ATi R600 Rumored Specs & Benchmarks

[AthlonXP1800]

Be prepare for a huge disappointment if the benchmarks VR-Zone obtained turn out to be true then R600 may be doomed, the flagship X2800 XTX 1GB DDR4 wasnt really the monster with massive 140GB memory bandwidth and huge 240W power as you think. X2800 XTX 1GB DDR4 at 700/2200 is just 10% faster than 8800 GTX 768MB at 575/1800 so Nvidia already has overclocked 8800 GTX 768MB at 630/2000 that will topple X2800 XTX.

http://www.vr-zone.com/index.php?i=4648

ATi Radeon X2800 XTX
4way× (MADD(2flops) + MUL(1flop))× 800 MHz × 64 SPs ＝ 614.4GFlops

Core Clock: 700 MHz
Memory Clock: 1100 MHz (2200 DDR)
Memory Bandwidth: 140.8 GB/sec
Shader Operations: 87200 Operations/sec
Pixel Fill Rate: 19600 MPixels/sec
Texture Fill Rate: 19600 MTexels/sec
Vertex Operations: 15600 MVertices/sec

I guess 512bit memory interface are far too much for ATI to handle, look like Nvidia won DirectX 10 war this time due to competitive pricing and lower power consumption. It really very shame ATI took so long to developed R600 and launch it 6 months late.

[mullet]

I wish I knew what all those numbers meant. (feels dumb).

[Xion X2]

Also using Core 2 Extreme QX6700 @ 5300MHz (353.3×15), Radeon X2800XTX 1GB 512bit 0.85ns GDDR4 @ 700MHz/2200MHz, the 3DMark06 score stood at 15032 3DMarks (SM 2.0 = 5754 Marks/ SM 3.0 = 5841 Marks/ CPU Score = 7615 Marks).

Is this a misprint? Because on the chart they have "3dMark05."

[nekrosoft13]

ok so its faster, pretty much everyone was expecting it to be faster.

good news is that the margin in very small, i'm quite sure nvidia has another card right behing the corner to hit ati back,

[Xion X2]

The 2800XT is only 700 points better in 3dMark06 (that's obviously supposed to read 06 and not 05, since they're using a quad-core. I scored 17,200 w/ my Conroe and a single OC'd 8800GTX in 3dMark05). Not good for ATi. That's not going to make any sensible 8800GTX owner dump the green team and go red (or AMD green).

And I'll bet you can't clock the XTX much higher seeing as it's already sky high on the clocks as it is (700/2200).

[AthlonXP1800]

Is this a misprint? Because on the chart they have "3dMark05."

Yes it a misprint.

http://www.fx57.net/

They tested it with 3DMark 2005, not 3DMark 2006.

[Xion X2]

Yes it a misprint.

http://www.fx57.net/

They tested it with 3DMark 2005, not 3DMark 2006.

That just can't be. Because I scored 17,200 on 3dMark05 w/ my Conroe and a single GTX. And they were using a quad-core in the benchmarks.

A quad-core will get around 3-4,000 points more in 3dMark06 than a dual-core will, and I scored a 12,200 with my Conroe and a single OC'd 8800GTX in 3dMark06. That would scale much better.

I think the chart is supposed to read 3dMark06. I'm just wondering why they skipped '05. I mean, who the hell benches a brand new videocard on 3dMark03!? Stupid..

[slaWter]

Is this a misprint? Because on the chart they have "3dMark05."

I think it is 06 because of this:

Intel(R) Core(TM)2 Extreme QX6700 @ 5300MHz (353.3x15)
ATI Radeon X2800XTX 1GB 512bit 0.85ns GDDR4 @ 700MHz/2200MHz
3DMark Score 15032 3DMarks
SM 2.0 Score 5754 Marks
SM 3.0 Score 5841 Marks
CPU Score 7615 Marks

It's another test, the FX-55 was 03 and 05 and the QX6700 was 06 imo.

[Xion X2]

Yeah, the chart should read 06.

[Lfctony]

look like Nvidia won DirectX 10 war this time

Yes, there are so many DX10 games out and the R600 is performing real bad... WTH man, do you actually believe the crap you say? Or does your fanboy ass forget that while the 5800 and the 9700 were equal in DX8, the 9700 wiped the floor with the 5800 in DX9? Does your fanboy brain block those bad memories?

[slaWter]

Yeah, the chart should read 06.

I think the chart is correct (03 and 05). That was with a FX55.
The other 15k score with the Quad Core is the only 06 test.

But most likely these are fakes anyway ;)

[Xion X2]

Why do you say they're fakes? VR-zone got the scoop before anybody else on G80.

[slaWter]

Why do you say they're fakes? VR-zone got the scoop before anybody else on G80.

Don't know we'll see. I think these are realistic numbers but we have no real proof.

[retsam]

http://resources.vr-zone.com/newspics/Feb07/12/R600XTX-Retail.jpg

[Mr_LoL]

I would like to see how this thing performs in games not 3dmark 06.

[retsam]

I would like to see how this thing performs in games not 3dmark 06.
ya, i feel the very same way, the problem im having now is the power requirments. one 8 pin and one 6 pins power connector is just crazy.

[tacos4me]

Well, this makes me feel better. There's really no reason to go R600 over a G80 if these numbers are true. An overclocked GTX will most likely be at least equal to a stock XTX, and it doesn't look there's too much headroom for overclocking on the R600, but we'll see I guess. Not to mention the GTX uses less juice and actually fits in my case. :rolleyes:

One thing is for certain though, G80 refresh > R600.

