Vega Rumors

[Ieldra]

TSMC 10nm is already in shipping products, like the iPad Pro 10.5". Why would things need to go painfully wrong for NVidia to use TSMC 10nm?

There's a big difference between a tiny mobile soc and a gpu that can approach reticle limits, they dont use the same libraries, which are essentially sets of operating parameters used in the design that determine cell density, routing density etc. Needless to say a large gpu like Vega 20 which surely will push past 4096 units and will support DP as well existing Vega features will require far more complex routing logic than apples iPad soc.

Nvidia announcing this 12nm FFN node they codovelpped with tsmc is an interesting development. They've been developing their libraries in house for years now and you can see that paying off big time when you look at maxwell and Pascal efficiency and the high clocks they can achieve. Now it appears they are working with the foundry to tune their process around their needs as well, I am quite optimistic about Volta's operating frequencies

 

[tybert7]

Really?

GF has had more problems with node transitions than ANY other foundry. That is why they licensed 14nm from Samsung lol. Its nice to rewrite history but doesn't work. Even with a more "advanced" 14 nm process from Samsung, they are still behind nV's chips on a "less advanced" 16nm process.

The rest of it is just not much sense, if a person that is in the market for a new card, what will they buy, last years products? No they will get the best for the money with today's product.

They are still way behind the mhz curve with their process, because their architecture needs more voltage to sustain those clocks. Much more then just pushing clocks man. Vega is like Intel's Pentium 4, obscene voltages for the clocks they are getting. Leakage is high too, check Pcper and GN articles on what temps do to Vega's leakage. The same temps don't do that to nV's chips. Why is that? The only reason to increase voltage is to maintain signal cohesion at a certain temp.

Had more problems. Global Foundries was like the Enterprises warp engines in star trek the motion picture before Spock (Samsung) came in with his enhanced warp field equations.

 

[razor1]

pun intended lol.

 

[tybert7]

Where do you all think Vega 64 performance will land? The air version? Right in line with a 1080? below? I assume the liquid version will try to go a bit above a 1080 since it costs more, but you never know based on what they sold the fury x for.

I am curious about games that are not heavily favored for AMD like Doom on Vulkan. I fully expect amd to be able to at least match a 1080 there, but what about other titles? Like frostbite 3 titles? I still need to finish mass effect andromeda, only did 2 planets and stopped due to being busy with other things and wanting to wait so I could actually play in 4k on my tv (not on ultra, I can turn some settings down)

 

[razor1]

the air version right around the 1080, the liquid cooled just above

 

[Gideon]

the air version right around the 1080, the liquid cooled just above

That would be what I expect at this point as well. Also the pricing seems to reflect this, but reviews will be out shortly.

 

[Hameeeedo]

I expect air Vega to be consistently less than 1080 in the majority of games, and slightly faster or on par with it in DX12 games. Liquid Vega is going to be on par with the 1080 in majority of games, but slightly and consistently faster in DX12 games. This is based soley on AMD's cherry picked games which they populated with DX12 games only.

 

[Snowdog]

Nvidia announcing this 12nm FFN node they codovelpped with tsmc is an interesting development. They've been developing their libraries in house for years now and you can see that paying off big time when you look at maxwell and Pascal efficiency and the high clocks they can achieve. Now it appears they are working with the foundry to tune their process around their needs as well, I am quite optimistic about Volta's operating frequencies

I remember the Video someone posted here about a field trip to a NVidia lab where they had all of this crazy expensive new state of the art equipment for testing chips. Some of the things they could check really surprised me, this was many years ago and it was clear then how much feedback they were giving TSMC.

Because of that and things like 12nm FFN design that I expect better process results from NVidia + TSMC, than AMD + GF.

But 12nm FFN barely moves the meter on transistor density. It should probably be called 15nm FFN. It seems more about power usage than anything else.

 

[JustReason]

Not blind assumptions though. There are plenty of measurements with the effects implemented. Devs have been getting 20-30% boosts with FP16. That's a fact. AMD had a slide with 25% gains from a similar effect for noise and lighting passes. AMD released a chart of bandwidth and work reductions with DSBR. Battlefield doing 30% less work resulting in higher FPS isn't much of a stretch. That should work out to a 43% performance increase. Nine other major titles on that chart with varying benefits. The big gains coming from the poorly optimized titles that have been Nvidia's bread and butter in regards to benchmarks. Nothing magic about this, just simple math and common sense. Calling something magic, fantasy, etc is a term used by the ignorant to describe something they don't understand.


