• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Turns out we've had GT200 all along - the 9800GX2

I know the 3870 X2 has 2x seperate dies, but surely if even they stuck the 3870 X2 onto 1x die, then it would run the same as it would if it was on 2x dies wouldn't it, just like the 2x 3870's on 2x cards linked by a bridge in Crossfire, run the same as a 3870 X2, as thats those 2x cards moved onto one PCB with the bridge moved onto there as well to become 1x card, it hasn't miracullously started beating itself in 2x card form by a huge amount of frames has it. :D
 
Last edited:
Im not sure loadsa, if the work is all done on die then it might be ****** realy fast, it's not SLI(which is poo regardless of card's) it's one chip. the best way to mesure the performane would be to double a 9800gts/8800gt's score then knock off 10%.
 
Putting the chips together would improve latency. Improved performance will depend on how the cores address the memory bus. Two cores addressing a single 512bit bus would be nice.
 
It would be better than itself in 2x core form though, as it wouldn't have to switch a core off for non SLi supported games, it would run at its full potential all the time being only the 1x core.

With 192 streams though, it just sounds like they have stuck 2x GX2's on the 1x chip, cut some streams off, as the GX2 has got 256 streams, and doubled the memory interface up to 512bit.
 
Last edited:
Aekeron said:
Having an internal bridge for the two cores will be far more efficient than using pcie or whatever the dodgy external bridge they normally use is called. The new ATI cards are rumoured to have a similar setup and one that allows the card to pool the graphics ram properly - no loner will 1 gig sli/xfire cards only actually really have 512MB ram.

That's the theory/rumour anyhow. Whether or not that's true, who knows.
Click to expand...

The 3870 X2 has that now though hasn't it, the 2x cores on that are linked by Crossfire internally between them on the board.
 
i think you guys may have jumped the gun here regarding crysis performance, the claim being given is this:

55nm TSMC process
Single chip with "dual G92b like" cores
330-350mm2 die size
900M+ transistors
512-bit memory interface
32 ROP
192SP (24X8)
6+8 Pin
550~600W PSU min
9900GTX SLI runs Crysis 2560x1600 VH 4XAA smoothly


now look carefully at the last line:
9900GTX SLI runs Crysis 2560x1600 VH 4XAA smoothly

the key word being SLI, which means 2 of the above dual core cards, which means 4 cores in total if that wording is clear. more believable if you look at it like that.
 
if you have 2 cores on one die along with an architectural change allowing them both to access the same memory/frame buffer, etc. then it would be a lot more effective overall than a plain old SLI implementation - as cbyer-mav said a major part of the rendering pipeline lends itself well to multi-threaded execution. I didn't think we'd see such a change inside the next 1-2 generations of GPU technology.
 
Cyber-Mav said:
now look carefully at the last line:
9900GTX SLI runs Crysis 2560x1600 VH 4XAA smoothly

the key word being SLI, which means 2 of the above dual core cards, which means 4 cores in total if that wording is clear. more believable if you look at it like that.
Click to expand...


huh and what? :confused:


The last line in that article says...

As a whole, it appears as the two-processor geForce 9800 GX2 in single-processor accelerator. Associate report that the performance Of geForce 9900 GTX will be sufficient for comfortable game play in Crysis with very high setting with 4- AA and resolution 2560 x 1600. But it is difficulty to believe such data .
Click to expand...



I canny see SLI any where in that article never mind sentence.
 
Rroff said:
if you have 2 cores on one die along with an architectural change allowing them both to access the same memory/frame buffer, etc. then it would be a lot more effective overall than a plain old SLI implementation - as cbyer-mav said a major part of the rendering pipeline lends itself well to multi-threaded execution. I didn't think we'd see such a change inside the next 1-2 generations of GPU technology.
Click to expand...


Wait just a minute, i think i may need that test after all:p
 
They must be seeing how far they can milk their old 2006 tech, bit like what they done with Crossfire/SLi, but instead of sticking the cards together to make one, they sticking the cores together instead. :D
 
Last edited:
LoadsaMoney said:
They must be seeing how far they can milk their old 2006 tech, bit like what they done with Crossfire/SLi, but instead of sticking the cards together to make one, they sticking the cores together instead. :D
Click to expand...

:D

At least it's some sort of progress :p.
 
You must log in or register to reply here.
Back
Top Bottom