[Treason]

Seems to me doomed from the get-go. Can AMD take this big of a financial hit?

And don't go on about how it's a bad thing for a nVidia monopoly. They've lowered the pricing on their 8800 GTS even with no competition for the last 4 months in that level bracket.

[TG01]

One thing is for certain though, G80 refresh > R600.
So far nothing is certain and if you're going to make any comment based on these benchmarks then it would be:
R600>G80


Oh wait.. this is the nvidia forum..
Excuse me..

[Xion X2]

So far nothing is certain and if you're going to make any comment based on these benchmarks then it would be:
R600>G80


Oh wait.. this is the nvidia forum..
Excuse me..


Oh please. You have absolutely nothing to back that up with, either.

The mighty 2800XT is a whopping 700 points better in 3dmark05. That's a 2-year old benchmark program. Meanwhile, you can easily overclock an 8800GTX to surpass that, easy. 650 on the core and 2100 on the memory like I had mine and you're good to go.

Until we find out how well these cards actually overclock and have some DirectX10 games to bench, both sides need to quit yapping. One thing's for sure, though, I'm less than impressed so far. Nearly 4 months after the release of G80 and this is the best they can come up with in DirectX9 performance (which will still make up the majority of titles)? And from what I'm hearing it'll be late March-April before it's on store shelves.

[Dazz]

LMAO great benchmarks tests using a Athlon 64 FX 55 a single core CPU urm can anyone say CPU limited.

From our forum post, apparently the specifications and benchmarks of R600 are leaked. The rumored ATi Radeon X2800 XTX specs are 4way× (MADD(2flops) + MUL(1flop))× 800 MHz × 64 SPs ＝ 614.4GFlops, Core Clock of 700 MHz and Memory Clock of 1100 MHz (2200 DDR). Here's the rumored benchmarks using an old Athlon 64 FX-55, WHQL driver and 1024x768 noAA/noAF :

Benchmarks X2800XTX (700/2200) 8800GTX (575/1800) X2800XT (625/1800) 8800GTS (500/1600) 7950GX2 (500/1200)
3DMark03 43606 41274 34623 32567 29257
3DMark05 17387 15467 16135 13773 12886

Also using Core 2 Extreme QX6700 @ 5300MHz (353.3×15), Radeon X2800XTX 1GB 512bit 0.85ns GDDR4 @ 700MHz/2200MHz, the 3DMark06 score stood at 15032 3DMarks (SM 2.0 = 5754 Marks/ SM 3.0 = 5841 Marks/ CPU Score = 7615 Marks).

Wow 15k Benchmarks on a stock clocked video card.

[Toss3]

LMAO great benchmarks tests using a Athlon 64 FX 55 a single core CPU urm can anyone say CPU limited.

Wow 15k Benchmarks on a stock clocked video card.
Yeah that sure is a monster! :D And AthlonXP1800, maybe you shouldn't jump the boat before the official benchmarks are released - benchmarks done on a computer without such a huge bottleneck? Besides, the r600 is the one winning, not the g80. It alls comes down to Vista support though, as nvidia has made some enemies by not having drivers fully functional for Vista yet. :headexplode:

[Xion X2]

Wow 15k Benchmarks on a stock clocked video card.

Yeah, but look at what's powering it. A Quad-core Xtreme @ 5.3gHz!

One has to wonder why they wouldn't bench '06 with all the cards? Why bench the comparison test with an old heap of graphics trash such as 3dMark05? That tells us nothing. IMO, the chart benchmarks are worthless. Why would anyone be basing performance on a 2-year old outdated benchmark program?

The score you need to look at is the 3dMark06 score, and it's not that great, IMO. 15,032 for an XTX paired with a quad-core OC'd to 5.3gHz? Sounds kind of low. I scored 12,200 w/ a single 8800GTX paired with my Conroe @ 3.6gHz, and a quad-core adds a lot of points to your score. I could've sworn it added 2,000+ points to your total, and the one in this test is overclocked all the way to 5.3gHz!

Where's slawter, HeavyH20 or Spartan? Any of you Quadmeisters out there that would be willing to bench 3dMark06 on a single 8800GTX for us as a comparison test? Heavy would be the best comparison since he's on phase @ 4.4gHz.

[tacos4me]

LMAO great benchmarks tests using a Athlon 64 FX 55 a single core CPU urm can anyone say CPU limited.

Wow 15k Benchmarks on a stock clocked video card.

The 15k comes from that processor. My 8800GTX clocked at 600MHz/2000MHz yields the following results:

3DMark Score 12163 3DMarks
SM 2.0 Score 5467 Marks
SM 3.0 Score 5589 Marks
CPU Score 2897 Marks

Notice my CPU score. The actual GPU tests aren't very far off from the R600's. I'm sure I could squeeze out another 300+ points for both SM2 and 3. So as it stands, if these benchmarks are legit, overclocked 8800GTX = R600. Right now I'm interested in the overclockability of the R600.

EDIT: Yeah, Xion beat me to it. :p

[slaWter]

Any of you Quadmeisters out there that would be willing to bench 3dMark06 on a single 8800GTX?

HeavyH20 has some single card results AFAIK. Maybe he'll post the result here.