So the card without the performance enhancing features enabled is trading blows with 1080 already. Thanks for making my point. Or is this an argument about magical drivers, like what Nvidia just did with Titan on pro apps, not being a valid argument? The math and facts bare out what I've been saying. I don't see what's so difficult about understanding that.


Then make memes if you like spreading more accurate information. The facts are pretty simple, Vega supports a superset of Pascal's features. Pascal isn't designed for async, bindless resources, and other features that DX12 wants. Vega supports features that WILL result in a significant performance increase. Those games are coming. Give it a year and Vega being on par or above Titan Xp isn't a stretch. Assuming Titan can even run everything at max settings. It's hard to be disillusioned when the facts already support the point being made. The only paycheck is probably going to you with all the spin you've been putting on the facts lately. Pretending stuff can't exist, then when it appears forgetting about it or pointing it out while forgetting you said it was impossible.

Case in point below.


So pure magic? Or are gains over 10% with drivers still fantasy?

I was thinking about the features/tech in Vega and thought maybe they put them in Vega as a testing bed and a way to optimize for them before Navi comes in. I do believe it is very possible that Vegas performance could well be climbing on the heels of the 1080Ti/TitanXP, unfortunately it will likely take months for that to happen, at least 6mths even possibly a year. Until reviews, and I mean the indepth kind, we may not know exactly where it stands now and what may be plausible in the future.

I am probably buying one at release. I want the watercooled one, would prove great for my air-conditioned case. I missed the FuryX, and I want to give my wife my 290 to replace the lackluster r7-370 she has now. I am still deciding on which CPU/MoBo combo to get as well as a monitor. Upgrade time.

 

[kcthebrewer]

I do believe it is very possible that Vegas performance could well be climbing on the heels of the 1080Ti/TitanXP.

In all seriousness, based on what?

I have seen no evidence of this and it appears they are putting everything into half precision shaders and hoping it pays off. Until games are released that use this feature, you will unlikely see any change in the cards performance. GCN is GCN. There isn't much left to optimize.

 

[JustReason]

In all seriousness, based on what?

I have seen no evidence of this and it appears they are putting everything into half precision shaders and hoping it pays off. Until games are released that use this feature, you will unlikely see any change in the cards performance. GCN is GCN. There isn't much left to optimize.

Honestly, and not a jab but simple truth, go to other forums for some more indepth discussions on the tech AMD has spoken of and what is possible and what is not. This AMD section is so vile and confrontational that there is no real debate or interest in doing anything more than bashing others and the sections products. Seeing some of the tech and how it works and the implementation of it helps see how it COULD happen, doesn't mean it will.

 

[razor1]

Honestly, and not a jab but simple truth, go to other forums for some more indepth discussions on the tech AMD has spoken of and what is possible and what is not. This AMD section is so vile and confrontational that there is no real debate or interest in doing anything more than bashing others and the sections products. Seeing some of the tech and how it works and the implementation of it helps see how it COULD happen, doesn't mean it will.

err doesn't happen that way man, the only reason it does happen is because AMD has stuck with GCN for so long, nV stops optimizations of older graphics cards once a new architecture come out. Of course they still make sure stability checks are done, but optimizating them, is low priority.

Talked about this before with the g80 and why it lasted so long cause well Tesla was there for 3 graphics card releases.

It all goes back to people wanting to see what they want to see. There is no such thing as AMD fine wine.

https://www.hardocp.com/article/2017/01/30/amd_video_card_driver_performance_review_fine_wine/14

AMD claimed its new AMD Radeon Software Crimson ReLive Edition would net between 4-8% performance increases depending on the game. We can say this is pretty much on the money, at 5-6% in the games we tested AMD’s claims seem valid. That means both the AMD Radeon R9 Fury X and AMD Radeon RX 480 have received a cumulative performance upgrade via drivers, but it is quite small in the grand scheme of performance.

 

[Bandalo]

It all goes back to people wanting to see what they want to see. There is no such thing as AMD fine wine.

I never got this idea of "fine wine"...I look at it as being ripped off. If 2 years after release you can give me 10% more performance, that's great. But that's 10% performance I haven't had for the past two years. Why couldn't you have delivered that when I bought the card?

