NVIDIA GeForce GTX 880 Detailed

[btarunr]

NVIDIA's next-generation GeForce GTX 880 graphics card is shaping up to be a true successor to the GTX 680. According to a Tyden.cz report, GTX 880 will be based on NVIDIA's GM204 silicon, which ranks within its product stack in the same way GK104 does to the GeForce "Kepler" family. It won't be the biggest chip based on the "Maxwell" architecture, but will have what it takes to outperform even the GK110, again, in the same way GK104 outperforms GF110. The DirectX 12-ready chip will feature an SMM (streaming multiprocessor Maxwell) SIMD design that's identical to that of the GeForce GTX 750 Ti, only there are more SMMs, spread across multiple graphics processing clusters (GPCs), probably cushioned by a large slab of cache.



This is what the GTX 880 is shaping up to be.

• 20 nm GM204 silicon
• 7.9 billion transistors
• 3,200 CUDA cores
• 200 TMUs
• 32 ROPs
• 5.7 TFLOP/s single-precision floating-point throughput
• 256-bit wide GDDR5 memory interface
• 4 GB standard memory amount
• 238 GB/s memory bandwidth
• Clock speeds of 900 MHz core, 950 MHz GPU Boost, 7.40 GHz memory
• 230W board power

View at TechPowerUp Main Site

 

[OC-Rage]

wow GODDAMN what is this {monster}

ready for 4K Games ready for O.C i think 256bit not enough for high resolutions

instead 7.40 Ghz memory speed

 

[john_]

I like coincidents. For example at wccftech they have possible specs for R300 series.
AMD Pirate Islands : R9 300 Series Alleged Specifications Detailed, Flagship Cores Bermuda XTX, Treasure Island XTX and Fiji XTX

Anyway, I thing 32 ROPs is pretty low for a hi end card today with 4K in mind, or am I wrong? 256bit already mentioned.
Maybe it is happening what I am afraid of. Hi end names and prices on mid range products.
That's why I was shouting that Titan was and is bad news for us gamers/desktop users no matter who's company fan you are and 295X2 is in fact a ridiculously overpriced card justified partly thanks to Titan Z's price. And it doesn't matter if you rush to defend the semi custom nature of Titan. As gamers WE DO NOT CARE FOR THE PROFESSIONAL FEATURES OF THE CARD.

Who ever was defending Titan, as a good Nvidia fanboy, gets as a present a card with 32 ROPs instead of 48 and 256bits data bus instead of at least 384bits. Congrats.

 

[Tsukiyomi91]

if it's true, then AMD has to come up with a "contingency" plan since the Maxwell's power efficient chips (as well as potential) proved to be one of Nvidia's main strongholds here. But what really surprises me is that a high-end chip, the GM204 chip inside the GTX880 has a sweet 230W power consumption! That's like 20W lower than GTX780Ti's full-blown GK110 chip!!! So, I say kudos to Nvidia once again of proving to be the best VGA vendors money can buy.

 

[Tsukiyomi91]

the Maxwell aka GM2xx chips have bigger memory cache than Kepler chips, so having a blazing-fast VRAM speeds & huge video memory compensates the bus width. 256-bit is quite standard so it's not of a big deal anyways. Remember, Maxwell is focused on energy efficiency, not raw power like AMD's super-hot Tahiti chips.

 

[Katanai]

3200?!? This will be a beast. If it won't cost a fortune, I will have to get this one.

 

[NC37]

And all the people who bought a $3000 GPU go...

NV stated they'd be doing a tick-tock setup like Intel. Kepler was a nice advancement but it was still the tock to Fermi. It isn't super different. Now we're going on the next tick, which is Maxwell.

Interested to see if NV gets it right again like they did with the 400s. The 750 is really mediocre and its a GM chip. Will be keeping a close eye and hopefully new employment to pay for it. Unless of course, AMD finally gets everything right.

 

[Tsukiyomi91]

what AMD needs to do now is to make their chips faster, lower the power consumption on new or refreshed chips & improving the software (drivers too) in order to push the card to it's maximum potential. Heat issues are inevitable, but if they managed to yield lower TDP, then they're getting on the right track for the 1st time in years.

 

[HumanSmoke]

I like coincidents. For example at wccftech they have possible specs for R300 series.
AMD Pirate Islands : R9 300 Series Alleged Specifications Detailed, Flagship Cores Bermuda XTX, Treasure Island XTX and Fiji XTX.

Ah, WCCF...

I have just recently gotten my hands on a somewhat outdated and alleged AMD GPU specification details.....all Pirate Island GPUs will have the DirectX 12 Hardware Feature Set.

That's some mighty fast work. According to AMD, DX12 only became a reality after Mantle showed up, so AMD are going from architecture tweak to design layout to tape out to risk silicon to production silicon and die packaging in 9-10 months...on a new process which doesn't seem conducive to high power IC's (TSMC's CLN20SOC) ?...and there's no way that TSMC's 20nm BEOL/16nm FEOL will be ready for primetime in that timeframe.

Should be interesting to see how big the die gets. 4224 cores are a 50% increase over Hawaii for a less than 30% smaller node transition.

Both of these estimations (Pirate Islands and GM204) look more like wish lists, unless the timeframe is further out to allow for 20nm/16nmFF node usage.

 

[Xzibit]

*Anyone else getting a VIRUS WARNING when clicking the ExPreview link ?

