab process? Whatever it is, it's killing AMD. :|

Over paying for ATI is what killing them.
07-11-2008, 07:39 AM
Caveman787

No at this point getting ATI was a good move because there gonna dominate the gpu industry thanks to price/performance.
07-11-2008, 07:41 AM
demonkevy666

Quote:

Originally Posted by drfedja

Maybe AMD engineers little tweaked K10 core with load ahead of store and they finally done right with OoO load execution.

Acutally it has 2 pipelines. One integer, and one floating point with separate schedulers for integer/memory and floating point pipeline. These pipelines has 12 stages for integer and 17 stages for floating point.
K10 has 3 integer units, 3 AGU units and 3 FP units.
Core 2 has also 3 128-bit wide FPU ,3 ALU and 2 AGU units, but it can decode up to 4 instructions per clock, usually 3 arithmetic + one memory, because Intel PR machinery says that they have 4-way CPU, and other guys haven't. :d
I can't tell how much impact on performance has 4-way decoder on Core 2, but I am sure in that the main reason for good performance is good prefetch in combination with big L2 cache. Also, low latency L2, inclusive architecture, good branch predictor and high associativity L1 may bring few percent of performance.

I agree with that, but it also needs better prefetch. When CPU needs data from memory IPC factor drops significantly. With better prefetch algorithms CPU can access data from L1, L2 or L3 without that significant performance drop. If CPU hasn't that smart algorithms IMC or better IMC can't help there.

hell then get a 16way L1 cache with 16 256 bit SSE1,2,3,4 preclock precore. lol
07-11-2008, 07:42 AM
GoThr3k

Quote:

Originally Posted by demonkevy666

Over paying for ATI is what killing them.

buying ATI did cost them too much money indeed, but it was needed to buy ATI, for platfrom, CPU+GPU and all other stuff.

The loss AMD is making is basicly because of their bad CPU business, they may have lost some money on ATI, but that is no excuse for all their failures
07-11-2008, 07:42 AM
Rammsteiner

Quote:

Originally Posted by informal

New AMD WR has been made :D.
The dudes who brought us these pictures from itocp.com did it :
http://www.itocp.com/redirect.php?ti...tpost#lastpost

All it was needed is 3484Mhz on Deneb to break it.Cooling was AIR.

New 45nm Phenoms are still not recognized properly so they show up as 9850.
The score is exactly 12.5% better clock/clock than tweaked Agena core (MILANS did the next best K10 result,took him 3621Mhz,with tweaked L3@2.1Ghz and low latency ddr2 to get 21.625s score).

Damn, it beat the K8 heavily as well there:eek:. w00t:cool:
07-11-2008, 07:43 AM
Wiggy McShades

Quote:

Originally Posted by gOJDO

What K10 needs is higher core, L3 and NB clocks, lower cache latencies, wider cache and some architectural improvements.

oddly enough on my phenom anything past 2.5ghz on the nb/L3 operating speed does nothing for any test or benchmark.
07-11-2008, 07:45 AM
GoThr3k

Quote:

Originally Posted by Wiggy McShades

oddly enough on my phenom anything past 2.5ghz on the nb/L3 operating speed does nothing for any test or benchmark.

at what CPU speed?
07-11-2008, 07:47 AM
informal

Quote:

Originally Posted by Rammsteiner

Damn, it beat the K8 heavily as well there:eek:. w00t:cool:

It utterly destroyed K8 in SPi :D.Around ~17.5% faster than K8.Deneb brought us a lot more than Agena did in this particular test.
So all they needed was 3.48Ghz to break a WR that needed almost 4Ghz 65nm K10 chip to do it :D.
Being an ES still means this is not quite "tweaked" from BIOS perspective.Par this thing up with upcoming SB750 motherboard and 2 4870X2 card,what you get is all-amd-uber-system :).K10.5 is starting to look a lot like RV770 ;).
07-11-2008, 07:54 AM
demonkevy666

Quote:

Originally Posted by informal

It utterly destroyed K8 in SPi :D.Around ~17.5% faster than K8.Deneb brought us a lot more than Agena did in this particular test.
So all they needed was 3.48Ghz to break a WR that needed almost 4Ghz 65nm K10 chip to do it :D.
Being an ES still means this is not quite "tweaked" from BIOS perspective.Par this thing up with upcoming SB750 motherboard and 2 4870X2 card,what you get is all-amd-uber-system :).K10.5 is starting to look a lot like RV770 ;).

well Wiggy McShades

that is odd I get a boost with my north bridge up 200mhz I get about 5-7% faster.
07-11-2008, 08:06 AM
Morais

I just hope they fixed the power consumption on these procs
07-11-2008, 08:14 AM
Sparky

Quote:

Originally Posted by Morais

I just hope they fixed the power consumption on these procs

Everyone complains about the power consumption but considering that my opteron 165 drank almost 180W at 2.8GHz, 4 cores at 2.6GHz and 140W doesn't sound that bad :p:

OK all joking aside, yes it would be nice. I'm curious as to how much it uses and how warm it runs.
07-11-2008, 08:26 AM
Donnie27

Quote:

Originally Posted by KeZzZu

yeah but 3ghz phenom beated 3ghz c2q in 3dmark? if i remember right

we need more other test's than just superpi, i would like to see cinebench

Yepp, real software is where AMD will also look better. I lost faith in most synthetic stuff a long time ago.
07-11-2008, 08:32 AM
informal

Hmm ok i see a trend here,regarding some of the comments.
AMD shows a significant improvement in a such a poor "benchmark" as SPi is,and now most people complain about TDP,even though this is C0 stepping,not the production one.
IF AMD showed zero IPC increase,even in poor SPi, BUT lower volts at those clocks we've seen,i bet most of the people that complained about TDP/power consumption would switch over to a no-IPC improvement argument.
My point is,which ever way you look at some of the comments,AMD just can't do right in some people's minds :rolleyes:

What we have here is an ES(C0 stepp.),a DVT sample back from April,that is around 12% faster in one poor synthetic benchmark(but thankfully we know that even in this one there is a great IPC increase),that clocks better than any 65nm Phenom and does this on a SB600 motherboard.All probably done on a beta BIOS and with low NB/L3 clocks.
07-11-2008, 08:34 AM
Sparky

Yeah it does look good for a beginning for sure. I am very interested to see how this all plays out.
07-11-2008, 08:36 AM
KeZzZu

Quote:

Originally Posted by informal

Hmm ok i see a trend here,regarding some of the comments.
AMD shows a significant improvement in a such a poor "benchmark" as SPi is,and now most people complain about TDP,even though this is C0 stepping,not the production one.
IF AMD showed zero IPC increase,even in poor SPi, BUT lower volts at those clocks we've seen,i bet most of the people that complained about TDP/power consumption would switch over to a no-IPC improvement argument.
My point is,which ever way you look at some of the comments,AMD just can't do right in some people's minds :rolleyes:

What we have here is an ES(C0 stepp.),a DVT sample back from April,that is around 12% faster in one poor synthetic benchmark(but thankfully we know that even in this one there is a great IPC increase),that clocks better than any 65nm Phenom and does this on a SB600 motherboard.All probably done on a beta BIOS and with low NB/L3 clocks.

yeah it's great results thought, id like to see how much that cpu has tdp rate
it really doesnt matter for me, but id like to see below 100w, if it get like 65w tdp... deneb really delivers then. with sub 200 euros price it would be superb in my opinion.
07-11-2008, 08:42 AM
Sparky

Yes yes we know that intel does better at superpi than AMD, they always have. But superpi is not a good indication of performance between the two. The only thing it is good for IMO is getting a little bit of an idea how a CPU compares to its predecessors, in the same company. Cross company comparison is not valid. Great example is P4 vs K8, P4 was faster in superpi but lost overall compared to K8.
07-11-2008, 08:56 AM
gallag