 

[Anarchist4000]

Anarchist4000 haha!

Yes. Let me know when Vega is outperforming Titan Xp in something that isn't an amd tech demo. I'm still waiting for confirmation of your claimed 40% performance improvements from async.

You have a long list of shit you've yet to prove

So basically no future games are allowed for the comparison? Or they'll be like Doom using "intrinsics", aka features Nvidia doesn't support, to get that advantage. As for the 40% we still haven't seen any games hammer away at async shading. They seem to have difficulty supporting it and bindless resources would help there. At this point I'm guessing Volta will have to release before we really start to see uptake. Anyways I'll be sure to get back to you once everything is enabled and the features are used.

Which games are those?

Anything Bethesda with that "long term strategic partnership" seems rather likely. So Wolfenstein, probably Doom in a patch, future Fallouts, TES if they make more, and some other upcoming titles from Bethesda. Any upcoming DICE titles most likely. I think Bioware mentioned it. Then anything coming from console that is already using packed math.

I was thinking about the features/tech in Vega and thought maybe they put them in Vega as a testing bed and a way to optimize for them before Navi comes in. I do believe it is very possible that Vegas performance could well be climbing on the heels of the 1080Ti/TitanXP, unfortunately it will likely take months for that to happen, at least 6mths even possibly a year. Until reviews, and I mean the indepth kind, we may not know exactly where it stands now and what may be plausible in the future.

Some features are definitely in there as a testing bed, likely at the request of developers. Primitive shaders being a big one as devs have already presented papers on the techniques. The DX12 features though will probably be seeing some use within the next year. Engines are finally starting to adopt them. At least Lumberyard just added conservative raster and OIT, Unreal and Unity in similar boats.

As for performance Vega has a slight advantage in technical specs and the software feature with TBDR that accounts for much of Nvidia's "more efficient" architecture. Only difference being with all the cache on Vega it might actually be more efficient than Nvidia's approach. Nearly twice the SRAM of even P100. So yeah I'm guessing driver support will take some time, but within a year be pretty damn close if not exceeding it with newer titles.

I am probably buying one at release. I want the watercooled one, would prove great for my air-conditioned case. I missed the FuryX, and I want to give my wife my 290 to replace the lackluster r7-370 she has now. I am still deciding on which CPU/MoBo combo to get as well as a monitor. Upgrade time.

I was really hoping for a Threadripper APU if those ever exist and some sort of ultimate ITX build. Like the Project Quantum AMD demoed in the past. Something with 16 cores and mid range GPU performance I might be able to use without dGPUs. Thanks to a dead system fan and now freezing every other day (RIP I7) I'm definitely feeling that upgrade itch as well. All the bundles kind of sucked (no threadrippers, HDR), or I'd have jumped on one. So I'll probably just grab an air cooled version and downclock or maybe get the Nano. Piece together a system from there if availability holds. Limited locations for a second radiator on the water cooled version or I'd definitely consider it. Guess I'm more or less waiting to see what else shows up atm. Even for a monitor I'd rather consider HDR options and ideally keep the two screens I have alongside. Then to get all that running on Linux!

go to other forums for some more indepth discussions

Couldn't agree more. There is very little informational value on anything around here to base decisions on.

 

[razor1]

I never got this idea of "fine wine"...I look at it as being ripped off. If 2 years after release you can give me 10% more performance, that's great. But that's 10% performance I haven't had for the past two years. Why couldn't you have delivered that when I bought the card?

even 10% performance, not really worth it in the long scheme of things. We are talking about depreciated products here, why buy something for the that 10% extra 2 years down the road when its worth less than half that by that time?

 

[razor1]

So basically no future games are allowed for the comparison? Or they'll be like Doom using "intrinsics", aka features Nvidia doesn't support, to get that advantage. As for the 40% we still haven't seen any games hammer away at async shading. They seem to have difficulty supporting it and bindless resources would help there. At this point I'm guessing Volta will have to release before we really start to see uptake. Anyways I'll be sure to get back to you once everything is enabled and the features are used.

nV has their own set of intrensics now fully functional, again, it was just that they wren't exposed in drivers early enough for ID to work on theirs. Even with those doesn't really matter does it?

