SFF.Network NVIDIA Set to Announce Pascal GTX GPU's Today

We've seen dribs and drabs of NVIDIA's upcoming Pascal refresh for their consumer graphics cards - from leaks of reference shrouds, to rumored (if not highly questionable) benchmarks. But the most recent slew of data has enough substance to suggest that it's the real deal, and the timing of the leak just before NVIDIA's announcement of a livestream event later today all but confirms that we're about to see the next generation of consumer flagships.

Eagle-eyed folks at videocardz.com caught some benchmarks published without identifiers indicating their GPU of origin (though driver names did spill the beans), and from that they've been able to construct a semi-complete specifications table that compares the rumored GTX 1080 and 1070 against the recently-released Tesla P100...

Read more here.
 

Phuncz

Lord of the Boards
Editorial Staff
Moderator
Gold Supporter
May 9, 2015
5,183
4,519
It would be a shame of the GTX 1080 was massively better than the Titan X. When you're "enthusiastic" enough to spend over a 1,000$ for a single GPU, you'd hope that it maintains relevance for a year or two.

The GTX 1070's specs have been announced and it looks like it's going to be performing near the Titan X, that must hurt more.
 

artimaeus

Apprentice
Apr 13, 2016
30
11
It would be a shame of the GTX 1080 was massively better than the Titan X. When you're "enthusiastic" enough to spend over a 1,000$ for a single GPU, you'd hope that it maintains relevance for a year or two.

The GTX 1070's specs have been announced and it looks like it's going to be performing near the Titan X, that must hurt more.

I'd say the Titan X lost relevance after a few months when the 980 ti came out XD
 

MarcParis

Spatial Philosopher
Apr 1, 2016
3,380
2,387
To finish on this upcoming gtx 1080, I wonder if there will be any short pcb version, even on gtx 1080.
Do you have any news related to that?
 

iFreilicht

FlexATX Authority
Feb 28, 2015
3,241
2,355
freilite.com
That would be absolutely amazing, but I think the 1070 will stay the strongest short PCB card. It has a higher TDP than the 980 and more RAM, so without something like HBM, getting a short PCB version would be very difficult. ASRock, I'm counting on you! :D
 

PlayfulPhoenix

Founder of SFF.N
Original poster
SFFLAB
Chimera Industries
Moderator
Gold Supporter
Feb 22, 2015
1,049
1,960
Well, sharing my opinion is hurting people..:)

Sharing your opinion is fine, but you were cherry-picking information to validate it. That's what I take exception to, because you're misrepresenting (intentionally or not) the reality of what this card means to folks. Which, granted, isn't breaking any rules or anything, but you better be prepared to be called on it.

Your opinion, if I understand it correctly, is that you're unhappy with the relative performance improvement. That's totally fine! People have different expectations and different thresholds for that sort of thing, so it's subjective. But that's not what you said, or at least not all that you said. You said that NVIDIA was lying to us about VR performance, and proceeded to misrepresent how their new technology works:

Dirty announcement of VR improvement perf >*2 vs Maxwell generation (Pascal is just generating one image and rotate it to adjust to each eye, thanks to async compute. Maxwell was generating 2 images, one for each eye)

And you said that a >2Ghz overclock required that a GTX 1080 max out its fan speed, even as there exist many counter-examples of reviewers surpassing that with lower fan speeds (not to mention that you stripped important context that even your sole source made sure to include in their analysis):

Annoucement of Overclock above 2Ghz...however, except if you force fan speed to 100% all time, it will throttle down 1.6Ghz

Those aren't opinions, those are stated facts that objectively aren't true. And that's not even all of the misleading or inaccurate stuff that you said!

I have no qualms with your opinion. But I do with bullshit. I'm going to call it, and others are going to call it. That's on you, not us or the community, and that's how great communities work - by filtering the bullshit so the discussions are civil and informed. In this instance, we had three people all bring up separate points that contradicted what you said, with data to demonstrate what you said was bullshit, and we all mentioned diverse sources or linked to them directly. Couching that as us having our feelings hurt is simply disrespectful.

