AMD's 3GHz K10 to break 30,000 3DMark06 (Inq)

[Theli]

Enough crap. Let's show the official spokespeople.
AMD's Quads to outperform Intel's by over 40%
http://www.youtube.com/watch?v=G_n3wvsfq4Y
3GHZ demo is TRIFIRE:
http://www.youtube.com/watch?v=R7EZmYth6TM
So, unless these people are in fact lying(hence the Fudzilla comment yesterday), maybe it is far more possible than most think, and then again, maybe not.

Why would anyone in any kind of official position lie?

[CraptacularOne]

I'm actually now more interested in their implementation of more than 2 GPUs in Crossfire. Since Nvidia's Quad SLI was a bust. Maybe they have figured out a better way to distribute the load between the GPUs than Nvidia did.

[BadNizze]

So 23k for 2*Xt and 30K for trifire not to shabby after all! Sounds plausibule to me, 2*16 full speed Pci-e and all that. Know where is my Rd790 Board? (Just kidding I don´t hava a Rd790 board...yet)

[cadaveca]

VictorWang yesterday posted singlecard 2900XT on core2 @ almost 30k in '05. I highly doubt the truth of these 30k '06 benches, seeing that.

Victor's thread is here:

http://www.xtremesystems.org/forums/...d.php?t=156898

So, victor might be able to get 45-50k in 05 on that system, I do not doubt he has the skill to pull it off. The more i dig, the better this is looking for the consumer!

[alfaunits]

Assuming this is true:
http://www.tweaktown.com/news/8067/w...emo/index.html
6400+ with 2x2900XT on 790G gave 12000 3DMark06.
Only the CPU change won't give 18K, but can the third card, and OCing the 3 cards give 12K even?

[savantu]

I'm not a cpu designer but I do however know a lot about the process of cpu design and manufacturing. It's my hobby. I love technology and innovation. I dislike "patches" and workarounds (all to an extend of course).

Anyway I'd like to share my 2Mhz.

IMO most posters forget one important fact. That is that noone knows how efficient a true quad core cpu is compared to a 2x dual core mcm solution.
Sure 30k is amazing and I also believe it was a trifire setup. But still it is not "impossible".

Intel and AMD know; according to Intel the difference is negligible.AMD OTOH claims native is better , doesn't say by how much.
I expect the truth to be in between.

Let's look at the current X2s and Conroe.We can call the K8 X2 as the best MCM-like aproach; 2 cores which are connected by an outside link.
C2D uses a shared L2 , the most "native" aproach.

One must know the disadvantages AMD's K8 has over intel's Conroe. And then compare that to the K10. Most of Conroe's advantages are taken care of by the K10. One of K8's biggest problem is the "interchip bandwidth". The memory throughput is bottlenecked. That's why you don't see much efficient use of DDR2's bandwidth on K8's.

Time and time again K8 showed it was in no way bottlenecked by its memory sysytem.It only cares about low latency.

Almost every bottleneck has been taken care of. And remember we are talking about A Quad CPU Core, not a Quad Core CPU It's 1 Core with 4 CPU Units sharing 1 L3 Cache of 2MB. So who can tell what performance advantage you can get? It's probably not much in some benchmarks but it could make a difference in ohters. Also the memory controler seems to be clocked either higher or slower than the core it self.
That will also mean that the system memory could be able to run at default clock no matter the cpu clock. There are many more enhancements that could make the K10 perform well beyond expectations.

K10 major flaw is in its cache size ; 512KB is small and the L3 will be a bottleneck.Also its latency if true ( 38 ) is horrible compared with K8/C2D/Penryn ( 12-14 ) .Latest reports put Penryn's massive L2 at 13 cycle latency which is nothing short of astounding.

But to get the most of a K10 software must be recompiled. To make fully use of the FPU you don't have any choice anyway.

Umh , not exactly.The changes made will offer sizable benefits for existing code.

[grunge100]

savantu,

What is AMD's L2 latency on K10

[T2k]

Hmmm...

K10 x RD790 x X2900XT CF x engineered and optimizated from the ground up within one Co. that has a proven track record for innovation = 30K 3DMark06? It's not impossible!

Hey... get away from my dream... it's all mine... *walks away laughing sinisterly*

2900 has almost literally NOTHING to do with AMD. Its design stems from former ATI and was done well before the merge.

[BeardyMan]

savantu,
i didn't knew intel and amd were using the same caches?

[T2k]

And you just bought yourself a 3 day vacation for your nasty mouth.
Enjoy!

I'm actually now more interested in their implementation of more than 2 GPUs in Crossfire. Since Nvidia's Quad SLI was a bust. Maybe they have figured out a better way to distribute the load between the GPUs than Nvidia did.

Wow, these XtremeDays go fast.

[CraptacularOne]

Wow, these XtremeDays go fast.

Please move on, there was a simple misunderstanding is all.

[xenolith]

2900 has almost literally NOTHING to do with AMD. Its design stems from former ATI and was done well before the merge.

Designs go through several iterations before production. Still, point taken.