Anything Bethesda with that "long term strategic partnership" seems rather likely. So Wolfenstein, probably Doom in a patch, future Fallouts, TES if they make more, and some other upcoming titles from Bethesda. Any upcoming DICE titles most likely. I think Bioware mentioned it. Then anything coming from console that is already using packed math.

Partnerships only last if AMD can deliever, just like Nintendo switched over to nV, it will happen if AMD can't make good products.

Some features are definitely in there as a testing bed, likely at the request of developers. Primitive shaders being a big one as devs have already presented papers on the techniques. The DX12 features though will probably be seeing some use within the next year. Engines are finally starting to adopt them. At least Lumberyard just added conservative raster and OIT, Unreal and Unity in similar boats.

rasterization routines that use CRV and ROV's nV has plenty of support of those till Volta comes out so no big deal there.

As for performance Vega has a slight advantage in technical specs and the software feature with TBDR that accounts for much of Nvidia's "more efficient" architecture. Only difference being with all the cache on Vega it might actually be more efficient than Nvidia's approach. Nearly twice the SRAM of even P100. So yeah I'm guessing driver support will take some time, but within a year be pretty damn close if not exceeding it with newer titles.

Won't happen any case
Within a year Volta will be out, what is the point of even bringing that up?

Couldn't agree more. There is very little informational value on anything around here to base decisions on.

Right so when [H] does articles about that very thing you ignore them......

 

[pandora's box]

This whole FP16 vs FP32 debate is giving me flash backs to 2003 when AMD (ATI) could only do FP24 and nvidia could do FP16 and FP32, but FP32 was incredibly slow on the FX5800.

 

[razor1]

This whole FP16 vs FP32 debate is giving me flash backs to 2003 when AMD (ATI) could only do FP24 and nvidia could do FP16 and FP32, but FP32 was incredibly slow on the FX5800.

Ironic isn't it lol. Granted this is a bit different AMD won't suffer from register pressure lol but none the less, its performance is sub par to the competition.

 

[OutOfPhase]

This is a strange thread.

I'm not sure I've seen such incredible speculation and optimism for possible future performance (not even hinted at in the company's own marketing) in a computer component before.

I hope everyone gets the part they want, I truly do. But it does seem a little weird to me how far we're willing to bend all observed evidence to continue towards the desired end goal.

 

[tybert7]

This is a strange thread.

I'm not sure I've seen such incredible speculation and optimism for possible future performance (not even hinted at in the company's own marketing) in a computer component before.

I hope everyone gets the part they want, I truly do. But it does seem a little weird to me how far we're willing to bend all observed evidence to continue towards the desired end goal.

My minimum tolerable expectation was for Vega to at least trade blows with a 1080, I'll probably get that at least so I'll go for it. But I still think AMD needs a page one rewrite of their gpus for navi. Or rather, I hope they did that when they started on it years ago. If Vega is a testing bed for newer features, great, but AMD needs larger leaps to stay in the game with nvidia. Hopefully those server chip profits will help amd lower their debt burden and beef up investments for r&d.

 

[Algrim]

This is a strange thread.

I'm not sure I've seen such incredible speculation and optimism for possible future performance (not even hinted at in the company's own marketing) in a computer component before.

I hope everyone gets the part they want, I truly do. But it does seem a little weird to me how far we're willing to bend all observed evidence to continue towards the desired end goal.

To some, perception is reality.

 

[Gideon]

This is a strange thread.

I'm not sure I've seen such incredible speculation and optimism for possible future performance (not even hinted at in the company's own marketing) in a computer component before.

I hope everyone gets the part they want, I truly do. But it does seem a little weird to me how far we're willing to bend all observed evidence to continue towards the desired end goal.

Well there is some truth to those that have such hopes. Ryzen when launched had some issues with performance and in two months some developers and AMD did some updates that pulled the performance of Ryzen up on lower resolutions where it had suffered before. Ryzen being new technology required some hands on tweaks and it helped it. So based on some new features on Vega it could also benefit with this same support from developers as well. Now there is just no guarantee that will actually happen. If you cut most of the garbage posts in this thread you have about 20 posts or so of useful information but yeah good luck sorting through all of it. But this forum is not the best for AMD info anymore cause far to often the get derailed and you will tend to notice who those posters are that do it. Will see what happens when the reviews post.