Quote:

Originally Posted by informal

Hmm ok i see a trend here,regarding some of the comments.
AMD shows a significant improvement in a such a poor "benchmark" as SPi is,and now most people complain about TDP,even though this is C0 stepping,not the production one.
IF AMD showed zero IPC increase,even in poor SPi, BUT lower volts at those clocks we've seen,i bet most of the people that complained about TDP/power consumption would switch over to a no-IPC improvement argument.
My point is,which ever way you look at some of the comments,AMD just can't do right in some people's minds :rolleyes:

What we have here is an ES(C0 stepp.),a DVT sample back from April,that is around 12% faster in one poor synthetic benchmark(but thankfully we know that even in this one there is a great IPC increase),that clocks better than any 65nm Phenom and does this on a SB600 motherboard.All probably done on a beta BIOS and with low NB/L3 clocks.

You think this is bad, You want to check out some of the "not happy" comments in the nehlam thread. A lot worse.

Here are a few, This one in particular is funny considering you are jizzing over AMD pi scores

Quote:

Originally Posted by informal

Around 10% faster than Penryn @the same clock,not bad but not earth shattering.We need something else apart from useless Spi1m(wPrime ie. where multicore shines)

Quote:

Originally Posted by informal

Shintai,Pinacolada summed it up well here:
Quote:
So until you really need new computer or you are into multithreaded apps upgrade to quadcore Nehalem does make sense.Otherwise you will pay 100% more for about 10-20% improvement - not worth IMO for typical user it's much wiser to upgrade GPU or wait for cheap SSDs.
Quadcore penryn is good enough to wait till 2H 2009 for 32nm shrink where price/performance ration against Penryn will be much better.Most probably octalcores will be available then.

People are expecting marvels from Nehalem but it will be more evolution than revolution.

Quote:

Originally Posted by informal

It was a quote from another user.Get it?!This is so amusing.

Quote:

Originally Posted by informal

Is it just me or the BOINC results are a little disappointing,being just on par with C2Q@65nm clock for clock?Or there is more than meets the eye?

Quote:

Originally Posted by informal

BOINC shows total scores(per CPU package),at least that's what i understand.Someone please confirm/deny this.
Now if this is the case,scale down the Kentsfield score to 2.93Ghz and you end up @ pretty much the same score with Nehalem,clock/clock.

I've looked at tomshardware.tw preview of Nehalem scores,and especially at Crysis CPU test.Managed to find a database of C2Q scores for the CPU test 2 at techarp:
http://www.techarp.com/showarticle.a...tno=499&pgno=3

Nehalem @2.93 scores:

while QX9650 gets around 18.7fps @1280x1024 noAA according to techarp database.Wolfdale e8400 even scores better than Qx9650with ~18.9fps in CPU test2.
Anyone knows what are the scores for QX9650 in CPU test 1 in Crysis bench tool @1280x1024?

Quote:

Originally Posted by informal

Hmm i think that JC was disappointed with BOINC since the scores where lower than Movieman expected.
Anyone else care to post their thoughts on BOINC benchmark results?

Quote:

Originally Posted by informal

Hmm exactly,something is off with this bench..Either HT isn't working correctly or it is working stellarly and is producing 2x the scores :confused::confused: ..Which in turn is highly unlikely (how can a HT enabled Nehalem chip score 2x the physical scores with HT enabled??).We need actuall points per day the Nehalem puts out,not the built-in tool to benchmark it.The actual ppd will tell us is the HT really this good or is it messing with the scores.

Hypocrite much? Perhaps instead of wondering why people are not jumping up and down with excitement and posting only positive things about 45nm phenom you should ask yourself why you where so negative in the nehlem thread?
07-11-2008, 09:01 AM
informal

Maybe because Deneb was supposed to be a shrink of 65nm Phenom while Nehalem is a whole new uarchitecture?Think before posting.Yeah.Thanks.

BTW,Nehalem in MT is a beast,this is not a question.
07-11-2008, 09:03 AM
eleeter

Quote:

Originally Posted by gallag

Hypocrite much?