Here the comparison chart and it looks like going to cost just as much as a 780 TI. Have to wonder how much the big chip will cost then

 

[bim27142]

Hoping for a 860 or 860 Ti with no external power connector...

 

[Tsukiyomi91]

AMD now wanted to use DX12 Hardware Feature Set?? that's something...

 

[HumanSmoke]

Here the comparison chart and it looks like going to cost just as much as a 780 TI. Have to wonder how much the big chip will cost then

An arm and a leg...and maybe a kidney for the fully enabled die.
The cost per wafer for 20nm BEOL+ 16nm FEOL (FinFET) is rumoured to be around $6000-6500 per- a sizeable cost increase over the existing 28nm process (< $4K) + costs for masks. If the Pirate Islands info from WCCF is ballpark then there probably wont be much difference in die size between AMD and Nvidia at the high end. A quick and dirty calculation says that AMD's chip (for a 50% core increase) would be ~460-470mm², while Nvidia's would probably decrease from its current flagship (551mm²) (GM204 not GM200 if it eventuates) since the core count is only increasing by 11% and uncore would decrease in relation to GK110 (less die space needed for memory controllers, I/O) unless Nvidia decide to beef up the SFU's to increase FP64 or dramatically increase cache.

 

[Rahmat Sofyan]

Titan Z not come yet for review, GTX 790 never come?

Now GTX 880 already rumoured..Jezzz

 

[sweet]

Tyden.cz? Never heard about it. An unpopular source from Czech leaks info about nVidia? So reliable
And the "leaked" specs is not that impressive compared to the "leaked" from AMD
http://wccftech.com/amd-pirate-islands-r9-300-series-bermuda-fiji-treasure-islands-xtx/

 

[buildzoid]

Maxwell is better at mining than current GCN so if scrypt ASICs don't show up soon these will cost 2x MSRP

 

[HumanSmoke]

Tyden.cz? Never heard about it. An unpopular source from Czech leaks info about nVidia? So reliable
And the "leaked" specs is not that impressive compared to the "leaked" from AMD
http://wccftech.com/amd-pirate-islands-r9-300-series-bermuda-fiji-treasure-islands-xtx/

WCCF are only any good when they're cutting and pasting from known leakers (Chinese forums) or another mainstream site...otherwise it's more like this (and I don't think there is a correct specification in the whole article)

 

[sweet]

WCCF are only any good when they're cutting and pasting from known leakers (Chinese forums) or another mainstream site...otherwise it's more like this (and I don't think there is a correct specification in the whole article)

Yeah, I never said WCCF is a reliable source. But is Tyden.cz THAT popular to be cited by TPU???

I expected some leaked nVidia's slides when I read the title "NVIDIA GeForce GTX 880 Detailed" (which has no question mark). What a mislead article.

 

[arbiter]

Tyden.cz? Never heard about it. An unpopular source from Czech leaks info about nVidia? So reliable
And the "leaked" specs is not that impressive compared to the "leaked" from AMD
http://wccftech.com/amd-pirate-islands-r9-300-series-bermuda-fiji-treasure-islands-xtx/

yea that one link shows 390x having 4224 Stream Processors seems bit far fetched for 1 gpu generation. Seems like be more like be about the same as nvidia's or 3500ish range for AMD's 390x. Otherwise would be a be a pretty massive gpu and being 300watts ish or more wouldn't shock me..

 

[rokazs1]

Hoping for a 860 or 860 Ti with no external power connector...

That's a bit unrealistic, even some of the 750Ti has 1 x 6pin
860Ti with 1 x 6pin and I'm sold .

 

[HumanSmoke]

Yeah, I never said WCCF is a reliable source. But is Tyden.cz THAT popular to be cited by TPU???

Probably not. As I said in an earlier post - Both of these estimations (Pirate Islands and GM204) look more like wish lists.
The silly season is upon us. People are itching to hear about something new after a relative ( two-and-a-bit years) eternity on 28nm, so the semi-pro guessers lay down the estimates and the forums go wild. That Czech site will probably get more hits in the next 24 hours than they normally get in 6 months. Ka-Ching!

 

[RejZoR]

Interesting, they'll be using less ROP's and less bus bandwidth than in older models.

 

[TheDeeGee]

3200?!? This will be a beast. If it won't cost a fortune, I will have to get this one.

According to a comparison table on Guru3D, it's aimed to be the same price as a 780 Ti.

GTX880 - 16000 CZK (583 Euros / 807 Dollars)
GTX780 Ti - 16100 CZK (586 Euros/ 881 Dollars)

 

[buggalugs]

AMD now wanted to use DX12 Hardware Feature Set?? that's something...

What are you going on about. This is Nvidia next gen hardware, wont be out for a while, we're still talking rumours. AMD will have something to compare soon enough. AMD will be built on 20nm too so performance will be +/- 10%.....just like every generation. Both are built at the same factory.

The plan is same as always, to milk the consumers. Nvidia will sell you cut down GTX 880 first with disabled ROPS,TMUs CUDA cores etc etc, then AMD will release something, as good or better, then Nvidia will unlock some features of same silicon and call it 880 Ti so consumers need to spend hundreds on a new card.

There is no "better" graphics card company, they are both in on it, everything is pre-planned, Nvidia knows exactly what AMD is doing and AMD knows exactly what Nvidia is doing.

 

[robert3892]

Looking forward to this but since tsmc is having difficulties with 20nm dies I expect to see 8XX series cards towards the end of this year or beginning of next year.