[wccftech] nVidia Previewed Unreleased Graphics Card

[railven]

im guessing dual gpu card. Whether 2x full gm204 or full gm200 is the question.

Why not both?

 

[96Firebird]

Why not both?

4 GPUs on one card. 2x GM200 for graphics processing, 2x GM204 for PhysX.

 

[railven]

4 GPUs on one card. 2x GM200 for graphics processing, 2x GM204 for PhysX.

Haha, that would at least justify a $2000 price tag.

 

[iiiankiii]

I predict dual titan x @ $2499.99.

 

[xorbe]

I predict dual titan x @ $2499.99.

I put on my robe and wizard hat, predict dual 980Ti @ $1500.

 

[LTC8K6]

I predict a memory surprise on the card.

 

[RussianSensation]

Mabe they will require the Niagara falls to cool a dual Fury X 500 watt+,$2,000 card?

It's actually worse for top-of-the-line performance to crippled dual-GM200 chips with an air cooler. An AIO CLC would have allowed NV even more TDP headroom for even higher GPU boost clocks, and as a consequence would have provided even lower noise levels because of the epic cooling power of AIO CLC.

Considering that a reference blower 980Ti's cooler is already inadequate for overclocking without noise levels exceeding 50 dBA, it's going to require insane binning processor to produce a cool and quiet dual-GM200 graphics card that has overclocking headroom and as much GPU boost as after-market 980Ti cards like Zotac AMP! Extreme have out of the box.

R9 295X2 was cooler at load with 2xR9 290X chips than a reference 980Ti is.

Once 980Ti is overclocked, we are talking reference blower 290X noise levels (yuck!)

This bodes very poorly for an air-cooled >$1000+ dual-GM200 card with an air cooler which in pictures appears very similar to the Titan Z's. Trying to cool down 2x250W ASICs with air cooling while maintaining the same 82-84C hard thermal throttle limit almost guarantees that a dual-GM200 card with an air cooler will never beat stand-alone GTX980Ti OC SLI, unless NV shoves dual Titan Xs on one PCB and they can boost to 1.45Ghz.

600 is a ton of cash for your average gamer though. Probably why they don't call FuryX or 980Ti "mainstream" GPUs.

He is discussing $600 R9 295X2s in the era when 980 cost $550. This phenomenon continued for months despite R9 295X2 wiping the floor with the 980 overall, while providing 92% of 980 SLI performance at 1440P and 95% at 4K for nearly half the price. Let's not forget XDMA's frame times.

In other words, his point is spot on - even when AMD completely dominated the market with its dual-chip solutions against nV's offerings, it mattered for squad. Let's not even forget the ownage HD6990 laid on VRM-crippled/dying GTX590s, HD7990 smashing GTX690 (2GB VRAM crippled) and of course R9 295X2 owning Titan Z for half the price.

im guessing dual gpu card. Whether 2x full gm204 or full gm200 is the question.

I don't think dual GM204 card make sense. Firstly, NV is milking GTX980 for $450+, while dual GTX970s are worse than GTX980Ti as an overall product. Since GTX980Ti can be purchased for $630 or sometimes below that, NV would be losing $ on dual 980s and dual 970s would have to be priced well below $630 to make any sense. But since SLI doesn't work in all games, that's not a favourable solution at all against 980Ti OC. I think it makes far more sense to package the large die Maxwell as the last hurray to 28nm.

I hope NV has improved the stock Titan Z cooler or did better power management/GPU binning because the Titan Z was a thermal throttling mess (GPU clocks fell to 706mhz over prolonged gaming) without manual fan control.

Hopefully NV has learned from the Titan Z's pricing fiasco and prices a dual-chip card at a far more reasonable price level.

TechSpot already showed that it takes after-market 980Ti SLI to actually beat Fury X CF:

"The average frame rate data saw the Fury X cards come out 4% ahead of the GTX 980 Tis based on the 10 games that we tested at 4K. Where the Fury X Crossfire setup won big was in Thief where it was 50% faster and Total War: Attila where it was 36% faster. Now for the interesting part, typically we expect Nvidia to have the edge when looking at frame time (99th percentile) performance, but this wasn't the case here. The R9 Fury X Crossfire cards were on average 22% faster when comparing the 99th percentile data."

I still think for consumers it would have been far better if NV released a 960Ti or a 980Ti Black Edition with fully unlocked shaders and 1.2Ghz base clocks. A dual-chip GK110 card (or Fury X2) seems like a very limited product given how we are on the verge of 16nm FinFET GPUs in 2016 with 8GB-16GB of HBM2 and a brand new architecture from NV.