I didn't think he was being hypocritical at all, and was pretty much spot on with his comments.
07-11-2008, 09:05 AM
gallag

Quote:

Originally Posted by informal

Maybe because Deneb was supposed to be a shrink of 65nm Phenom while Nehalem is a whole new uarchitecture?Think before posting.Yeah.Thanks.

BTW,Nehalem in MT is a beast,this is not a question.

Dont change the argument:), You got annoyed because people where not singing and dancing about these results and where questioning how they could bring any negative comment into it and I was just pointing out that every comment you made in the Intel thread was negative. Simple as that.
07-11-2008, 09:12 AM
informal

Quote:

Originally Posted by gallag

Dont change the argument:), You got annoyed because people where not singing and dancing about these results and where questioning how they could bring any negative comment into it and I was just pointing out that every comment you made in the Intel thread was negative. Simple as that.

Change the argument?Evading much now?
The point is that Nehalem is a whole new uarch. while Deneb was supposed to be just a shrink.I never "jizzed" over SPI scores as you put it,i even said it's a lame benchmark.

BTW my comments are not negative in Nehalem thread.You are from N. Ireland,right?You speak English?Read my comments again and point where i was wrong?Boinc benchmark tool showed us some amazing score so i questioned them and said we need a real world ppd not a tool to bench HT.
Crysis was also there so i found a techarp database and looked at scores C2Qs put out,whats wrong with that??
The only really exciting score i saw was the Vantage CPU score which is really amazing.All other questions i asked were valid.And remember it was the op in the Nehalem thread(JC) that said "far from your expectations" for some of the results,not me.Is he a hypocrite?
07-11-2008, 09:15 AM
terrace215

Quote:

Originally Posted by informal

Maybe because Deneb was supposed to be a shrink of 65nm Phenom while Nehalem is a whole new uarchitecture?Think before posting.Yeah.Thanks.

BTW,Nehalem in MT is a beast,this is not a question.

Deneb *is* basically a shrink. Some very minor tweaks, and the larger L3 cache. It's the bigger L3 which is responsible for the sPi improvement over Phenom at the same clock. 1M sPi has an 8MB footprint, right? So now 3/4's of it fits in cache vs 1/4 on Phenom. That probably explains just about the entire speedup right there.
07-11-2008, 09:16 AM
[XC] hipno650

i am still not impressed with AMD here. the power consumption and heat with the vcore's being used is still going to be sky high. and the performance will still be behind Kentsfield clock per clock not to mention yorksfield and what lies beyond. they need an extra 20% boost to get up to competitive levels of performance clock per clock with intel and even if they got that far they still can't over clock. ohhh 500mhz overclock big whoop. my Q6600 can do a 1.5ghz overclock. so my Q6600 for $210 can do 3.9ghz on air and be faster clock per clock and this may not even hit those speeds on LN2 and will be more expensive.

i agree that it is good for them to be making progress but they are still hurting in performance.
07-11-2008, 09:17 AM
Kasparz

Guys, guys, take it easy. We all know that SuperPi is Intel bench. Even in Netburst vs K7/K8 times netburst could outperform AMD in SuperPi, but in other benches/realtime applications netburst was way behind. So even with this AMD improvement in SuperPi, if we compare to equal 65nm Agena, Deneb shows major improvement. Compared to 3.6Ghz very tweaked Agena platform, Deneb is still way faster. So if AMD brings out new SB and Deneb BEFORE LGA1160 platform, they can earn some sweet bucks and get back in business. And as we know only thing that basically held back Agena OC is HTT overclocking and skyhigh power consumption and load for power phases. 45nm will solve this, so i believe 4GHz is achievable with Deneb.
Intel needs this move from AMD, we need this move, whole IT industry need this move. As fast as AMD will make very competitive product for mainstream, as fast Intel will reply with Lynnfield. And we need some healthy competition, right?

Show 100 post(s) from this thread on one page