** The Official Nvidia GeForce 'Pascal' Thread - for general gossip and discussions **

[layte]

Most of these 'credible' people, don't want to lose their nvidia partnerships by posting something negative.

And there we go..... Time to bail until something actually credible appears folks.

 

[Mauller]

And there we go..... Time to bail until something actually credible appears folks.

Like I said, you can check it yourself, you just have to build the dx12 async examples and run gpuview.

But live in your own little bubble if you wish, it wont change things.

 

[LambChop]

****! I was going to buy pascal, now I dont know what to do.

 

[layte]

****! I was going to buy pascal, now I dont know what to do.

Ask your imaginary friends on disqus. Mine say AMD are stinky poop heads who smell of cheese. So no Polaris for me.

 

[Ayahuasca]

****! I was going to buy pascal, now I dont know what to do.

The futures bright, the futures Polaris.

 

[JediFragger]

The futures bright, the futures Polaris.

Brighter than a thousand Suns

 

[LambChop]

The futures bright, the futures Polaris.

Not knowing what to do doesnt mean I'm doing to make a braindead descision

 

[Mauller]

I never said Pascal lacked async. We still know nothing on pascals capabilities.

 

[Orangey]

When I see "disqus" I think wccf comments...

 

[DmitryD]

****! I was going to buy pascal, now I dont know what to do.

Wait for release first, before jumping to conclusions.

Pascal will be improved Maxwell with extra DP processing, which might only be available in compute oriented versions with NVlink - game targeted cards might have a lot less DP as it's not necessary.

I don't know if they could fix async in it - hopefully that's the case, but no way to know now without inside info - don't forget they had extra time to deal with it, maybe what they release is updated Pascal already.

 

[ICDP]

Wait for release first, before jumping to conclusions.

Pascal will be improved Maxwell with extra DP processing, which might only be available in compute oriented versions with NVlink - game targeted cards might have a lot less DP as it's not necessary.

I don't know if they could fix async in it - hopefully that's the case, but no way to know now without inside info - don't forget they had extra time to deal with it, maybe what they release is updated Pascal already.

Get your sarcasm meter checked mate

This is Lambchop he NEVER posts anything negative about Nvidia or ANYTHING good about AMD.

 

[Gregster]

The sources I have are from a disqus chat. People using microsoft dx12 async examples. The nvidia 980ti tested having same fps with async on and off. While the furyx gained 20%+ performance.

Them using gpuview themselves showed only 1 queue being used when async is on or off on the latest nvidia drivers, doing all the above that I mentioned.

You can even run the examples and check yourselves, cant post links atm as at uni on phone.

Don't take this the wrong way bud, but seriously??? Disqus??? Come on lol

And take from this post what you will...

http://www.overclock.net/t/1590939/...ync-compute-yet-says-amd/350_50#post_24898074

I can confirm that the latest shipping DX12 drivers from NV do support async compute. You'd have to ask NV how specifically it is implemented.

 

[Mauller]

As I Sai, get the dx12 async examples yourself and run them. You will get the same fps for async on or off when running a nvidia card. Hence async in the driver just dumps everything async into a single queue as if async was disabled.

You can even use gpuview and observe the behaviour yourself.

All I see is people being ignorant rather than looking into it. Try growing up.

 

[maonayze]

As I Sai, get the dx12 async examples yourself and run them. You will get the same fps for async on or off when running a nvidia card. Hence async in the driver just dumps everything async into a single queue as if async was disabled.

You can even use gpuview and observe the behaviour yourself.

All I see is people being ignorant rather than looking into it. Try growing up.

Well Greg, that there is the chucking down of The Gauntlet. And you always do say that it is better to test things yourself rather than rely on other peoples performance results.

Go get them DX12 Async examples and do some hands on bonefide testing. You could always do a video for us too, so we can take a look to see if the claims are true or false

 

[Gregster]

As I Sai, get the dx12 async examples yourself and run them. You will get the same fps for async on or off when running a nvidia card. Hence async in the driver just dumps everything async into a single queue as if async was disabled.

You can even use gpuview and observe the behaviour yourself.

All I see is people being ignorant rather than looking into it. Try growing up.

Well Greg, that there is the chucking down of The Gauntlet. And you always do say that it is better to test things yourself rather than rely on other peoples performance results.

Go get them DX12 Async examples and do some hands on bonefide testing. You could always do a video for us too, so we can take a look to see if the claims are true or false

I am happy to test anything and show what is what but no idea where to start and what to use. I typed "how to test Async on NVidia?" in Google and got nothing but discussions about Async.

 

[Mauller]

I will find links once i get home.

 

[bru]

So is there a link to these samples and software so we can download them?

If not, you might as well of said just learn to programme and make a test to show the results.

If there is, my google fu has failed me so far.

 

[humbug]

Wait for release first, before jumping to conclusions.

Pascal will be improved Maxwell with extra DP processing, which might only be available in compute oriented versions with NVlink - game targeted cards might have a lot less DP as it's not necessary.

