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

[cadaveca]

No i didn't ,, sorry no offense ,, all i meant is that you keep tweaking , testing so results may vary a lot so it is not apples to apples ,,, we are not talking about a "run 3D mark 06 test " guy

Well, unfortuantely, that's just what I did, install new driver, run tests. That's why i modded the vid bioses..to ensure clockspeeds and the like weren't a contributing factor.


Now, currently, yes, dependant on chipset, 3ghz Core2 will score MORE...but we are talking a mere 400-500 points only, which, in turn, means that the Core2 system pulls ahead, albeit marginally(assuming that clockspeeds and overall bandwidth remain similar, as they are in the benches i posted).

But i chose those benches for a reason...the driver used for Core2 is one edition higher than A64.

So, like-wise thinking...Theo is running a pre-release driver, vs the quoted Core2 scores that use current. big deal.

The fact remains that K8, clock for clock, is not far behind Core2, in 3Dmark. Keeping this in mind, to not expect signifigant leaps in perforamnce from a new AMD platform is a little bit close-minded. I think peopel are just far to used to seeing 5ghz Core2 3Dmarks...but noone has posted 5ghz A64 3DMarks, so no real comparison can be made.

[madcho]

I think Phenom is going to be good (at least it better be or AMD is done) but I highly doubt it at 3Ghz is going to out do a 5Ghz Kentsfield with the same hardware. I do hope in fact that is is faster because competition is good for everyone. It drives prices down and forces the competitors to develop even faster hardware. And in the end, it's us the consumers who will reap the rewards

5ghz kentfield is not 100% faster that a 2.5ghz kentfield. It's false. More you increase, less the the cpu is faster/mhz.

...

[Gautam]

Sampsa, can you verify that it's an 05 score? I've done a bunch of runs in the mid/low 4.6 gig range that I wish I could show for a direct comparison. If it's a true 05 score done at 1024x768 and default settings then K10 is still sounding very strong clock for clock. Unfortunately the lowest score I have on hand right now is this one ...but I'll see if I can dig up some screenies of 30.6k runs I've got at 4.7GHz. Either way we're looking at a large difference in clock for clock power, as, at 3 GHz, the best I can muster is in the ~20k range....

[w0mbat]

Stop acting like idiots. It's 3DMark05 score, not 3DMark06 like Fugger already said.

Even if it´s 2k5, k10 is much faster than core2.

[CraptacularOne]

5ghz kentfield is not 100% faster that a 2.5ghz kentfield. It's false. More you increase, less the the cpu is faster/mhz.

...

Thank you Captain Obvious! ......Where did you see anything about that?

[Vapor]

30k in 05 at 3GHz is almost harder than in 06, no?

If either are true, this CPU is a beast beyond what any rumor said.

[Stapler]

30K is a 2K5 score

Theo is mistaken

Quoted for emphasis.

30k on air though is still really impressive.

[CraptacularOne]

30k in 05 at 3GHz is almost harder than in 06, no?

If either are true, this CPU is a beast beyond what any rumor said.

Lets not also forget the weight that ATI cards pull in the 05 benchmark either, I have a pair of X1950XTX's that can almost keep up with my 8800s in 05

[kemo]

Well, unfortuantely, that's just what I did, install new driver, run tests. That's why i modded the vid bioses..to ensure clockspeeds and the like weren't a contributing factor.


Now, currently, yes, dependant on chipset, 3ghz Core2 will score MORE...but we are talking a mere 400-500 points only, which, in turn, means that the Core2 system pulls ahead, albeit marginally(assuming that clockspeeds and overall bandwidth remain similar, as they are in the benches i posted).

But i chose those benches for a reason...the driver used for Core2 is one edition higher than A64.

So, like-wise thinking...Theo is running a pre-release driver, vs the quoted Core2 scores that use current. big deal.

The fact remains that K8, clock for clock, is not far behind Core2, in 3Dmark. Keeping this in mind, to not expect signifigant leaps in perforamnce from a new AMD platform is a little bit close-minded. I think peopel are just far to used to seeing 5ghz Core2 3Dmarks...but noone has posted 5ghz A64 3DMarks, so no real comparison can be made.

Surprisingly i am still on K8 coz they are not far behind like you said
PS: Theo is not running anything

Even if it´s 2k5, k10 is much faster than core2.

i wont use this word till real benches come out

[Periander6]

Sampsa, can you verify that it's an 05 score? I've done a bunch of runs in the mid/low 4.6 gig range that I wish I could show for a direct comparison. If it's a true 05 score done at 1024x768 and default settings then K10 is still sounding very strong clock for clock. Unfortunately the lowest score I have on hand right now is this one ...but I'll see if I can dig up some screenies of 30.6k runs I've got at 4.7GHz. Either way we're looking at a large difference in clock for clock power, as, at 3 GHz, the best I can muster is in the ~20k range....

