NVIDIA Pascal Thread

[xpea]

Isn't that with 300W power consumption though ? 980ti TDP is 250. I can clock one of my cards to 1600 on the core and it uses 350W+ This is why I don't think clocks will be much better.

300W T100 TDP is absolute maximum value with all FP64 units running and 4 x 40Gb/s NVLinks. Typical FP32 gaming scenario is much less taxing.
Top GP104 SKU with GDDR5X should be very close to 980Ti but we don't compare apples and apples. GP104 is mid range Pascal die where GM200 is big Maxwell

 

[USER8000]

I expect GP104 with GDDR5 will beat a GTX980TI. Remember:
1.)The GTX980TI does not use a fully enabled GPU
2.)The boost is not very high

Now,consider a GP104 should probably have a similar amount of transistors to a GM200,and probably will be a fully enabled part with higher clockspeeds. Then you need to add all the improvements made too.

 

[ShintaiDK]

I expect GP104 with GDDR5 will beat a GTX980TI. Remember:
1.)The GTX980TI does not use a fully enabled GPU
2.)The boost is not very high

Now,consider a GP104 should probably have a similar amount of transistors to a GM200,and probably will be a fully enabled part with higher clockspeeds. Then you need to add all the improvements made too.

I think you have to use cherry picks for that. Remember GTX980Ti got 50% more memory bandwidth, ROPs etc.

Its the GDDR5X edition that will fly.

 

[Qwertilot]

If stock cards are not representative of the actual cards sold, I don't see why it's not "fair". The consumer should get benchmarks which are representative of the real world performance of the card.

Because, if you don't have a reference card how on earth are you going to do reliable comparative reviews in the future? Or now for that matter You'd have to state specifically which OEM card you're talking about on each chart.

Also very handy for judging power efficiency as the huge overclocks the OEM put on chips tend to wreck that etc.

 

[linkgoron]

Because, if you don't have a reference card how on earth are you going to do reliable comparative reviews in the future? Or now for that matter You'd have to state specifically which OEM card you're talking about on each chart.

Also very handy for judging power efficiency as the huge overclocks the OEM put on chips tend to wreck that etc.

Here's a solution: use both. If you don't want to buy some OEM's, OC the reference card (my question remains the same, can nvidia dictate that reviewers can't benchmark the 1080 against a reference OCd 980ti). Regarding the future, I'm not sure how reliable comparisons of future benchmarks are, with driver updates and patches changing performance significantly.

Of note - sites are already using arbitrary GPUs for benchmarks, and people aren't complaining. For example, pcgamershardware is using the EVGA superclocked, 390 PCS and MSI 970.

 

[Ajay]

I think you have to use cherry picks for that. Remember GTX980Ti got 50% more memory bandwidth, ROPs etc.

Its the GDDR5X edition that will fly.

Yes, since, if rumors are correct, only the fully enable GP104-400 dice will be used with GDDR5X. I am wondering what the power consumption will be with this part. The clocks look to be very high and NV will have have to had made a trade between clockspeed, density and power consumption (as in, pick two to optimize and the third will suffer).

 

[xpea]

Yes, since, if rumors are correct, only the fully enable GP104-400 dice will be used with GDDR5X. I am wondering what the power consumption will be with this part. The clocks look to be very high and NV will have have to had made a trade between clockspeed, density and power consumption (as in, pick two to optimize and the third will suffer).

As I already said, Nvidia won't increase Pascal reference card TDP over equivalent Maxwell. It will be suicide in terms of marketing (especially with the node shrink !)
Official reference GTX980 TDP is 165W so you can be sure that GTX1080 won't be higher. And if the leaks are correct, this GTX1080 should offer performances close to 980Ti, maybe even higher at 4k due to the extra bandwidth provided by GDDR5X

 

[tential]

Here's a solution: use both. If you don't want to buy some OEM's, OC the reference card (my question remains the same, can nvidia dictate that reviewers can't benchmark the 1080 against a reference OCd 980ti). Regarding the future, I'm not sure how reliable comparisons of future benchmarks are, with driver updates and patches changing performance significantly.

Of note - sites are already using arbitrary GPUs for benchmarks, and people aren't complaining. For example, pcgamershardware is using the EVGA superclocked, 390 PCS and MSI 970.

And if you want to get early access to Nvidia and amd cards you play by their rules. So I don't know why you're surprised some reviewsites do so.

