NVIDIA GF100 & Friends speculation

[Alexko]

That's not necessarily a voltage problem, it just means they are operating just above Vmin.

Yes that's what I meant: Vmin is too close to operating voltage to allow for significant power savings when idling.

Unless I got something wrong, that is.

 

[leoneazzurro]

I think is not a good test (better: pre-test), but 3d vantage is not much cpu-dipendent

Sure? Maybe this is true for the Extreme score but

http://www.guru3d.com/article/core-i5-750-core-i7-860-870-processor-review-test/17

it seems that Performance test is quite affected by the CPU. Look at the i7 965 vs i7 920 : a 560 MHz nominal frequency increase leads to a 1200 pts in the benchmark. And the VGA is a GTX280 OC, it is better than the 5xxx mobility cards but not by 50% or so...

 

[CarstenS]

Even worse - the CPU part of the score is heavily influenced by Physx in the normal running mode. Or was it disabled in this case?

 

[leoneazzurro]

I think it´s disabled as there is also the CPU score listed in the first graph. But we don´t know if it´s disabled in the "leaked " benchmarks. Anyway, we can easily browse and compare the benchmarks for the desktop 5750 (1 Gbyte), as the mobility 5870 has the same core clock and almost the same memory clock. As far as I can see, probably the 5870 mobility could reach around 9000 pts with that processor and if the 480M score is with Physx, then things are not so good for Nvidia. But, anyway, the 480M is a downclocked GTX465 card, with clocks that are 75% of the desktop counterpart. So, it will probably perform around desktop 5770 level (a little less IMHO). If it´s so, and ATI needs to counter this card, they probably won´t need even a mobile Cypress at 100W. An 850 MHz - 75W Juniper would be enough.

 

[Kaotik]

I think it´s disabled as there is also the CPU score listed in the first graph. But we don´t know if it´s disabled in the "leaked " benchmarks. Anyway, we can easily browse and compare the benchmarks for the desktop 5750 (1 Gbyte), as the mobility 5870 has the same core clock and almost the same memory clock. As far as I can see, probably the 5870 mobility could reach around 9000 pts with that processor and if the 480M score is with Physx, then things are not so good for Nvidia. But, anyway, the 480M is a downclocked GTX465 card, with clocks that are 75% of the desktop counterpart. So, it will probably perform around desktop 5770 level (a little less IMHO). If it´s so, and ATI needs to counter this card, they probably won´t need even a mobile Cypress at 100W. An 850 MHz - 75W Juniper would be enough.

IIRC the GTX480M clocks are 70% for core & shaders, mem was higher though, 80% or something?

 

[leoneazzurro]

IIRC the GTX480M clocks are 70% for core & shaders, mem was higher though, 80% or something?

You are right regarding core and shaders, I thought they were 450/900 but they seems to be 425/850. About the memory, the advantage is the 256 bit bus, but memory is also clocked very low (power consumption again?) at 2400 MHz effective speed. So bandwidth is 76,8 Gbytes/s, or 75% of the desktop counterpart.

 

[Scrat]

yes....and?

...and it's a comparison between laptops*, not between video cards, as the systems are completely different and uncomparable. So nothing useful comes from that pre-test regarding relative performance of nVIDIA's and AMD's mobile GPUs.

* provided that desktop hardware inside a notebook form factor can be defined "laptop"

 

[fellix]

GTX485 Launch: Less hot and noisy in the sack?

 

[Silent_Buddha]

Possible new record for fastest death of a product? GTX 480 killed off by successor GTX 485?

Regards,
SB

 

[RobertR1]

Possible new record for fastest death of a product? GTX 480 killed off by successor GTX 485?

Regards,
SB

A repeat of x1800 > x1900 series?

 

[BRiT]

GTX485 Launch: Less hot and noisy in the sack?

I sure hope they have more than just this to offer:

KitGuru has heard that nVidia’s been through the first iteration of the GTX480 with a fine tooth comb, to create a GTX485 product that can run at least 20MHz more on the core clock, with improved cooling and performance. It’s also worked through some neat driver improvements that will also help. Nice.
<<...snip...>>
we’re now expecting the GTX485 to come in around $20 higher into distribution than the launch price of the GTX480 It will also overclock easier, because more of the pesky electrons will be going where they are supposed to go. Nice.

 

[Cookie Monster]

I sure hope they have more than just this to offer:

Im not sure how reliable kitGuru is (especially when they use a pic of the EVGA's high flow bracket in the article), but could this GTX485 be based on the new B1 revision of GF100? There could be a possiblity for the GTX485 to have all 512SPs enabled.

 

[Ailuros]

Im not sure how reliable kitGuru is (especially when they use a pic of the EVGA's high flow bracket in the article), but could this GTX485 be based on the new B1 revision of GF100? There could be a possiblity for the GTX485 to have all 512SPs enabled.

Theoretically 1475 vs. 1344 GFLOPs or around 10%. The author states "at least" 20MHz so even 40-50MHz could be within expectations, but the result is still nothing to write home about. If they wouldn't have to battle to reduce the power consumption of GF100 things would be different. Under those conditions having all SMs enabled, increasing frequencies and lowering power consumption at the same time for the entire package isn't exactly easy. For GT200 the GTX280 went from 600MHz/236W to GTX285, 648/204W and there was also a die shrink involved for the latter.

 

[Silent_Buddha]

A repeat of x1800 -> x1900 series?

Oct 2005 - Jan 2006, so a 3 month window. Will be interesting to see. And then do we go by Press launch of Mar 2010 or Product availability of Apr. 2010 for GTX 480? Either way, Nvidia will have to work hard if they want to beat ATI's record for fastest product replacement.

Regards,
SB

 

[Squilliam]

Oct 2005 - Jan 2006, so a 3 month window. Will be interesting to see. And then do we go by Press launch of Mar 2010 or Product availability of Apr. 2010 for GTX 480? Either way, Nvidia will have to work hard if they want to beat ATI's record for fastest product replacement.

Regards,
SB

Does that mean that everyone who predicted a very quick refresh ala GTX 280/285 was infact correct?

 

[Ailuros]

Does that mean that everyone who predicted a very quick refresh ala GTX 280/285 was infact correct?

I'd say give or take early fall.

 

[xDxD]

GTX485 Launch: Less hot and noisy in the sack?

20mhz more on the core clock? It's a joke?

 

[Lonbjerg]

Silly season just restarted?

 

[Alexko]

Does it ever really end?

 

[Erinyes]

Does it ever really end?

On that note, ive got a bit of info from a source. He says that a respin of fermi is on the way, which fixes the power problems. According to him, it will release in fall. And he says that there's no rumoured GF119 chip, only GF 104, 106 and 108