NVIDIA Maxwell Speculation Thread

Dumb question: if I am to reduce performance to roughly 1/4 from the original resolution, then why not just use garden variety 4x Supersampling instead?
Simplest answer: DSR - like ordinary downsampling - works even on games that refuse all types of AA. You can opt for other AA methods, if you prefer.

BTW - was there every any doubt about four FP64-Units since the GM107 release?
 
I asked for a product to get me to upgrade (sub $400 + 50% perf over 670) and that's what I got. This is the exact product NVIDIA needed at this time.

It wouldn't surprise me if they had a "970Ti" to slot in between the 970 and 980 at a $450 price point in the near future as well.
 
Simplest answer: DSR - like ordinary downsampling - works even on games that refuse all types of AA. You can opt for other AA methods, if you prefer.

BTW - was there every any doubt about four FP64-Units since the GM107 release?

But why blur the image?

Does the 0% smoothness setting take out the blur, making the image more like ordinary downsampling?
 
I suspect that anything referred to as FL 12 is actually FL 11_3, seeing as how FL 11_3 hasn't officially been defined and named yet.

Is that confirmed though? I'll admit 11_3 seems to have all the features that 12_0 was suppose to have, but intel does list 12_0 on their skylake slide. So I'm wondering if there will be a few more features in 12_0 over 11_3.
 
I guess though power draw isn't actually lower than that of a gtx 770 after all despite the lower TDP. Still, it's quite an improvement on perf/w. tomshardware has some very interesting power draw numbers, apparently with the right load these cards can be stressed to a level which is easily beyond the rated pcie power connectors even (50% above TDP, right at the level of r9 290). I guess that puts some things into perspective.
 
Simplest answer: DSR - like ordinary downsampling - works even on games that refuse all types of AA. You can opt for other AA methods, if you prefer.

If that really works universally, including OpenGL 1.x, D3D 9, D3D8 and lower etc. then it's very decent to have around. Though the example given of 1.78X ordered grid supersampling is not very appealing!

If it's GM2xx specific, that's a shame for low end guys / farts like me no really keeping up with games. I would want to play some quake 3, counterstrike etc. downscaled from triple the res on both axises or more. Maybe the GM206 will trickle down about a year from now, giving a card similar to what the 650ti was.
 
The interesting this is, those that will go on and on about the power efficiency but then overclock their card or cards and end up in the same neighborhood as the 290/290x.
I understand that it's not easy to be an uncompromising AMD fan today, but when you see a GPU released like this that beats its own predecessor and the competition in every metric possible, and perf/W in particular, isn't it possible to just stand aside an marvel at such a technical achievement? This is a mostly technical forum, if various metrics don't mean anything to you, then what are you doing here? AMD will come back, don't worry.
 
If it's GM2xx specific, that's a shame for low end guys / farts like me no really keeping up with games. I would want to play some quake 3, counterstrike etc. downscaled from triple the res on both axises or more. Maybe the GM206 will trickle down about a year from now, giving a card similar to what the 650ti was.

NVIDIA mentioned that this feature may make it into older cards via a driver update.
 
So I'm wondering if there will be a few more features in 12_0 over 11_3.
I haven't seen any such indications, but I don't follow it closely. Looks like DX12's existence is purely to remove driver overhead, nothing more. I wonder how to will evolve in the future: will there be 2 APIs? One for ease of use and one for performance that have the same features?
 
It will still be possible to do things DX11 style in DX12. I assume indie/low budget devs will go that route.
 
NVIDIA mentioned that this feature may make it into older cards via a driver update.

Ah, that makes sense thanks. Dunno why I was thinking that was linked to some scaling hardware or something in the display controllers.
It reminds me of old supersampling as a new feature on Geforce 2 GTS (as well as register combiner and maybe some other thing) which was from a driver update and worked on Geforce 1.
 
I guess though power draw isn't actually lower than that of a gtx 770 after all despite the lower TDP. Still, it's quite an improvement on perf/w. tomshardware has some very interesting power draw numbers, apparently with the right load these cards can be stressed to a level which is easily beyond the rated pcie power connectors even (50% above TDP, right at the level of r9 290). I guess that puts some things into perspective.

Yes that's very interesting. It looks like most of Maxwells superior performance/watt actually comes from smarter and faster clock/voltage switches? Which is great when gaming but makes little difference when doing GPGPU-stuff since then everything is pretty much maxed out anyhow.

Dave B, start your copy machines :smile:
 
Has anyone seen 3 DisplayPorts on a 970? Seems like I'm only seeing that config on the 980s...
I wonder just how expensive that connector-plate is, because not a single 970 card that I see for sale is using it. Bummer, but I guess the gift that I got for my birthday was $300. :shrug:

Is that confirmed though? I'll admit 11_3 seems to have all the features that 12_0 was suppose to have, but intel does list 12_0 on their skylake slide. So I'm wondering if there will be a few more features in 12_0 over 11_3.

http://www.anandtech.com/show/8544
Direct3D 12 will be the low level API, and Direct3D 11 will continue to be developed to offer the same features through a high level API.
 
