• 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.

GTX 485

Woo, so Nvidia's answer to a 6870, will be a proper Fermi, with maybe 10% more power while AMD should be aiming for a 30-40% bump in performance with their 6870.

AMD will likely hit the same price points while Nvidia are planning to release a likely significantly slower card, for yet another increase in price?

As for yields being fixed, I really wish people would not make assumptions. They will be getting 512 sp Fermi's off every wafer, just not enough to make a releasable part. If the yield was fixed and they could release it now, THEY WOULD, they can't, which means yields aren't fixed, they are just hoping to have enough numbers stockpiled by Q3(I'd guess late Q3 which will slip to Q4) to make a small release of a part no one will want by then. 480gtx sales have stalled due to price/performance and the guys who only guy Nvidia all have them already. A 485gtx will have even worse sales at a higher price for little noticeable difference in performance.

A 485gtx was always coming though just a matter of when they would have enough to make a real launch, the fact its taking so long really does confirm how terrible the yields are.

Thing is a dual gf104 will likely come out aswell offering similar/better performance for similar/lower cost.

The 480gtx can't sell well at all now, if the 485gtx launches the same time as a 6870 at £300 with 20% higher performance, the 485gtx will end up the worst Fermi yet.
 
Based on those specs raw hardware wise it would only be 6-7% faster - hardly any point for the extra power usage, etc.

To call it nVidia's answer to the 6870 tho is a little... misguided at best...
 
Oh come on DM stop being so cynical! :D

All the nay-sayers who said fermi was 'impossible' to manufacture and 'broke', well I see a sheetload of the cards everywhere and they are actually selling, not as many as the 5xxx series I bet but selling nontheless!

It will be interesting to see how much of a diff those extra SP's make though. 10% maybe?
 
I guess we could just think of it as the 'Ultra' version.. So probably not. :D

I shall be waiting in the wings though probably, for a price drop or bargain on fleabay to add it to my new rig when I build it soon.. :P
 
Oh come on DM stop being so cynical! :D

All the nay-sayers who said fermi was 'impossible' to manufacture and 'broke', well I see a sheetload of the cards everywhere and they are actually selling, not as many as the 5xxx series I bet but selling nontheless!

It will be interesting to see how much of a diff those extra SP's make though. 10% maybe?
No mate I see loads of fermis sitting on shelves not selling, never seen such availability in a latest generation launch before
 
NVidia can quite easily make Fermi 20% faster by increasing default clocks and unlocking the additional cores. A 20% increase may allow the "GTX485" to compete with 68xx.
 
Woo, so Nvidia's answer to a 6870, will be a proper Fermi, with maybe 10% more power while AMD should be aiming for a 30-40% bump in performance with their 6870.

AMD will likely hit the same price points while Nvidia are planning to release a likely significantly slower card, for yet another increase in price?

As for yields being fixed, I really wish people would not make assumptions. They will be getting 512 sp Fermi's off every wafer, just not enough to make a releasable part. If the yield was fixed and they could release it now, THEY WOULD, they can't, which means yields aren't fixed, they are just hoping to have enough numbers stockpiled by Q3(I'd guess late Q3 which will slip to Q4) to make a small release of a part no one will want by then. 480gtx sales have stalled due to price/performance and the guys who only guy Nvidia all have them already. A 485gtx will have even worse sales at a higher price for little noticeable difference in performance.

A 485gtx was always coming though just a matter of when they would have enough to make a real launch, the fact its taking so long really does confirm how terrible the yields are.

Thing is a dual gf104 will likely come out aswell offering similar/better performance for similar/lower cost.

The 480gtx can't sell well at all now, if the 485gtx launches the same time as a 6870 at £300 with 20% higher performance, the 485gtx will end up the worst Fermi yet.

I reckon it will see a fair speed boost from the 480, but i think a 6870 would still be the winner. it depends on if all nvidia are just stockpiling the 512 capable chips (which i reckon is most likely), or actually making minor tweaks to the process to crank out a bit more performance

NVidia can quite easily make Fermi 20% faster by increasing default clocks and unlocking the additional cores. A 20% increase may allow the "GTX485" to compete with 68xx.

and by doing that, you end up with a card that could quite possibly exceed a 300w TDP, meaning its outside of the ATX spec (75w from board, a 6pin pcie and an 8pin pcie), meaning OEM manufacturers will have no interest in it whatsoever. Even for high end cards, that will probably account for a sizable chunk of the sales.
 
