Poll: ** The AMD VEGA Thread **

[pagepack]

It depends how long it takes to get through final validation? Unless you know how AMDs internal process works you can't really guess that.
Also I would like to point to games getting day 1 patches.

I've no idea how amd work, but a point is reached when final testing begins and unless they are complete amatures, that would have started already. With incremental performance improvements being implemented at the same time for future driver releases.

 

[Mauller]

I can't believe that with less than 4 weeks before launch, gaming RX drivers are not in final release testing. Sure amd will be working on future release drivers, but launch drivers will be in their final form with no serious changes unless a major bug is found.

From what I've read FE game mode does exactly zero. Why? Now amd could be completely incompetant, which is a possibility. Or they want to save gaming performance for the official RX launch.

Pro type drivers have an emphasis on rendering accuracy and stability. They have to be code stable on release, Vega FE might not even get the performance boosting features activated for a month or two after RX Vega launch. Since the code base needs to be considered stable before being implemented. The gaming mode just changes radeon settings to allow access to certain gaming features, it still uses the same driver in the end.

But the above is considering the rumours and information being found that Vega FE drivers aren't actually using any of the new vega hardware features yet. The ones that require software to be used at least.

 

[LoadsaMoney]

Who are you, and what have you done to LoadsaMoney?!?!

 

[D.P.]

Pro type drivers have an emphasis on rendering accuracy and stability. They have to be code stable on release, Vega FE might not even get the performance boosting features activated for a month or two after RX Vega launch. Since the code base needs to be considered stable before being implemented. The gaming mode just changes radeon settings to allow access to certain gaming features, it still uses the same driver in the end.

But the above is considering the rumours and information being found that Vega FE drivers aren't actually using any of the new vega hardware features yet. The ones that require software to be used at least.

The pro drivers are independent of the gaming drivers.
The drivers for Vega FE are the same gaming drivers, and then with additions for pro software. They aren't independent.


Blaming Vega FE on drivers is just a complete red herring.

And there is zero evidence that Vega isn't running many of the new features. And in fact many of the improvements shouldn't even need driver support working at the hardware level only. The utility that potentially shows TBR on Maxwell and Pascal GPUs is not guarantee to.work on a different TBR architecture. It is only by exploiting a hack that results in undefined behavior that allowed the utility to work, it was specially designed for Maxwell.

TBR may or may not be working in games. There is zero evdience either way.

For.AMD's sake I hope it is the case but it.doesnt seem likely. And if it is not working in actual games then it points to.some potentially serious issues with the implementation.

 

[pagepack]

But the above is considering the rumours and information being found that Vega FE drivers aren't actually using any of the new vega hardware features yet. The ones that require software to be used at least.

Absolutely, all supposition. I want amd to succeed as one major player in the discrete market would be a disaster for everyone, even those who hate amd.

 

[LoadsaMoney]

We've only had one major player for the past few years now.

 

[TheRealDeal]

We've only had one major player for the past few years now.

You are correct but the really funny thing is AMD grabbed some market share back in the time they couldn't really compete at the high end. Really hoping the RX card is not the wet fart that the FE appears to be. That ti you ordered is taking a very long time though . The 80% extra shader's seem to be working in the non gaming tests most of the time so surely they will do something in the gaming world that we are not seeing so far. I have lost hope in AMD and the whole graphics card pricing as Nvidia will fire at will without competition. Common RX Vega do something bar this crap.

 

[FoxEye]

You are correct but the really funny thing is AMD grabbed some market share back in the time they couldn't really compete at the high end. Really hoping the RX card is not the wet fart that the FE appears to be. That ti you ordered is taking a very long time though .

It probably is (a wet fart). People here are clutching at straws.

Nobody has adequately explained why they would ship a halo card with all/most new features disabled at the driver level.

 

[LoadsaMoney]

That ti you ordered is taking a very long time though .

Cancelled that while ago

 

[TheRealDeal]

Cancelled that while ago

Bad move by the looks of it. As at this Rate i only see Nvidia as an option as if i am shelling out it won't be on this performance.

 

[TheRealDeal]

It probably is (a wet fart). People here are clutching at straws.

Nobody has adequately explained why they would ship a halo card with all/most new features disabled at the driver level.

It's mainly based on what the hell are those 80% extra transistors doing. They appear to be doing nothing in the gaming world. Sure they could just be aimed at compute and stuff but surely some would translate to gaming performance. That's not happening but in a lot of synthetics that are non gaming they are translating to a lot of performance over the fury X. I have seen 4 x faster but that could be memory but in a lot it's around 100%. Again memory could be playing it's part.

