[SeniorEditor]

Source;
http://forum.donanimhaber.com/m_23295300/tm.htm

---

Sponsored Links:

[Niceone]

So according to that it wouldn't have DX10.1 support.

and that 50% more gaming performance.. compared to 9800 GX2 ? That would be dissapointing. On other hand Nvidia already has huge lead in game benchmarks (minimum and average fps numbers specially):
http://plaza.fi/muropaketti/artikkel...rce-9800-gtx,2

50% increase over 9800 GX2 would make that card Fast

[G|an]

Originally Posted by Niceone

So according to that it wouldn't have DX10.1 support.

and that 50% more gaming performance.. compared to 9800 GX2 ? That would be dissapointing.

are u smoking pot or what

50percent increase on 9800gx2 if that ever happens means
smaking a quad 9800gx2 performance.. and thats bliss..

basically u can say crysis has been conquered.

[Mobious]

512-bit memory, 1Gb of DDR3, almost double the stream proc count as the 9800GTX, physics support on board, and CUDA; step in the right direction if anything. One thing I'm wondering about: is it called "PhsyX", or "PhysX", cuz last I remembered "physics" is "phys".

Originally Posted by G|an

basically u can say crysis has been conquered.

Thats a big IF there. There's been no benchmarks or any test cards made yet, to my knowledge (and with the release date only a month away, I'm surprised we have yet to see any). With any luck it will though.

-Mobious-

[Niceone]

Originally Posted by G|an

are u smoking pot or what

50percent increase on 9800gx2 if that ever happens means
smaking a quad 9800gx2 performance.. and thats bliss..

basically u can say crysis has been conquered.

Well HD4870 is around 30% faster than 9800 GX2 already, not to mention HD4870 X2..

[G|an]

Originally Posted by Niceone

Well HD4870 is around 30% faster than 9800 GX2 already, not to mention HD4870 X2..

physic in the house

weird all i saw was 1.25x of 9800gtx...

wow..

[Niceone]

Originally Posted by G|an

physic in the house

Well they have to use one cluster to do physics calculations (bye bye 24 SP units) --> more object to render --> more need for calculating power --> huge performance drop.

[Earl088]

50% in paper so thats about 20~25% in real world performance, considering best case scenario compared to a GX2, sounds about right to me.

[Araris]

Originally Posted by Niceone

Well HD4870 is around 30% faster than 9800 GX2 already, not to mention HD4870 X2..

That is incorrect, the 4870 is 1.25x better than a 9800GTX, which is about 50%-60% worse than a 9800Gx2. The HD4870x2 should be about as good as a 9800Gx2 or just a bit better if ATi can link the two cards in a way that is different than the 3870x2 (since the way the two cores were linked has been reported to bottleneck the cards true potential).

So if the specifications for the GeForce GTX 280 are correct then it should be about TWICE as fast as a HD4870 and about 50% faster than a HD4870x2, not to mention you can then SLI them.

All I know is that I'm SLI'ing the GTX 280 now with this information. Crysis shall fall, since a GTX 280 should get 37.5fps in 1920x1200 with 2xAA on Crysis (hypothetical since the 9800Gx2 gets 25 with 2xAA), and then you SLI and BAM! 55-60fps. That's also if the 50% performance over a 9800Gx2 actually translates into 50% performance increase in games like Crysis.

EDIT: I would like to remind you that the 4870 was only meant to be 1.25x better than a 9800GTX since it is retailing for $200-$220 in the USA, which is $80 cheaper than the 9800GTX. ATi isn't going for the performance crown, they are going for the price/performance crown, since the GTX 280 is obviously going to be $550-$650 USD at launch.

[Anarion]

Originally Posted by Niceone

Well they have to use one cluster to do physics calculations (bye bye 24 SP units) --> more object to render --> more need for calculating power --> huge performance drop.

You fail to take account the CPU bottleneck. If these specs are true it can definitely be similar performance jump then what we saw with GeForce 6 series cards.

[TecnoWorld]

I'll buy one of these (moving away from my 8800 gts 512) if and only if they'll allow playing crysis at very high detail at 1680x1050 with AA 2x or better. All the other titles run just great on my configuration!