P.S. TPU is hilarious - "It remains to be seen if NVIDIA gets the pricing wrong the second time."

 

[Spanners]

Powercolor made a custom dual 290X and I guess they're repeating it with a dual 390X.

http://www.hardwareheaven.com/2014/06/powercolor-devil13-r9-290x2-graphics-card-review/

https://www.techpowerup.com/215755/...r9-390-x2-devil13-dual-gpu-graphics-card.html

And to vendors making odd contraptions, XFX is making a watercooled (which still makes no freaking sense) Fury.

http://wccftech.com/xfx-releasing-liquid-cooled-amd-r9-fury-nonx/


So, it stands, they can basically make things if they think there is a market for it.

Those Powercolor cards aren't dual mid-range chips though. I wasn't saying dual GPU is entirely pointless. I just don't think something like dual Tahiti or GM206 has a market or a reason to exist.

Vendors can make anything they think will sell, (that's so evident I'm not sure it needs pointing out) I'm in no way saying they shouldn't or can't I just don't think dual mid-range cards will or have sold well.

 

[Zodiark1593]

Those Powercolor cards aren't dual mid-range chips though. I wasn't saying dual GPU is entirely pointless. I just don't think something like dual Tahiti or GM206 has a market or a reason to exist.

Vendors can make anything they think will sell, (that's so evident I'm not sure it needs pointing out) I'm in no way saying they shouldn't or can't I just don't think dual mid-range cards will or have sold well.

If Nvidia isn't putting a cut-down GM204 between the GTX 960 and 970, a dual GM206 gpu for ~$250 could potentially be a big $/performance hit.

Though, odds of such a dual GPU board would be, at best, 2%.

 

[3DVagabond]

Powercolor made a custom dual 290X and I guess they're repeating it with a dual 390X.

http://www.hardwareheaven.com/2014/06/powercolor-devil13-r9-290x2-graphics-card-review/

https://www.techpowerup.com/215755/...r9-390-x2-devil13-dual-gpu-graphics-card.html

And to vendors making odd contraptions, XFX is making a watercooled (which still makes no freaking sense) Fury.

http://wccftech.com/xfx-releasing-liquid-cooled-amd-r9-fury-nonx/


So, it stands, they can basically make things if they think there is a market for it.

I think he said the AIB's sometimes make them but not the IHV's. The Powercolor actually proves his point, not disputes it.

 

[3DVagabond]

It's actually worse for top-of-the-line performance to crippled dual-GM200 chips with an air cooler. An AIO CLC would have allowed NV even more TDP headroom for even higher GPU boost clocks, and as a consequence would have provided even lower noise levels because of the epic cooling power of AIO CLC.

Considering that a reference blower 980Ti's cooler is already inadequate for overclocking without noise levels exceeding 50 dBA, it's going to require insane binning processor to produce a cool and quiet dual-GM200 graphics card that has overclocking headroom and as much GPU boost as after-market 980Ti cards like Zotac AMP! Extreme have out of the box.

It would be terrible though if it required water cooling, wouldn't it? /sarc

 

[xthetenth]

It would be terrible though if it required water cooling, wouldn't it? /sarc

Yeah, I don't know how the market could stand another card setting the record for open air cooler performance :awe:.

 

[railven]

If Nvidia isn't putting a cut-down GM204 between the GTX 960 and 970, a dual GM206 gpu for ~$250 could potentially be a big $/performance hit.

Though, odds of such a dual GPU board would be, at best, 2%.

I think he said the AIB's sometimes make them but not the IHV's. The Powercolor actually proves his point, not disputes it.

Yes, it seems I mis-read his post. My apologies on that one.

 

[RussianSensation]

If Nvidia isn't putting a cut-down GM204 between the GTX 960 and 970, a dual GM206 gpu for ~$250 could potentially be a big $/performance hit.

Though, odds of such a dual GPU board would be, at best, 2%.

960 2GB is VRAM gimped and 2 of those can't really beat a single after-market 290/970. NV is making higher profits by selling 960 2GB and 960 4GB individually than if the went with dual GM206 cards for $250.

Dual-GM200 card will likely allow NV to retain the performance crown against Fiji X2 and even though it's a niche market segment, they could expand it by including 2 fully unlocked Titan Xs with 24GB of VRAM. If they do that, the card will find its audience even outside of gamers.