I apologize for making an example out of you, but I think the creeping of incorrect information in discussions like this can seriously pollute how beneficial they are for all of us. And, to be clear, everyone does it. Heck, I did it a few days ago, with a broad statement that @Necere correctly called out as misrepresenting the preference/efficacy of tower coolers. And far from being "hurt", I was thankful! Necere's comment ensured that readers were informed, and I (and everyone else) was better from that.

We should all feel free and comfortable to express our opinions and our knowledge and understanding alike. But we should all also be content with being accountable for what we've said, and be able to recognize when we're either biased or simply wrong, because that's true of all of us at some point. The day we can't maintain that dynamic is the day that SFF Forum becomes a lot less special for all of us.


Sharing your opinion isn't hurting people. You have a right to share your opinion and they have a right to criticize it. Simple stuff. No harm no foul.

Yes, of course. Additionally, if someone states a fact or an objective statement that isn't true, I encourage anyone and everyone to correct that and - importantly - show your sources!

To finish on this upcoming gtx 1080, I wonder if there will be any short pcb version, even on gtx 1080.
Do you have any news related to that?

My expectation is that they won't since they need the space on the 1080 for the memory, and since a shorter profile would be another potential plus in the Titan/1080 Ti column if and when we see such cards. We'd probably see a leak or an early announcement of a "short 1080" if a manufacturer decides to go for it, though, at least in the near future.

That would be absolutely amazing, but I think the 1070 will stay the strongest short PCB card. It has a higher TDP than the 980 and more RAM, so without something like HBM, getting a short PCB version would be very difficult. ASRock, I'm counting on you! :D

I'd agree. Fingers crossed!
 
  • Like
Reactions: Vittra

K888D

SFF Guru
Lazer3D
Feb 23, 2016
1,453
2,902
www.lazer3d.com
Personally I think the performance leap with this new generation is great.

Most people can't afford enthusiast level parts, which personally I would class anything above a 980, mainly because I can't even justify paying for a 970 in my own setup (currently running a 960).

This next generation will bring what I currently consider enthusiast level performance to within my mid range budget. The GTX 1060 may not quite get to 980 levels, but if it gets close that will open up so much potential for the mid range market.

Personally I'm very excited for choosing my next card sometime in the next 6 to 12 months, and hopefully AMD can bring some competition to the mid range market sector aswell to make my decision more difficult!
 

EdZ

Virtual Realist
Gold Supporter
May 11, 2015
1,578
2,107
My expectation is that they won't since they need the space on the 1080 for the memory, and since a shorter profile would be another potential plus in the Titan/1080 Ti column if and when we see such cards. We'd probably see a leak or an early announcement of a "short 1080" if a manufacturer decides to go for it, though, at least in the near future.
It should be possible. The short GTX 670 PCB supported 8 dies per side, and the same layout got carried as far as the GTX 970. The GTX 670 was a 170W TDP card, so the compact power delivery circuitry should be more than sufficient for at least the GTX 1070's 150W TDP, and possibly for a downclocked GTX 1080 (or slightly beefed up components with improved cooling to handle 180W, e.g. a compact AIO-cooled card).

The wrinkle is the GDDR5X traces. Nvidia have said that the signal transmission improvements needed for GDDR5X on the GTX 1080 were also applied to GRRD5 on the GTX 1070 ("NVIDIA tells us that the work they put into the G5X integrate on the GTX 1080 will actually improve performance for the GTX 1070 with G5 memory."), which means the 1080 and 1070 reference PCBs use a different routing than past PCBs. This means that to produce a non-reference short card, OEMs will need to also redo their GDDR trace routing if they make significant layout changes. For the normal "make the PCB substrate bigger and spread some peripheral components out to make the card look more impressive" OEM PCB design this isn't really a problem, but depending on board layers required it may mean the cost of design work for a short PCB 1070/1080 may be higher than for previous generations.
 

PlayfulPhoenix

Founder of SFF.N
Original poster
SFFLAB
Chimera Industries
Moderator
Gold Supporter
Feb 22, 2015
1,049
1,960
The wrinkle is the GDDR5X traces. Nvidia have said that the signal transmission improvements needed for GDDR5X on the GTX 1080 were also applied to GRRD5 on the GTX 1070 ("NVIDIA tells us that the work they put into the G5X integrate on the GTX 1080 will actually improve performance for the GTX 1070 with G5 memory."), which means the 1080 and 1070 reference PCBs use a different routing than past PCBs. This means that to produce a non-reference short card, OEMs will need to also redo their GDDR trace routing if they make significant layout changes.

