First GP100 "benchmarks" spotted

[Djole123]

So, we know that in the future, some of the Pascal cards will be equipped with the GP100 chip, but here are some spotted benchmarks* for it.

Now why the asterisk? Well, they ain't no 3DMark or PerformanceTest results, no, this is some kind of scientific workload, so raw power of GPU's are tested, and here are some results:

As we can see, raw performance of one of this card is on par with 4 Titan X's! Amazing!

 

I really can't wait and I certiandly look forward to it!

 

Source:

http://videocardz.com/60881/first-and-only-benchmarks-of-pascal-gp100-spotted

[2FA]

There is no GP100, only P100. The G indicates Geforce which there won't be any of since this is being used in the Tesla card.

[Starelementpoke]

3 minutes ago, Djole123 said:

So, we know that in the future, some of the Pascal cards will be equipped with the GP100 chip, but here are some spotted benchmarks* for it.

Now why the asterisk? Well, they ain't no 3DMark or PerformanceTest results, no, this is some kind of scientific workload, so raw power of GPU's are tested, and here are some results:

As we can see, raw performance of one of this card is on par with 4 Titan X's! Amazing!

 

I really can't wait and I certiandly look forward to it!

 

Source:

http://videocardz.com/60881/first-and-only-benchmarks-of-pascal-gp100-spotted

Well, with sli scaling it's more like 2 titan xs.

[vanished]

no 1070?

[Energycore]

8 minutes ago, Djole123 said:

As we can see, raw performance of one of this card is on par with 4 Titan X's! Amazing!

Those tests don't scale very well at all so I'd be careful calling P100 = 4 GM100.

[Starelementpoke]

1 minute ago, Energycore said:

Those tests don't scale very well at all so I'd be careful calling P100 = 4 GM100.

Actually wait, would those tests be done through SLI? Or is that not how this workload works?

[vanished]

1 minute ago, Energycore said:

Those tests don't scale very well at all so I'd be careful calling P100 = 4 GM100.

Yeah that's an understatement!   You can see how "well" these scale by the fact 2 Titan X's was 30% better than a single one and 4 Titan X's was 48% better than a single one.  

[Energycore]

1 minute ago, Starelementpoke said:

Actually wait, would those tests be done through SLI? Or is that not how this workload works?

Probably not. if they're just massively paralellized floating point operations, they can be performed without the cards communicating with each other.

[Notional]

As stated, there is only a P100 compute card with HBM2 (and cut down due to low yields). There won't be a GP100 (Geforce Pascal 100), as they are dropping the HBM2 for gaming cards. So instead we will get a GP102 with Gddr5x. If it will have fewer or more Cuda cores than p100 is difficult to say. Probably less, as P100 is a 300W tdp chip.

[pyrojoe34]

8 minutes ago, Energycore said:

Probably not. if they're just massively paralellized floating point operations, they can be performed without the cards communicating with each other.

Yea, it's unlikely that it's SLI, our computational biology core doesn't run any of our multi-GPU systems in SLI.

[Senzelian]

21 minutes ago, DeadEyePsycho said:

There is no GP100, only P100. The G indicates Geforce which there won't be any of since this is being used in the Tesla card.

We had this nonsense before where everyone was saying that the 980Ti would never exist.

So of course there will be a GP100. 

[Energycore]

Just now, Senzelian said:

We had this nonsense before where everyone was saying that the 980Ti would never exist.

So of course there will be a GP100. 

The thing is, the 1080Ti and Titan P are already more or less locked to be GP102. And P100 is a huge die. I could see a future where P100 doesn't make it to gaming.

[Senzelian]

4 minutes ago, Energycore said:

The thing is, the 1080Ti and Titan P are already more or less locked to be GP102. And P100 is a huge die. I could see a future where P100 doesn't make it to gaming.

Can you not just wait until there is an official statement from NVidia or until the cards are released before making those assumptions? You know that everyone and their dog will simply copy and paste what you said all around the world... And then we end up with a ton of rumors again <.<

[Trixanity]

5 minutes ago, Senzelian said:

We had this nonsense before where everyone was saying that the 980Ti would never exist.

So of course there will be a GP100. 

If that is so, expect it in at least a year, if not later. Not any earlier with the yields they are having. And if they decide to pull the trigger on it, I'm certain it will be a last minute decision because the effort might not be worth the reward.

[Energycore]

1 minute ago, Senzelian said:

Can you not just wait until there is an official statement from NVidia or until the cards are released before making those assumptions? You know that everyone and their dog will simply copy and paste what you said all around the world... And then we end up with a ton of rumors again <.<

I do it on purpose though. Speculation is fun, but if anyone is actually looking to make a purchase, I will be the first one to tell them to wait for official announcements / benchmarks. Until then, shitposting.

[2FA]

23 minutes ago, Senzelian said:

We had this nonsense before where everyone was saying that the 980Ti would never exist.

So of course there will be a GP100. 

That was also before the Titan X was even announced, and I specifically remember people saying there is a possibility that it won't exist, not that it wouldn't definitively. Nvidia doesn't mess around when it comes to codenames,  naming it P100 is very intentional. Many rumors also point towards the GP102 die being used in the next Titan and it's cutdown variant. 

 

FYI the graph benchmark says P100, not GP100. That is the Tesla card in action, not some consumer card that will be used in gaming.

[Big-G]

All I can see it this means dick-all to me as a consumer...

[Coaxialgamer]

We wont be getting gp100 in a titan.  It os simply to expensive to manufacture and has a lot of useless features for consumers