[Ish718]

Originally Posted by Araris

That is incorrect, the 4870 is 1.25x better than a 9800GTX, which is about 50%-60% worse than a 9800Gx2. The HD4870x2 should be about as good as a 9800Gx2 or just a bit better if ATi can link the two cards in a way that is different than the 3870x2 (since the way the two cores were linked has been reported to bottleneck the cards true potential).

So if the specifications for the GeForce GTX 280 are correct then it should be about TWICE as fast as a HD4870 and about 50% faster than a HD4870x2, not to mention you can then SLI them.

All I know is that I'm SLI'ing the GTX 280 now with this information. Crysis shall fall, since a GTX 280 should get 37.5fps in 1920x1200 with 2xAA on Crysis (hypothetical since the 9800Gx2 gets 25 with 2xAA), and then you SLI and BAM! 55-60fps. That's also if the 50% performance over a 9800Gx2 actually translates into 50% performance increase in games like Crysis.

EDIT: I would like to remind you that the 4870 was only meant to be 1.25x better than a 9800GTX since it is retailing for $200-$220 in the USA, which is $80 cheaper than the 9800GTX. ATi isn't going for the performance crown, they are going for the price/performance crown, since the GTX 280 is obviously going to be $550-$650 USD at launch.

If a single HD4870 is better than a 9800GTX, how is a HD4870 x2 gonna be equal to a 9800GX2 especially when crossfire scales better than SLI.

No way a GTX280 is TWICE as fast a HD4870, wtf are you talking about? It will be faster of course but don't exaggerate things.

HD4870 has super fast GDDR5 so it doesn't need 512bit memory bus like the GTX280 since the GTX 280 still uses GDDR3 it needs 512bit memory bus to avoid bottleneck.

[Niceone]

Originally Posted by Ish718

No way a GTX280 is TWICE as fast a HD4870, wtf are you talking about? It will be faster of course but don't exaggerate things.

Depends how much faster HD4870 is than HD3870 X2..I mean 9800 GX2 is in some gamestwice as fast as HD3870 X2
http://plaza.fi/muropaketti/artikkel...rce-9800-gx2,3

In Bioshock it's more tight, but one has to remember that they are testing without AA since it doesn't seem to work with AMD+Unreal3 based games. Nvidia's cards don't take huge performance impact on these games when using AA.

----

Also we don't know about what kind of architectural updates there have been; some changes for those few specs that can't be seen on those popular spec lists [For example Nvidia's stream processors are more capable so you cannot tell their shader performance from just SP amount and shader clocks.

[ptr1aa]

This official information leaves a lot to interpretation like "50 more performane than 1st generation" what this means 50% more than GTX ? or
GX2 ? or ???? what is first generation.
This 50% more is so left to speculations and interpretations that it can be erased from your minds and better read some numbers like how many transitors if it will have for those 240 shaders ? if they are twice the number of 9800GTX than it means it will be twice as fast as 9800GTX and a bit faster than 9800GX2 but not that wild 50% more.

[Niceone]

ptr1aa:
I'd say they are both using 1st gen unified shaders. 9800 GX2 and 9800 GTX uses G92 which is dieshrink from G80.
---
Just look at 8800 Ultra's performance on 8xAA or with super high resolutions. You'll notice that it will get really close to 9800 GX2. In some freakish test situations 8800 ultra will butcher 9800 GX2 (2560x1600 8xAA 16xAF (neither card can run games on those settings, but 8800 GTX will post three times higher fps numbers))

Now why is this? Because 8800 GTX/Ultra has 384-bit memory interface + 768MB usable vram when 9800 GX2 has only 512MB usable vram + 256-bit memory interface (and because there is not big enough difference in memory clocks).
------
GTX 260 compared to 8800 GTX Ultra:
-50% more SP units
-3 times more TMU units (3 times more Texture adressing units, 1.5 times more texture filtering units)
-1/3 times more ROP-units (16 ROPS would be already enough for 2560x1600)

GTX 280:
-87,5% more SP units
-3,75 times more TMU units
-1/3 times more ROP units
----

9800 GX2 will have more processing units than these two cards, but there won't be SLI issues (around 65-70% ish efficiency) and there won't be memory bottleneck that already hinders 8800 GTS 512's performance.

Now we don't yet know clock frequencies on GTX 260 and GTX 280. But if GTX 260 will have same memory clocks as 9800 GX2..then GTX 260 will have 75% more wide memory bandwidht..