I don't know if they could fix async in it - hopefully that's the case, but no way to know now without inside info - don't forget they had extra time to deal with it, maybe what they release is updated Pascal already.

I hope not, Maxwell needs updating (As does GCN)

I don't want a rehash and AMD does appear to be updating GCN.

 

[Dygaza]

Or you can try MDolenc's Async test from beyond3d forums, and see how new drivers have made difference

https://forum.beyond3d.com/posts/1869354/

 

[Gregster]

lol, same people as usual attempting to rewrite history. nvidia guys, in particular one of you, kept excusing the first Titan's price as being cheap because it offered big DP performance which meant it was great for guys who do compute at home (which is basically no one at all). In fact when everyone else laughed at you guys for throwing £800-900 on a card because it was cheap for home compute/DP performance, the guys desperate to defend Nvidia kept banging on and on and on about how DP was fantastic.

When Titan X came along and was the same price without DP, we pointed that out because the same guys were suddenly billing it as a cheap gaming card and how pointless DP was.

The only people who made a fuss about DP performance itself, was you guys, the people who 'made a fuss' about lacking DP performance in Maxwell were the guys pointing out the hypocrisy of you guys defending the Titan price entirely down to DP/professional card being cheap argument then doing a complete u-turn on this argument when DP went missing in the follow up cards.

DP isn't important for gaming, never was, it also was a horrific reason to 'defend' the first Titan's pricing which as above was compounded when the arguments changed with Titan X losing that single advantage but staying the same price.

The fact also remains that removing compute performance saves power and performance which is a large reason Maxwell is 'more efficient' and smaller compared to Kepler. The fact also appears to remain that as expected Pascal has DP performance and hardware scheduling put back which will have the opposite effect and harm efficiency and increase die size as opposed to a similar gaming powered Maxwell card.

Well let's have a look at history and what you said about the Titan and Titan X.

Spoiler

What on earth are you talking about.

http://www.anandtech.com/show/6774/nvidias-geforce-gtx-titan-part-2-titans-performance-unveiled

The original Titan was often up to 60% faster than the previous faster Nvidia card in the 680gtx, and that wasn't even a full part, the Titan was a salvaged part with cut down shaders/rops/etc.

This card has a much smaller performance gap over the 980gtx, it's mostly between 30-35% faster. It offers NONE of the compute performance, it's hotter, louder, and while 6GB was a sensible amount 12GB means paying for 6GB you aren't going to use for two years... at which point there will be cards with that much memory and twice the performance with lower noise and power that are much more suited to high res/memory usage gaming to begin with.

This card is drastically less impressive in gaming performance, embarrassingly worse in compute performance and worse in terms of heat/noise/value for money as a result.

Then you factor in what happened around the Titan's launch, there was no rumours of a 290x, there was no hints that AMD would come along and challenge performance of a 550mm^2 Nvidia core. This time around the 390x is relatively well known about, not that far away, carries technology on it that is completely up to date, will offer superior(afaik, tbh outside of performance I read very little of the Titan X review) DX12 support and could well end up noticeably faster.

Titan X looks pretty embarrassing as compared to the original Titan, which even at it's absurd cost at least looked like it would be the performance leader for some time into the future, the Titan X can't even claim that, though retains the ridiculous price.

As for the price and not meant to be this or that. Despite Titan and Titan Z being squarely marketed under Nvidia's gaming brand, 99% of the Nvidia guys on here jumped to defend it's cost by saying that it offered professional level DP performance at a lower price for home users..... no one really believed them but it was the only even remotely logical argument to defend it's cost... even that point has been completely removed with terrible DP performance.

Spoiler

Those who find it interesting are pointing out the irony. When we said the Titan was hilariously overpriced and laughed at those that overpaid for them, or laughed at the literally insanely overpriced Titan Z, categorically we were told they weren't aimed at gamers, they were aimed at home workers who wanted DP performance and that is why the price was so high.

Every time the price was brought up an influx of Nvidia users pointed out these WEREN'T aimed at gamers... despite using GeForce, the gaming product brand and not Tesla/Quadro. Look up the Titan Z launch thread, every Nvidia guy was bending over backwards to explain these WEREN'T gaming cards and as such not badly priced.

So Nvidia's users excuse for the entire range of Titan's prices was it being a 'budget' high performance non DP crippled card..............

But when it's pointed out the latest Titan effectively has no DP... it doesn't matter, no one buying a Titan cares about DP performance.

Just a bunch of Nvidia users begging to be ripped off, making excuses one year that directly contradict the excuse they were making the year before. I won't overpay for a graphics card AMD or Nvidia, but I'm consistent in calling a spade a spade. The length Nvidia users will go to justify the absurd pricing and pretend they aren't being ripped off is hilarious.

Seriously last year "it's not expensive, it's a budget professional DP card", this year "it's not expensive, you don't need DP, you don't buy a Titan for DP performance".

I can go on but it was only you who was kicking up a fuss about DP being missing from the TX and claiming everyone was laughing at us Titan/X buyers and insulting post after insulting post from you ensued and now you have the cheek to claim it was only us going on about DP being an issue. Get out with ya. Always personal with you