We would need a quad core for comparison, not a dual core like what you got on that screen.

[Gautam]

30k in 05 at 3GHz is almost harder than in 06, no?

If either are true, this CPU is a beast beyond what any rumor said.

That's actually a pretty reasonable statement. I would say between 3-3.5 with a Kentsfield, you could get 05 and 06 each scoring around 20k, but with heavy enough card clocks (subzero for sure) to feed the 06 well....whereas they would do jack for the 05 at such speeds. Even at stock you'd be pretty close to being CPU bottlenecked. But then again, that wouldn't be the case if this CPU is as powerful as it is being made out to be here...

[CraptacularOne]

We would need a quad core for comparison, not a dual core like what you got on that screen.

05 is not multi threaded so it doesn't matter how may cores you have

[Vapor]

We would need a quad core for comparison, not a dual core like what you got on that screen.

Quads don't help in 05 more than a few % (if that).

30k in 05 or 06 on all-air is INSANE.

[Gautam]

30k in 05 or 06 on all-air is INSANE.

QFT.

True or not, this article sure is drawing quite a crowd to the News section.

[CraptacularOne]

QFT.

True or not, this article sure is drawing quite a crowd to the News section.

Well everybody loves a good run though the hay...lol

[HopJokey]

Hey everyone, this could be true, this could be false. What does it mean? Who knows. Let's just wait and see (not that long of a wait) because after all this article was written by the same person who reported this:

http://www.theinquirer.net/?article=32589

[Frank M]

I suggest a mod locks this thread, as we have no other data apart from the inq,
which is not a reputable source; and the situation here is getting a bit messy.

[Periander6]

05 is not multi threaded so it doesn't matter how may cores you have

Ah, I see. Thanks.

[CraptacularOne]

Hey everyone, this could be true, this could be false. What does it mean? Who knows. Let's just wait and see (not that long of a wait) because after all this article was written by the same person who reported this:

http://www.theinquirer.net/?article=32589

LMAO @ reverse hyperthreading.....this whloe thread is basically

[Sampsa]

Sampsa, can you verify that it's an 05 score?

Sorry I can't verify but lets do quick math:

http://service.futuremark.com/compare?3dm05=3266193

When you change graphics card clocks from 743/828 to 830/900, switch old Asus P5W64 WS motherboard to AMD RD790 reference board with PCI Express 2.0 support and 2 x PCI Express X16 slots and K10 architecture @ 3 GHz, I think the score will be pretty close to 30k in 3DMark05. (edit: or 25k :P)

30k in 3DMark06 would require insane CPU score.

And remember, the source is The Inq.

[RedBull78]

3Ghz X2? More like closer to 4 ghz. I can't remember exactly what good ol' MM had on his FX for his old 20k WR, but it could be crushed on air and stock voltage by the lowest C2D without even trying, even with the cards on air.

4GHZ? Dude, pass some of that stuff around.

[Vapor]

I suggest a mod locks this thread, as we have no other data apart from the inq,
which is not a reputable source; and the situation here is getting a bit messy.

What, we can't dissect what it takes to get 30k and what that means for (potential) CPU power?

There may very well be more to it though....triple/quad R600 and lowered resolutions come to mind first.

[CraptacularOne]

What, we can't dissect what it takes to get 30k and what that means for (potential) CPU power?

There may very well be more to it though....triple/quad R600 and lowered resolutions come to mind first.

Ahh, good point about the more than 2 GPUs....I remember them touting that a while back....hmmm maybe with 4 R600s...... never the less, it is a possibility

[cky2k6]

can't wait to see the truth behind all this. 30k in 05 is really pretty much just as hard though... i think there is more to this, cause no matter how good k10 is, its really hard to imagine it being that much faster, the difference would be greater than k8 vs p4...

[Gautam]

Sorry I can't verify but lets do quick math:

http://service.futuremark.com/compare?3dm05=3266193

When you change graphics card clocks from 743/828 to 830/900, switch old Asus P5W64 WS motherboard to AMD RD790 reference board with PCI Express 2.0 support and 2 x PCI Express X16 slots and K10 architecture @ 3 GHz, I think the score will be pretty close to 30k in 3DMark05.

30k in 3DMark06 would require insane CPU score.

And remember, the source is The Inq.

You're a sly dog Sampsa diggin' up my score like that. I can say this much, there's no way I could see 30k happening on either P35 (w/ 8x/8x of course) or 975X at anything below ~4.4GHz. Regardless of card clocks. (Even at 4.6GHz+ '05 is very very CPU biased, even stock card clocks perform decently) On that platform it takes me past 4.6. I have no idea how X38 will fare.

30K in 05 under any circumstances at 3 GHz impresses me. Whether it's true or not remains to be seen.

Fun thread this is, quite a curveball this threw at XS.