[KonKorT]

Geforce GTX 260: 192 SPs, 64 TMUs and 448 Bit?

Yes, it is already known, that Geforce GTX 260 gets 192 shaderunits and probably an 448 Bit interface, but it is new that the chip will have 64 TMUs and 28 ROPs. You can check this by having a look at the architecture construction. 64 TMUs and 28 ROPs are the only options, if Nvidia do not change the architecture construction and the informations are correct about the 192 SPs and 448 bit interface.

Here is a good overview:

http://www.hardware-infos.com/news.php?news=2078

regards,
Florian

[penguin1]

oh SHIT. this cards are gonna pwn shit

[bruce555]

F*ck my donkey, I want one!! Wow, let's see the retail for these.

[penguin1]

yea hand ichy liao.. time to throw away my 7900 time for it to retire for good.

[hwaChris]

What's new? Yesterday there was already a link to dailytech ...

http://www.dailytech.com/Nextgen+NVI...ticle11842.htm

[trek554]

this has been said almost since the GT200 was first mentioned so whats with the question mark in the title? 28 ROPs with a 448bit bus should be common sense based on the way that Nvidia does things. 16 ROPs for 256bit, 12 for 192bit, 32 for 512bit and so on.

[Niceone]

Well TGdaily claimed that GT200 would have 16 clusters and all of them would have 16 SP units and in GTX 280 there would be 15 enabled and in GTX 260 15.

If TGdaily is correct:
GTX 280 - 120 TMU
GTX 260 - 96 TMU

Amount ROP's are rather meaningles. 16 would be enough already

[trek554]

Originally Posted by Niceone

Well TGdaily claimed that GT200 would have 16 clusters and all of them would have 16 SP units and in GTX 280 there would be 15 enabled and in GTX 260 15.

If TGdaily is correct:
GTX 280 - 120 TMU
GTX 260 - 96 TMU

Amount ROP's are rather meaningles. 16 would be enough already

thats what I thought for TMUs also. I do disagree with you that 16 ROPs is enough though.

[Haikusan]

Any news from the 55nm versions ?

[Ske7ch]

Hot damn.

[Reaperrr]

Originally Posted by trek554

I do disagree with you that 16 ROPs is enough though.

Me too.

Actually I think that the larger number of ROPs, and NOT the wider memory interface/larger memory amount was the main reason why Geforce 8800 GTX is still faster than 9800 GTX in some situations, despite the lower clockspeeds and lower amount of TAUs.

I think the importance of the ROPs is widely underestimated, while the impact of memory bandwith is overrated.

[NextGen_Gamer]

Originally Posted by Reaperrr

Me too.

Actually I think that the larger number of ROPs, and NOT the wider memory interface/larger memory amount was the main reason why Geforce 8800 GTX is still faster than 9800 GTX in some situations, despite the lower clockspeeds and lower amount of TAUs.

I think the importance of the ROPs is widely underestimated, while the impact of memory bandwith is overrated.