This is Nvidia by the way. They are going to make their new cards look as good as possible. They specifically make features that only work on the latest gen cards. It's not even a big deal just look at the launch reviews with a grain of salt honestly and wait for the dust to settle.

I think we put way too much credit on launch reviews given the amount of outside influences on many reviewers.

I'm interested in the launch review but I'm more interested in when the aib cards come out what reviews are like then.

 

[ShintaiDK]

Yes, since, if rumors are correct, only the fully enable GP104-400 dice will be used with GDDR5X. I am wondering what the power consumption will be with this part. The clocks look to be very high and NV will have have to had made a trade between clockspeed, density and power consumption (as in, pick two to optimize and the third will suffer).

I would guess in the 150-175W range.

 

[airfathaaaaa]

http://www.nordichardware.se/nyhete...s-till-computex-och-borjar-saljas-i-juni.html

they are expecting full market availability at june

 

[Sweepr]

We got a date for the press event, May 6.

EA will be unveiling Battlefield 5 at its world premiere event this week on May 6, while NVIDIA is reportedly inviting select press from around the world to unveil its new next-gen Pascal-based GeForce GTX 1080, and GTX 1070 video cards.

Coincidence?

 

[S.H.O.D.A.N.]

We got a date for the press event, May 6.



Coincidence?

Probably, given DICE has a long standing relationship with AMD.

 

[USER8000]

Probably, given DICE has a long standing relationship with AMD.

Would be hilarious if Polaris launched the same day too!!

But DICE does have a decent relationship with Nvidia too if you look at BF3 and the latest Mirror's Edge.

 

[S.H.O.D.A.N.]

Fair point, ME does seem to be an NVidia title.

 

[tential]

We got a date for the press event, May 6.



Coincidence?

Unveiling Battlefield 5 on 4x Big Pascal in 8K resolution. (4K + resolution scale this is battlefield guys not some other crappy game that doesn't include these obvious features)

Please?

Oh ya, and unveiled on an HDR 4K resolution monitor.

 

[RussianSensation]

They always use stock cards when they compare them to the other cards. It would be far from being fair to compare overclocked GPU with stock GPU.

They should include after-market 980Ti cards in every review. Nearly every non-reference 980Ti performance 15-20% faster out of the box. They also have another 8-10% overclocking headroom on top of that. If GP104 can also overclock 20-25%, then reference vs. reference is fair to 980Ti. However, if GP104 isn't as good overclocker, then putting reference 1080 vs. 980Ti purposely shows Pascal in good light while ignoring the key reason 980Ti was recommended over Fury X for a year -- overclocking headroom.

Having said that, generally speaking, 980Ti OC owners aren't the target market for GP104. If NV gets its user base to upgrade 2x in a generation, that's a win for them. But I think the target market are Fermi/Kepler and GCN 7900-R9 390 owners.

 

[raghu78]

They should include after-market 980Ti cards in every review. Nearly every non-reference 980Ti performance 15-20% faster out of the box. They also have another 8-10% overclocking headroom on top of that. If GP104 can also overclock 20-25%, then reference vs. reference is fair to 980Ti. However, if GP104 isn't as good overclocker, then putting reference 1080 vs. 980Ti purposely shows Pascal in good light while ignoring the key reason 980Ti was recommended over Fury X for a year -- overclocking headroom.

Having said that, generally speaking, 980Ti OC owners aren't the target market for GP104. If NV gets its user base to upgrade 2x in a generation, that's a win for them. But I think the target market are Fermi/Kepler and GCN 7900-R9 390 owners.

Actually the GTX 980/GTX 970 owners must be feeling the need to upgrade more than any R9 290/R9 290X/R9 390/R9 390X owners. Maxwell GPUs are getting spanked in the recent DX12 titles. 980 Ti seems to be the exception holding up better while the rest are thumped badly.

http://www.overclock3d.net/reviews/...son_16_4_2_-_directx_12_performance_boosted/4

R9 390 is getting 50-70% higher framerates in Quantum Break compared to GTX 970. Thats a whole freaking 2 tiers higher in performance.

https://www.youtube.com/watch?v=zK2BUeYqLVI
http://www.pcgameshardware.de/Quantum-Break-Spiel-15745/Specials/Technik-Test-1191358/

In Hitman again R9 390 is 25% faster than GTX 970

http://www.hardocp.com/article/2016/04/17/hitman_2016_performance_video_card_review/8#.VyUT5TArLIU
https://www.youtube.com/watch?v=fq6GyUzyuJQ

In Ashes the story is similar

http://www.hardocp.com/article/2016/04/01/ashes_singularity_day_1_benchmark_preview/3#.VyUViDArLIU

Right now AMD GCN users are getting better performance than Kepler / Maxwell counterparts. 980 Ti is the one card which is slightly an exception and competitive but thats more due to Fury X scaling poorly wrt Hawaii.