Based on those specs raw hardware wise it would only be 6-7% faster - hardly any point for the extra power usage, etc.

To call it nVidia's answer to the 6870 tho is a little... misguided at best...


Unfortunately its not, I think we can all safely agree Nvidia can NOT add any more transistors on 40nm, I mean seriously can we agree on that? IN which case AMD have a new chip for 40nm, and Nvidia do not, both will have a new gen out on 28nm, which will compete with each other.

AMD will bring out new chips late Q3/early Q4, Nvidia are apparently releasing this 485gtx in Q3.

So the 485gtx will be Nvidia's top card, and AMD's top card will be a 6870.......... honestly I'm failing to see how these cards won't be competing with each other.

Nvidia might not have planned it to be the answer to the 6870, or want to compete with it, but thats how its playing out.


Oh come on DM stop being so cynical! :D

All the nay-sayers who said fermi was 'impossible' to manufacture and 'broke', well I see a sheetload of the cards everywhere and they are actually selling, not as many as the 5xxx series I bet but selling nontheless!

It will be interesting to see how much of a diff those extra SP's make though. 10% maybe?

Fermi was never said to be impossible to manufacture as in, "the circuits can't even be physically printed onto wafers. Its called, artistic licence in writing. He VERY clearly meant its not manufacturable in the sense that theres a certain yield, size, cost, power leakage and general targets you need to hit when producing a chip of any kind. He very clearly listed the specs and power usage of said chip, which are regarded as a completely failed design. The simple fact is AMD are making HUGE profit on each chip while selling more chips and Nvidia are making no money and probably a loss, while selling a lot less chips while losing market share.

It IS a failure, in every sense of manufacturing, the ultimate goal being simple, to make a wafer of chips you can sell at a profit, they've failed to do that.

In reality it would be surprising if it got more than 10% faster, and its incredibly unlikely to be a new spin, but just the stockpile of chips they've managed to get off each wafer being made. If so its got no chance of being a 4890 with tweaks and better clocks/better overclocking.
 
I don't think fermi one (GTX480) did itselfs any favours in performance by having less TMU's less cache per shader processor than the old GTX280 dispite the doubling in SP's and increase in bandwidth. Which seemed like the wronge path to take when u look at the GTX465 specs and compare them to the similar performing GTX280/285 DX11 or not

Probably would have been better off with only a 256mbit buss leave die space for more TMU's and cache. but what do i know I'm no chip designer lol
 
NVidia can quite easily make Fermi 20% faster by increasing default clocks and unlocking the additional cores. A 20% increase may allow the "GTX485" to compete with 68xx.

Your post is correct bar two words 'easily make', if it were that easy Nvidia would have made a 512 core version by now and one that was priced to compete against the HD5970 but it's widely believed that a 512 cores was really diffcult to manufacture so it was put on ice.

If Nvidia can get another 7-8% from this card and price the same as a regular GTX480 it would make it a lot more competitive when it's being compared to the HD5870 which will hopefully be enough to get AMD to slash prices (finally).
 
Unfortunately its not, I think we can all safely agree Nvidia can NOT add any more transistors on 40nm, I mean seriously can we agree on that? IN which case AMD have a new chip for 40nm, and Nvidia do not, both will have a new gen out on 28nm, which will compete with each other.

AMD will bring out new chips late Q3/early Q4, Nvidia are apparently releasing this 485gtx in Q3.

So the 485gtx will be Nvidia's top card, and AMD's top card will be a 6870.......... honestly I'm failing to see how these cards won't be competing with each other.

Nvidia might not have planned it to be the answer to the 6870, or want to compete with it, but thats how its playing out.

Bare in mind that (refreshes aside) the next proper step up from ATI is looking like its just evergreen shaders with some optimisation and a redesigned core - this isn't a bad thing tho if you look at graphs of benchmark results over time comparing against a breakdown of what the card is rendering at the time - shader performance is very much NOT the area the 5 series cards fall down on - infact its what pulls their numbers up as other areas let it down.

As for Fermi - I agree its going to be very hard to add any more transistors... I do believe tho that with a rejig of the design, i.e. designing it in as a 480SP process rather than 512 - they will be able to boost performance in other areas and gain much better clockability at the same power level - possibly even 20-30% higher clocks.
 
Back
Top Bottom