 

[Hameeeedo]

Ryzen when launched had some issues with performance and in two months some developers and AMD did some updates that pulled the performance of Ryzen up on lower resolutions where it had suffered before.

In some titles yes, in the majority of titles Ryzen 7 is considerably behind 7700K, especially games that are single threaded bound, and there alot of those around. AMD knows that very well, that's why they used 7700K when benchmarking Vega to put it in the best light possible. So yeah Ryzen 7 is still not the best choice for gamers.

Anything Bethesda with that "long term strategic partnership" seems rather likely. So Wolfenstein, probably Doom in a patch, future Fallouts, TES if they make more, and some other upcoming titles from Bethesda. Any upcoming DICE titles most likely. I think Bioware mentioned it. Then anything coming from console that is already using packed math.

Yeah don't count on AMD partnerships, they always fail. They had one with EA for Mantle, and it failed miserably, NV took Mirror's Edge Catalyst, Titanfall, Mass Effect Andromeda and now NFS Payback/ Battlefront 2 right from underneath them. They had one with Square Enix, and NV crushed them with Rise of Tomb Raider, Just Cause 3, and Nier Automata. Final Fantasy 15 is coming as well. So yeah AMD's partnerships don't last long. and NV is mighty aggressive with their support for games. They now have DX12 async for PhysX, so expect a big push there.

s for performance Vega has a slight advantage in technical specs and the software feature with TBDR that accounts for much of Nvidia's "more efficient" architecture.

It doesn't, it was used for Vega FE in pro apps and the closest they have ever been is GP104 Quadro level of performance, AKA 1080. Also AMD told anand that TBDR will marginally improve performance and power consumption. It's maximum effect is in APUs and mGPUs.

The DX12 features though will probably be seeing some use within the next year. Engines are finally starting to adopt them.

Yeah DX12 had like 13 titles since it's inception, half of them work better on NVIDIA hardware, the other on AMD, really not a great strong point for either of them.

You are talking like NV is a setting duck watching helplessly AMD reaping all the rewards, when in fact the opposite is true, look up how many games had GameWorks effects last year alone and you will know the difference. ME Andromeda, Ghost Recon Wildlands, Conan Exile, Ark Survival. all had NV tech in them, and performance was mighty impressive for the green time. 980Ti completely trashed FuryX in these titles with leads reaching 20%+, even Vega suffers compared to the 1080 in these titles. NV doesn't kid around. And they press their advantages aggressively in the gaming space.

 

[Snowdog]

Well there is some truth to those that have such hopes. Ryzen when launched had some issues with performance and in two months some developers and AMD did some updates that pulled the performance of Ryzen up on lower resolutions where it had suffered before. Ryzen being new technology required some hands on tweaks and it helped it. So based on some new features on Vega it could also benefit with this same support from developers as well. Now there is just no guarantee that will actually happen. If you cut most of the garbage posts in this thread you have about 20 posts or so of useful information but yeah good luck sorting through all of it. But this forum is not the best for AMD info anymore cause far to often the get derailed and you will tend to notice who those posters are that do it. Will see what happens when the reviews post.

AMD's own slides have likely already presented the most AMD friendly outcome. Most reviewers are not as likely to stick to AMD friendly titles as AMD themselves.

So I don't think anyone should get their hopes up that reviews are going to paint a significantly more positive picture than AMD has painted in it's slides/presentation.

 

[exdriver]

I've read a lot of "in depth discussions" that do not lead to "real world performance" today. As such they are largely a waste of time for those that are not involved in development.

So yeah I'm guessing driver support will take some time, but within a year be pretty damn close if not exceeding it with newer titles.

In another year, Q3 2018, it will not matter at all what improvements have been made to this product. I don't think anyone will be interested in stepping back to Vega at that time. By then it's going to be the Vega vs Volta show and I think the previews are bad news for AMD.

Couldn't agree more. There is very little informational value on anything around here to base decisions on.

If you or anyone else can give me a concrete reason today to buy Vega I'm listening. I think all you've got is promises of a future on a product that so far has shown draw more power for the same performance at the same price.

Now, if the price was lower, the power draw less, or the performance better, maybe you'd have something. I don't need to build a new system, so we're just talking about a video card here. Giving credit where it is due I think AMD has a good release with their processors.