 

[tential]

Using Ashes/Hitman as a reason to go AMD is lame.

Might as well read Pcper's review of the Radeon Pro Duo....

 

[Arachnotronic]

Actually the GTX 980/GTX 970 owners must be feeling the need to upgrade more than any R9 290/R9 290X/R9 390/R9 390X owners. Maxwell GPUs are getting spanked in the recent DX12 titles. 980 Ti seems to be the exception holding up better while the rest are thumped badly.

http://www.overclock3d.net/reviews/...son_16_4_2_-_directx_12_performance_boosted/4

R9 390 is getting 50-70% higher framerates in Quantum Break compared to GTX 970. Thats a whole freaking 2 tiers higher in performance.

https://www.youtube.com/watch?v=zK2BUeYqLVI
http://www.pcgameshardware.de/Quantum-Break-Spiel-15745/Specials/Technik-Test-1191358/

In Hitman again R9 390 is 25% faster than GTX 970

http://www.hardocp.com/article/2016/04/17/hitman_2016_performance_video_card_review/8#.VyUT5TArLIU
https://www.youtube.com/watch?v=fq6GyUzyuJQ

In Ashes the story is similar

http://www.hardocp.com/article/2016/04/01/ashes_singularity_day_1_benchmark_preview/3#.VyUViDArLIU

Right now AMD GCN users are getting better performance than Kepler / Maxwell counterparts. 980 Ti is the one card which is slightly an exception and competitive but thats more due to Fury X scaling poorly wrt Hawaii.

Yes, Hitman (in DX12 mode) and Quantum Break are the big genre-defining titles that many a PC gamer will make her/his PC graphics card decisions on

 

[ShintaiDK]

Yes, Hitman (in DX12 mode) and Quantum Break are the big genre-defining titles that many a PC gamer will make her/his PC graphics card decisions on

Lol!

Cant resist to add this:
http://www.techpowerup.com/222100/hitman-episode-two-update-breaks-directx-12-support

 

[Gundark]

Lol!

Cant resist to add this:
http://www.techpowerup.com/222100/hitman-episode-two-update-breaks-directx-12-support

http://www.guru3d.com/articles-pages/hitman-2016-pc-graphics-performance-benchmark-review,7.html

So, I guess that techpowerup is wrong? Or not?

 

[Sweepr]

Usual dose of salt required.
Ps: As a reference, Mobile Geforce GTX 980 nearly matches desktop Geforce GTX 980.

 

[xthetenth]

Yes, Hitman (in DX12 mode) and Quantum Break are the big genre-defining titles that many a PC gamer will make her/his PC graphics card decisions on

Total Warhammer.

It's gonna be a fun bloodbath.

 

[JDG1980]

Yes, Hitman (in DX12 mode) and Quantum Break are the big genre-defining titles that many a PC gamer will make her/his PC graphics card decisions on

Of course not, but they may well be harbingers of things to come. It's undeniable that titles designed for DX12 lean more towards AMD performance, and this will be a larger percentage of titles as time goes by. The real question is whether this is limited to Kepler/Maxwell or if Pascal will have the same limitations.

 

[IllogicalGlory]

Yes, Hitman (in DX12 mode) and Quantum Break are the big genre-defining titles that many a PC gamer will make her/his PC graphics card decisions on

Hitman episode 2 has been getting almost miraculous reviews, at least for the PS4 version. 8.0 user score on metacritic is pretty shocking, and hell, above 6.0 on the PC (with very positive on Steam) is pretty good too in this age when every (big studio) game ends up with red user scores/mixed reception on steam.

Quantum Break looks like hot garbage, but I hate anything pointlessly scripted like that (to the point where it's basically impossible to play the game differently from anyone else), but apparently I'm in the minority. Everyone seems to love Naughty Dog(sh*t) . There are probably a lot of PC gamers who'd like to play it.