Given that, do you think a short card is likely if HBM2 isn't that far off?

(I don't really have thoughts or an opinion either way, just curious what you think)
 

MarcParis

Spatial Philosopher
Apr 1, 2016
3,380
2,387
I dream of nvidia gtx 1080 nano or amd vega nano..:) both using hbm 2..:)

They will fit perfectly in my s4 mini, even if they have major power peak due to hbm management..:)
 

EdZ

Virtual Realist
Gold Supporter
May 11, 2015
1,578
2,107
Given that, do you think a short card is likely if HBM2 isn't that far off?

(I don't really have thoughts or an opinion either way, just curious what you think)
That's down to whether the GTX 1070 will work on previous board designs (dependant on how GDDR5 is implemented on die, and anyone who could tell you that will probably be under NDA not to), and whether a particular OEM thinks the market for short cards is big enough to fund development of one.
The GTX 6xx, 7xx and 9xx series were almost 'interchangeable' between PCB designs. You can't just desolder a 670 chip and solder on a 970 chip and have things work, but the board redesigns between generations is in peripheral components rather than the high-speed signalling, so the development work once you already have a short PCB is relatively easy. If this carries over to the GTX 1070 (i.e. you can take an existing short PCB design ,tweak the non-signal elements, and drop in a 1070 chip) then there's a good chance at least one OEM will make a short card to take the 'short card crown' from the R9 Nano. If it requires reworking the PCB design to accommodate tighter requirements for the GDDR5 traces (even if they don't need to hit GDDR5x data rates, the 1070 may have tighter tolerances) then those fixed design costs are higher, and the payoff can be considered to be under a time limit; GDDR5 is on it's way out, and GDDR5X is almost an 'interim' memory for the lucrative high-end. If you design a PCB to handle GDDR5, or even GDDR5X, then if the next generation switches to all-HBM then your PCB is worthless. Costs may be recouped if a future '11xx series' low-mid-range chips stay on GDDR5, but then you're in the volume driven market and margins are lower. Or if an OEM gets wind that some big system integrator wants to put out a compact powerhouse on the mass market and wants a short card to go in it, that may be enough of a 'guarantee' of a future market to reduce the risk of developing a short card.

tl;dr There's nothing technical to prevent it, but it may not be economically worthwhile. You'd have to ask an OEM, and they probably wouldn't be able to tell you without tipping their hand to confidential information (future architecture designs) or their own future market predictions.
 
  • Like
Reactions: iFreilicht

MarcParis

Spatial Philosopher
Apr 1, 2016
3,380
2,387
Don't forget that us prices are annouced without vat...that's explaining most of price convertion..:)
 

GuilleAcoustic

Chief Procrastination Officer
Moderator
LOSIAS
Jun 29, 2015
2,298
3,396
guilleacoustic.wordpress.com
Don't forget that us prices are annouced without vat...that's explaining most of price convertion..:)

Yep, I'm used to that sadly :). I'm currently looking at second hand Quadro K4000 price for my workstation. This is the best single slot single PEG GPU I can find in my price range. I'd love a quadro M4000, but they are the current gen and still go for too much second hand.
 
Last edited:

MarcParis

Spatial Philosopher
Apr 1, 2016
3,380
2,387
On same leaks, we have also Polaris 10 & 11 first feedback, confirming it will be like R9 390X for polaris 10.

And also, Nvidia GP102, GTX 1080ti and GTX Titan revealed...that's true pascal power...they will keep it till AMD shows more stuff..:):)
 

MarcParis

Spatial Philosopher
Apr 1, 2016
3,380
2,387
Well...is GTX 1070 faster than GTX Titan X??? well nope...I know, I'm a bad guy, just reminding Nvidia promises..:)
Clealry GTX1070 is at the same level GTX 980ti...that's pretty good for 200€ cheaper.

Also note that founders edition on GTX1070 is way below than GTX 1080 founder edition. No more vapor chamber on GTX1070.

Anyway, I'm eager to see AMD answers...because leaving Nvidia alone is never good for consumers...