 

[NKD]

Nvidia launch cycle is not just about consumer but where they face serious challenges; HPC/automobile-transport/DL but these will require Volta architecture and solutions that apply across all segments.
The idea that Nvidia can delay because AMD is not competive is not 100% true, if that was the case Nvidia would not had released the Titan Pascal back in August 2016 and the HPC equivalents GP102 that included the full die P6000/P40, and why even bother with the GTX1080 back in May 2016 when the 980ti was still outselling Fury X and the 970 was the best selling mid-to-performance card so why replace as well with 1070....
The GTX1080 was sold because there is still synergy (R&D and manufacturing and logistics) overall in GPUs between consumer and professional/HPC; so consumers got the GTX1080/1070 May and the HPC world received the P40 and P4 in September, consumers received the cut down Titan in August and HPC world received the full die P40 in November and at the same time a very efficient GP104 model called the P4 (75W and 5.5 FP32 TFLOPs).

So AMD is not fully holding back Nvidia for detailed reasons given in the past, yeah nvidia do in some part occasionally such as we saw with the GTX1080ti but Nvidia still launched that in March 2017 because it would be pointless to delay that after the Titan xP launching in April and used in DL segment as well as consumers.
By the logic Nvidia strategy revolves around AMD products then Nvidia still should not had released the 1080ti and Pascal Titan refresh until say a month ago rather than March and April, nor in fact quite a few of the Pascal products even before that.


Cheers

at what point did I say nvidia can delay! I just said nvidia doesn't have to release volta early if they don't have the quantities ready. They barely did a test run with V100 where they gave away 15 tesla v100 cards. So they are barely hitting that and its almost q3. I seriously doubt their focus right now is volta for gaming ahead of time. I said they will stick to their schedule. You don't release product ahead of schedule especially when there is no competition. They will release Volta when its ready and ready in good quantities. They will make sure its a perfect launch not half ass launch like AMD.

They released the 1080 when it was its time. Oh wait did you compare this situation to 980ti vs fury x. That shit wasn't even close to the tragedy AMD has on their hands now! common man, we are on the same page here. But you are actually saying nvidia will release volta for your already when they barely are getting ready to launch v100. if v100 launched q1 I would have been all ears. Trust me volta wont be here until 2018 for gaming, until its ready for mass quantities with gddr6. Which will be here in q1. It will be very similar to pascal launch.

 

[JustReason]

In some titles yes, in the majority of titles Ryzen 7 is considerably behind 7700K, especially games that are single threaded bound, and there alot of those around. AMD knows that very well, that's why they used 7700K when benchmarking Vega to put it in the best light possible. So yeah Ryzen 7 is still not the best choice for gamers.

Yeah don't count on AMD partnerships, they always fail. They had one with EA for Mantle, and it failed miserably, NV took Mirror's Edge Catalyst, Titanfall, Mass Effect Andromeda and now NFS Payback/ Battlefront 2 right from underneath them. They had one with Square Enix, and NV crushed them with Rise of Tomb Raider, Just Cause 3, and Nier Automata. Final Fantasy 15 is coming as well. So yeah AMD's partnerships don't last long. and NV is mighty aggressive with their support for games. They now have DX12 async for PhysX, so expect a big push there.

It doesn't, it was used for Vega FE in pro apps and the closest they have ever been is GP104 Quadro level of performance, AKA 1080. Also AMD told anand that TBDR will marginally improve performance and power consumption. It's maximum effect is in APUs and mGPUs.

Yeah DX12 had like 13 titles since it's inception, half of them work better on NVIDIA hardware, the other on AMD, really not a great strong point for either of them.

You are talking like NV is a setting duck watching helplessly AMD reaping all the rewards, when in fact the opposite is true, look up how many games had GameWorks effects last year alone and you will know the difference. ME Andromeda, Ghost Recon Wildlands, Conan Exile, Ark Survival. all had NV tech in them, and performance was mighty impressive for the green time. 980Ti completely trashed FuryX in these titles with leads reaching 20%+, even Vega suffers compared to the 1080 in these titles. NV doesn't kid around. And they press their advantages aggressively in the gaming space.

Just wish to correct your very ignorant point on Mantle. It didn't fail at all, it gave birth to DX12 and Vulkan. And DICE asked for it AMD didn't bring it of their own accord. And had you any experience with it you would understand better that it brought absolutely great performance in the games that utilized it. Hell the best Multi-GPU frametime graphs are the ones for BF1 with Mantle, nearly a perfectly straight line. It served it purpose and got MS to begin to move from the serial nature of DX11 and then moved on to greener pastures.

