[ PCGH ] Maxwell GTX 880 specifications leaked

[Genx87]

If these specs are legit, then this is a repeat of Kepler and I'll be annoyed. This is going to hurt their sales with the people who actually want to build at the top end. What motivation do I have to upgrade if this is all they're offering for Maxwell out the gate? I've been planning a Haswell-E/DDR4/Maxwell build for 2015 for doing 4k Project CARS, but what's the point if this is it?

The specs are fake

 

[f1sherman]

The specs are fake

considering CCores and TMUs - its 5x times GM107, which is huge (Maxwell cores are significantly more potent than Kepler's)

but ROPs are 1/2 of that and BW is close to that too, so I guess that makes no sense

but who knows...

 

[ShintaiDK]

20nm Maxwell chips are called GM200 series as well. (Taken from nVidia driver).

 

[norseamd]

If this is accurate the 880 is going to be a weird card that is faster than the 780ti in benchmarks and shader intensive games but gets killed at high resolutions and w/ AA.

do they think they can bank by forcing sli for 4k?

first thing that came to my mind when i saw the low bandwidth and rops

 

[TreVader]

Chiphell appears to be tentatively buying it.



I think they have it right, except for clocks. I think it'll be clocked closer to 1100.

 

[toyota]

Chiphell appears to be tentatively buying it.



I think they have it right, except for clocks. I think it'll be clocked closer to 1100.

no way in hell those specs are right. no point in repeating why if you have not listened by now.

 

[tviceman]

I agree with Keys. NV's designs for the last several years have been quite modular (the smallest complete unit being the SMX for Kepler) and I expect there to be pretty good scaling. Not perfect scaling, because nothing is ever 100.00% perfect and there may be bottlenecks here or there, but pretty good scaling.

This is actually a good way to indicate performance, if nothing else, than looking at Kepler. GK104 was about ~333% faster than GK107, and consumed about that much more power as well. So, with things being as they are and assuming similar scaling, it's completely not unreasonable to expect GM204 to be 333% faster than GM107. In fact, if GM204 is coming on 20nm, it's not unreasonable to expect it to be even more than 333% faster than GM107 (a 28m, chip) since 20nm will offer more perf/watt.

Anyways, 333% faster than GM107 would be about 20% faster than GTX 780 TI. If Nvidia takes advantage of 20nm's power characteristics, then GM204 should be about 35% faster than GTX 780 TI.

 

[Ajay]

Anyways, 333% faster than GM107 would be about 20% faster than GTX 780 TI. If Nvidia takes advantage of 20nm's power characteristics, then GM204 should be about 35% faster than GTX 780 TI.

That would be pretty nice - especially if it came in at around $350-400 with at least 3 GB of RAM. Not sure if Maxwell is any more efficient with RAM, but I understand it will have more cache per GPU than an equivalent Kepler GPU. Oh and the clusters are now called SMMs w/128 CCs/SMM - I'd imagine that'd mean more ROPs too, and wider buses, but maybe not (even if NV uses one ROP per SMM, the can always disable some for product segmentation/cost reasons).

Didn't find anything useful on NV's CUDA Dev site - just a compatibility guide on changes to compile for the GM107.

 

[tviceman]

That would be pretty nice - especially if it came in at around $350-400 with at least 3 GB of RAM. Not sure if Maxwell is any more efficient with RAM, but I understand it will have more cache per GPU than an equivalent Kepler GPU. Oh and the clusters are now called SMMs w/128 CCs/SMM - I'd imagine that'd mean more ROPs too, and wider buses, but maybe not (even if NV uses one ROP per SMM, the can always disable some for product segmentation/cost reasons).

Didn't find anything useful on NV's CUDA Dev site - just a compatibility guide on changes to compile for the GM107.

It's almost a foregone conclusion that it will cost more than $350-400. It will more than likely follow in the foot steps of GTX680 and be Nvidia's initial 20nm flagship chip, and will likely cost $550-600. Also, if it is a 256-bit bus then it will have 4gb of vram. Anyways, my guess is $599 and ~30% faster than gtx 780 TI. Nvidia controls their prices well for the most part, so the gtx 780 TI will likely still be priced right at $650+ right up until this card launches, so complain now but $50-75 cheaper for 30+% more performance, with 40-60 less watts in power consumption, and an extra gb of vram will make this new card look like a good deal to many.

 

[Erenhardt]

I will make this new card look like a good deal to many.

Titan-Z will be nvidias new deal maker - "Look! Its almost as fast as Titan-Z but costs only half of that!"

 

[rgallant]

It's almost a foregone conclusion that it will cost more than $350-400. It will more than likely follow in the foot steps of GTX680 and be Nvidia's initial 20nm flagship chip, and will likely cost $550-600. Also, if it is a 256-bit bus then it will have 4gb of vram. Anyways, my guess is $599 and ~30% faster than gtx 780 TI. Nvidia controls their prices well for the most part, so the gtx 780 TI will likely still be priced right at $650+ right up until this card launches, so complain now but $50-75 cheaper for 30+% more performance, with 40-60 less watts in power consumption, and an extra gb of vram will make this new card look like a good deal to many.

true but at that time the spec for big max should be known and coming down the pipe @ plus 100% faster for maybe +$100.

if amd takes on the fight for the crown , the release for big max will be asap , not well we see sales are slowing down on this batch of sku's , so lets release the next one the list.

I know with myself , I'll be telling peeps to wait for the big max when the time comes.

 

[Ajay]

It's almost a foregone conclusion that it will cost more than $350-400. It will more than likely follow in the foot steps of GTX680 and be Nvidia's initial 20nm flagship chip, and will likely cost $550-600. Also, if it is a 256-bit bus then it will have 4gb of vram. Anyways, my guess is $599 and ~30% faster than gtx 780 TI. Nvidia controls their prices well for the most part, so the gtx 780 TI will likely still be priced right at $650+ right up until this card launches, so complain now but $50-75 cheaper for 30+% more performance, with 40-60 less watts in power consumption, and an extra gb of vram will make this new card look like a good deal to many.

Actually, you are probably right, in which case, I'll be waiting for the first price drop before jumping in. Although, if I decide to get more serious about writing CUDA apps, I may be shelling out for a Titan, in which case I am screwed anyway