 

[N19h7m4r3]

Cancelled that while ago

Pre-ordered 2 Frontier Editions to sidegrade from Furies eh?

 

[silversurfer]

Thoughts? Stick with my original pick? Chicken out and send back? Buying stuff is torture

Use the 1070 and choose between volta or Navi when it comes out, seems Vega is going to take forever to arrive, in hardware and afterwards in software development. Unless you need higher performance then 1070 or wanted some special feature of the Vega. I expect it to take till September for most people wanting vega to get a suitable card, worse case is HBM2 shortage and even further out into autumn

I wanted the latest GCN version plus a lot of what Raja mentions sounds like good features to have. Either he is just a showman, misleading or you'll need a lot of patience to see any of it and if you have the 1070 already, who would wait then. From Frontier so far we have no clue really, it runs the SPEC tests fine seems like so I would guess Frontier is considered an ok launch by AMD. They'll enhance other points later but its doing its main job?

It appears so bad on first contact I'm going to label it as an ironic good sign, this is not even Vegas final form

Spoiler: corporate excess :p

The market for Frontier reminds me of this pic I saw. Some guy goes to his GF's office and she has 4x 1080 in her office pc.

As someone who couldn't careless which manufacturer makes my next gpu, I think AMD are sandbagging, and that FE will get updated drivers on RX release.

I think they just need more time and even after release they will be bringing in improvements. From others comments it appears that they are completely skipping some of the Vega hardware ability till it can be done reliably. So I would agree its going to be a staggered launch and an unclear picture for anyone wanting to buy immediately.
I dont really pay attention on (pre)launches that much for this reason, dust has to settle but if others say this is how Fury went down and how the 480 first appeared then I'll go with that and its probably going that way again this launch

 

[Steampunk]

I guess we have to wait a few more weeks for the RX. Everyone is trying to make guesses based on the FE, but AMD has said they are different beasts aimed at different workloads. I think we'll see a lot of people have made assumptions based on very shaky grounds because the FE is not the same as RX, and you can't extrapolate from one to the other.

 

[muon]

My prediction is RX Vega will trade blows (i.e. equal or marginally behind) with the GTX 1080. However, won't overclock anywhere near as well. AMD will have maxed out the clocks.

 

[TNA]

My prediction is RX Vega will trade blows (i.e. equal or marginally behind) with the GTX 1080. However, won't overclock anywhere near as well. AMD will have maxed out the clocks.

1070/1080 do not overclock well anyway and I have had both. They hit ~1900MHz out of the box; when you OC to 2100MHz you do not even get 10% extra fps...

 

[D.P.]

I guess we have to wait a few more weeks for the RX. Everyone is trying to make guesses based on the FE, but AMD has said they are different beasts aimed at different workloads. I think we'll see a lot of people have made assumptions based on very shaky grounds because the FE is not the same as RX, and you can't extrapolate from one to the other.

BS. It's the same GPU,.and even if we entertain it's for pro use only it works out slower than an 1080

 

[muon]

1070/1080 do not overclock well anyway and I have had both. They hit ~1900MHz out of the box; when you OC to 2100MHz you do not even get 10% extra fps...

I've tested overclocks. Whilst it doesn't scale perfectly, it does scale reasonably.

1900mhz consistent boost on a reference car isn't likely. What you may be seeing is a result of factory OCd versions of the card which have power limits increased for you already over the reference version. That would be why you don't think it isn't scaling properly because it is boosting so high. Boost clocks is as much a function of power limits as it is actually setting higher clocks.

10%+ performance is easy to get on any Pascal card and should be pushing closer to 15% over reference for clocks.

 

[nashathedog]

Pre-ordered 2 Frontier Editions to sidegrade from Furies eh?

Dear God let's hope not.

 

[Rroff]

I've tested overclocks. Whilst it doesn't scale perfectly, it does scale reasonably.

1900mhz consistent boost on a reference car isn't likely. What you may be seeing is a result of factory OCd versions of the card which have power limits increased for you already over the reference version. That would be why you don't think it isn't scaling properly because it is boosting so high. Boost clocks is as much a function of power limits as it is actually setting higher clocks.

10%+ performance is easy to get on any Pascal card and should be pushing closer to 15% over reference for clocks.

My 1070 boosts to 1911MHz out the box and mostly holds it though wanders down 1-2 speed bins at times - slightly nudging the voltage it largely holds around 2100MHz but drops back to 202xMHz at times - actual performance gain averages about 5%.