 

[Ocellaris]

Just wish to correct your very ignorant point on Mantle. It didn't fail at all, it gave birth to DX12 and Vulkan. And DICE asked for it AMD didn't bring it of their own accord. And had you any experience with it you would understand better that it brought absolutely great performance in the games that utilized it. Hell the best Multi-GPU frametime graphs are the ones for BF1 with Mantle, nearly a perfectly straight line. It served it purpose and got MS to begin to move from the serial nature of DX11 and then moved on to greener pastures.

Mantle didn't give birth to DX12. DX12 was in development long before AMD announced Mantle (for one game that mattered).

 

[Hameeeedo]

Just wish to correct your very ignorant point on Mantle. It didn't fail at all

Failed here is for EA's partnership with AMD, not for Mantle. Nevertheless though, Mantle was short lived because AMD couldn't keep backward compatibility in check, new cards didn't run Mantle code well, and drivers had to account for that. Mantle wasn't really a great success either in the games that supported it, Mantle required too much VRAM, increased loading times on VRAM limited cards, and offered minimal fps increases outside of very CPU limited scenarios.

 

[JustReason]

Mantle didn't give birth to DX12. DX12 was in development long before AMD announced Mantle (for one game that mattered).

Prove it.

 

[razor1]

Prove it.

if you have the xbox one sdk and ps4 sdk, easy to see it isn't

 

[Ocellaris]

Prove it.

Intel, AMD, Qualcomm, and Nvidia were all partners with MS announcing DX12. They were working together long before the announcement (which came a few months after Mantle was announced).

 

[Digital Viper-X-]

Vulcan incorporated mantle or parts of it into it's release.

 

[TaintedSquirrel]

MS has probably been working on DX12, in some capacity, since DX11 launched.
The popular argument is that AMD's announcement of Mantle somehow spurred Microsoft to pull DX12 out of their asses. At best, it MIGHT have bumped up their release/announcement schedule.

Because as we all know, if a company hasn't announced something yet, it doesn't exist.
I suppose there's a half-hour long AdoredTV video about it somewhere. Otherwise I don't know why people keep saying it.

 

[JustReason]

Intel, AMD, Qualcomm, and Nvidia were all partners with MS announcing DX12. They were working together long before the announcement (which came a few months after Mantle was announced).

Wrong again but whatever. AMD/DICE announced Mantle in late one year and implemented by March I think the following with DX12 announce late that year. With a great deal of Mantle code within. But had you any real knowledge of the facts you would also know that a DX12-like API was in its infancy back before 2009, especially the WDDM 2.0 part. It was scrapped for some odd reason (insert conspiracy here). But the actual results you see are AMD Mantle implementations. But none of any part of this changes the fact that Mantle gave birth to DX12/Vulkan, in part for features and getting whatever DX12 or DX11.X to finally hit the door.

 

[razor1]

Vulcan incorporated mantle or parts of it into it's release.

Yes Vulkan was based on mantle, but IHV's have bastardized it now, in a good way. Live long and prosper btw

 

[razor1]

Wrong again but whatever. AMD/DICE announced Mantle in late one year and implemented by March I think the following with DX12 announce late that year. With a great deal of Mantle code within. But had you any real knowledge of the facts you would also know that a DX12-like API was in its infancy back before 2009, especially the WDDM 2.0 part. It was scrapped for some odd reason (insert conspiracy here). But the actual results you see are AMD Mantle implementations. But none of any part of this changes the fact that Mantle gave birth to DX12/Vulkan, in part for features and getting whatever DX12 or DX11.X to finally hit the door.

Almost all the features mantle had was already in the DX version in Xbox one and they grabbed some more from the sony's ps4 API.

Look API's mean little, cause the hardware must be able to use those features, they go hand in hand. So for GCN to work right with mantle, guess what, the consoles had to have all those features in place cause the consoles had GCN. And Mantle was backwards compatible with older GCN products.

Vulkan never gave birth to DX12 it pushed DX12 forward to an earlier than expected release. How else do you explain DX12 compatibility with Fermi, even for AMD cards they can't go that far back time wise for compatibility.