[NapalmV5]

^wow

goes to show what fanaticism does to people lol

the guy clearly stated: 1 x quad barcelona @ 3GHz + 2 x 2900 oced = 30K

wonder what ~4ghz does.. 4x4 @ ~3.5ghz

[SOLDNER-MOFO64]

I'm just happy that AMD are finally gonna release K10 very soon. My benchmarks are beginning to look like some from my AXP days....

Can't wait to get my hands on one

Intel's CSI sounds interesting......and regardless of mistakes made in the past...Intel don't seem to make many blunders these days.

[CraptacularOne]

^wow

goes to show what fanaticism does to people lol

the guy clearly stated: 1 x quad barcelona @ 3GHz + 2 x 2900 oced = 30K

wonder what ~4ghz does.. 4x4 @ ~3.5ghz

Yeah and he was also the guy that told us about "reverse hyperthreading" I think that guy is just a few cards short of a full deck if you ask me......

[NapalmV5]

Yeah and he was also the guy that told us about "reverse hyperthreading" I think that guy is just a few cards short of a full deck if you ask me......

i cant believe youre making me defend him, lol

do you really think he made that up outta the blue?

[TigeriS]

Assuming this is true:
http://www.tweaktown.com/news/8067/w...emo/index.html
6400+ with 2x2900XT on 790G gave 12000 3DMark06.
Only the CPU change won't give 18K, but can the third card, and OCing the 3 cards give 12K even?

Why not? I think it's a combination of both. How exactly does 3DM06 calculates to the cpu score? Does the L3 give the score an extra boost? Does the wider datapath give it an extra boost? Or is the redesigned memory controler more efficient with memory bandwidth? Perhaps it's the native quad core that works more efficient. Well I've never digged into 3DM06's calculations so I don't know.

My point is that there are so many unknown factors that it's not easy to say if it's "impossible" or not on this new platform with only 2 cards. As said before, Conroe is not much faster than the K8 in 3DM06.

So until we get a confirmation all we can do is speculate about it.
One thing is sure, AMD's new platform looks promising. Even if 3 Cards is needed for those scores, you bet it will be a success for AMD.

True or not, TriFire or not, it doesn't matter. Many are now even more interested than ever and all focussed on AMD. Penryn? Intel gave too much of it away already, there's nothing "new" about it anymore. In fact some are disappointed by it's performance. So even if Phenom turns out to be 5% faster, most will go for Phenom and the TriFire option.

I'm more interested in the performance between AMD's chipset and nVidia's.
On this new platform.

[CraptacularOne]

i cant believe youre making me defend him, lol

do you really think he made that up outta the blue?

All I'm saying is that his "reports" haven't exactly panned out if you know what I mean The INQ will post anything and they continue to do so. I don't even know why people even bother to respond to their "reports" like they have any validity. I'm sure we'll all find out that it was a system using 3 (or more) GPUs to achieve that score in 06.

[NapalmV5]

All I'm saying is that his "reports" haven't exactly panned out if you know what I mean The INQ will post anything and they continue to do so. I don't even know why people even bother to respond to their "reports" like they have any validity. I'm sure we'll all find out that it was a system using 3 (or more) GPUs to achieve that score in 06.

if turns out it was all a lie.. they clearly stated 2 x 2900.. if its more than 2 x 2900, INQ should get buried

[CraptacularOne]

if turns out it was all a lie.. they clearly stated 2 x 2900.. if its more than 2 x 2900, INQ should get buried

I agree, and if (more like when) this all hits the fan, they should ban any links to the INQ from being posted by putting them on their filter of banned words/sites here at XS.

[Donnie27]

You are back in this topic now it seems.

Anyhow,2 weeks to go until we find out the truth.
Based on TechArp article,K10 will have 20-30% advantage against intel's counterparts(clock /clock) with excesses of 170% advantage in rare cases.
Sounds better than C2D did against K8.

Doesn't matter how K-10 does, Fanboys such as yourself will love it anyway LOL! What did you call it on that other thread?

[Motiv]

Regardless if it's true or not. I wouldn't be surprised if we see a few benchmarks accidentally being seen before next week is up.

[informal]

Doesn't matter how K-10 does, Fanboys such as yourself will love it anyway LOL! What did you call it on that other thread?

You just can help yourself,can you...

[Canibuz]

why do people always think oh no company in the world would keep such performance a secret they would make an intentional *leak* who knows maybe it broke 30k on 06 personally i think they mean 05 but why wouldn't a company like a complete surprise win? I would keep 30k in 06 a secret till release if i headed amd but then again, thats just my 2 cents

[Carfax]

K10 major flaw is in its cache size ; 512KB is small and the L3 will be a bottleneck.Also its latency if true ( 38 ) is horrible compared with K8/C2D/Penryn ( 12-14 ) .Latest reports put Penryn's massive L2 at 13 cycle latency which is nothing short of astounding.

I agree that Penryn's memory sub-system is incredible, but lets have a bit more faith in AMD.

One of the greatest strengths of the Athlon/Hammer and now Barcelona, is that it has a large L1 cache.

The K10 improves upon the L1 cache in Hammer by making it two port.

Also, Shanghai (the 45nm successor for Barcelona) still has the same L2 cache size, while the L3 cache is expanded to 6MB.

Because of this, it seems that the AMD engineers didn't think the size of the L2 cache in K10 was going to limit it's performance, otherwise they would have addressed it in Shanghai.