Last edited by a moderator:
Direct3D 12 will be the low level API, and Direct3D 11 will continue to be developed to offer the same features through a high level API.

That's orthogonal to the question I'm asking. Remember, feature level != Direct3D level. It's quite possible that Direct3D 12 could support feature levels 12_0, 11_3, 11_1, and 11_0.
 
I doubt D3D12 (feature level) will support anything older than Kepler & GCN...

(Just based on what I assume the API to be, I have no access to it or its documentation whatsoever, unfortunately :(.)
 
I guess though power draw isn't actually lower than that of a gtx 770 after all despite the lower TDP. Still, it's quite an improvement on perf/w. tomshardware has some very interesting power draw numbers, apparently with the right load these cards can be stressed to a level which is easily beyond the rated pcie power connectors even (50% above TDP, right at the level of r9 290). I guess that puts some things into perspective.

Any device in this class can, because it's not a challenge for these chips to physically draw way more power than what they are rated for. It's not particularly challenging for them to draw enough power to damage themselves. The real challenge is finding ways to not let chips kill themselves.
I don't dispute the likelihood that Maxwell has improved its voltage and clock management at least somewhat, but if its DVFS is improved, I have missed whereTomshardware's numbers measure clock and voltage levels.
Using a scope to give a jaggy power graph doesn't contribute anything but noise, because of course complex power supplies for complex chips running complex workloads don't draw a flat line.
If they want to claim Maxwell is doing the same thing as Hawaii, they should use their new setup on both and log their clock/voltage steps before stating it.

Power spikes happen at extremely short intervals with or without enhanced DVFS, and if there are particularly high spikes it can be a sign that the solution is actually too slow to catch abrupt ramps in demand.
If the higher than TDP power numbers for their compute load are sustained, then that's definitely not a sign of a significantly improved DVFS.

Spikes above TDP are actually fine within the constraint that they not take more than some thermally significant period where the thermal solution is expected to absorb transients. Beyond that, more advanced turbo methods can exceed it for longer if they've modeled the cooler's thermal capacity and realize there's slack.
If it can be forced into a sustained draw above TDP, such as was given for the reference design tested at Tomshardware, that's what happens when the chip isn't doing anything particularly leading-edge in terms of DVFS.

If there were an area where I do think AMD does better, it would be in this realm, since it likely buys them several speed bins that would be lost due to guard-banding and a more conservative TDP.
However, like Mantle, such measures only let you get as far as the architectural laurels AMD has rested on for several years can take you.
 
I doubt D3D12 (feature level) will support anything older than Kepler & GCN...

(Just based on what I assume the API to be, I have no access to it or its documentation whatsoever, unfortunately :(.)

I think it's quite clear at the moment that D3D Feature Level 12_0 will require Maxwell as minimum, GCN (at least some of them) might do, as well as possibly Intel's latest, but Kepler & Fermi at least won't - heck, even Maxwell, GCN and Intels latest might not cut it.
It has to require at minimum what the latest D3D Feature Level 11_x requires, now we just need to find out whether GCN (some form of it) supports the supposed 11_3 stuff or not, and same for Intel.

Supporting DX12/D3D12 API is a whole another thing, obviously

That's orthogonal to the question I'm asking. Remember, feature level != Direct3D level. It's quite possible that Direct3D 12 could support feature levels 12_0, 11_3, 11_1, and 11_0.

I think his point was that devs can still use all the latest features with D3D11 API without adopting D3D12 API, there's no reason why D3D12 API wouldn't support lower Feature Levels, too.
 
I think people are getting very confused (perhaps so am I :p). Direct3D 12 will almost certainly allow you to target feature level 11_3, 11_1, and 11_0 class cards (that doesn't mean all 11_0/11_1 cards will get D3D 12). The question is: does feature level 12_0 exist (and if so what new features will be available) or have we been mislabeling 11_3 as 12_0? My guess is 12_0 still exists (but perhaps not finalized).

Also, I don't see how current 11_1 cards could support 11_3 through a software update. 11_3 includes new non-trivial hardware requirements.
 
More cards coming based on GM204

Another version of the chip, with 13 SMs, will ship concurrently and be called GeForce GTX 970. In the future we plan to offer additional products based on GM204 that will ship with different combinations of GPCs, SMs, and memory controllers to address various segments of the graphics market.

---

From page 6 of this whitepaper

http://international.download.nvidi...nal/pdfs/GeForce_GTX_980_Whitepaper_FINAL.PDF
Note the plural "products".

So probably a GTX 970ti and the GTX 960 at least.
 
Back
Top