 

[Anarchist4000]

I'm not sure I've seen such incredible speculation and optimism for possible future performance (not even hinted at in the company's own marketing) in a computer component before.

It's in AMDs marketing slides. They provided numbers for ten top titles with DSBR enabled. There's one Specview benchmark where that feature provided over 100% gains. These are conservative estimates so I wouldn't call them incredible. The only difference is that AMD hasn't committed to when all the features and performance will be there. Could be in a week, month, or year. I tend to think more in the months range.

In another year, Q3 2018, it will not matter at all what improvements have been made to this product. I don't think anyone will be interested in stepping back to Vega at that time. By then it's going to be the Vega vs Volta show and I think the previews are bad news for AMD.

In another year there will be a refresh of products likely with improved titles to play. The people with a Vega wouldn't be stepping back either as they would already have hardware to run the titles.

It doesn't seem unreasonable to see Vega ahead if Titan at that time, nor is it unreasonable to expect a hardware refresh to compete with Volta. Just look at what the 580 did. Faster HBM, core clocks, etc just from process improvements. No reason there isn't a bigger Vega to compete with Volta. AMD has 10, 11, 12, and 20 on the books already.

If you or anyone else can give me a concrete reason today to buy Vega I'm listening. I think all you've got is promises of a future on a product that so far has shown draw more power for the same performance at the same price.

"Concrete" is the current problem. You're looking at a product that with a refresh in six months and zero hardware changes could claim significant performance increases. The problem is there is no time frame for those changes, or guarantee, but we've seen nothing from AMD indicating the hardware is seriously flawed. Just that they need time to improve drivers.

As for power, I'd wait for reviews of RX with the voltages tuned and power saving features fully enabled. So far all anyone has done is benchmark the worst case scenario with max voltages.

The popular argument is that AMD's announcement of Mantle somehow spurred Microsoft to pull DX12 out of their asses. At best, it MIGHT have bumped up their release/announcement schedule.

The more likely argument is that Microsoft was creating DX11.2/3 as DX12. Most of those features stuck, but the current underlying submission model was adopted after tests revealed how efficient it was.

 

[Digital Viper-X-]

Yes Vulkan was based on mantle, but IHV's have bastardized it now, in a good way. Live long and prosper btw

vulcan for life forget that K nonsense

 

[razor1]

It's in AMDs marketing slides. They provided numbers for ten top titles with DSBR enabled. There's one Specview benchmark where that feature provided over 100% gains. These are conservative estimates so I wouldn't call them incredible. The only difference is that AMD hasn't committed to when all the features and performance will be there. Could be in a week, month, or year. I tend to think more in the months range.

Those are best case scenarios. Will not happen to everything they touch.

In another year there will be a refresh of products likely with improved titles to play. The people with a Vega wouldn't be stepping back either as they would already have hardware to run the titles.

It doesn't seem unreasonable to see Vega ahead if Titan at that time, nor is it unreasonable to expect a hardware refresh to compete with Volta. Just look at what the 580 did. Faster HBM, core clocks, etc just from process improvements. No reason there isn't a bigger Vega to compete with Volta. AMD has 10, 11, 12, and 20 on the books already.

err yah nope. rx580 what did it have 5% more performance for 20% more power usage? Where are they going to go with Vega, a 400 watt card out of the box?

Vega 20 is Vega 10 with DP units, AMD stated Vega 10 isn't for HPC market, Vega 20 is, whats the difference? DP units.

"Concrete" is the current problem. You're looking at a product that with a refresh in six months and zero hardware changes could claim significant performance increases. The problem is there is no time frame for those changes, or guarantee, but we've seen nothing from AMD indicating the hardware is seriously flawed. Just that they need time to improve drivers.

Same thing was said about the FX and r600.......

As for power, I'd wait for reviews of RX with the voltages tuned and power saving features fully enabled. So far all anyone has done is benchmark the worst case scenario with max voltages.

LOL max voltage, when AMD is giving us the TBP's? Come on.

The more likely argument is that Microsoft was creating DX11.2/3 as DX12. Most of those features stuck, but the current underlying submission model was adopted after tests revealed how efficient it was.

The submission model was already there in dx11.x (dx next)

LLAPI's kinda of need that submission model, this wasn't the first time we saw similar models either, all consoles since they started using modern API's have had that type of access.