Is Vega going to be DOA!?

[Arachnotronic]

RX Vega will be better than the 1080 Ti. Volta is a 2018 product, and will probably bring no big innovations (they may use TSMC's new 12nm FinFET rebrand of 16nm FinFET). IMHO NVidia is making the same mistake Intel made.

Volta is an entirely new architecture with substantial changes.

 

[Mopetar]

I guess it depends on your definition of DOA. To me, Fury was DOA. It couldnt even beat a stock 980ti at perf/dollar. If an AMD card doesnt win on perf/$, then it is clearly DOA since they dont have the brand power. The Nano technically wasnt DOA since it filled a niche but AMD isnt going to stay afloat by serving niche markets. They need a 1080ti killer for under $600, and more importantly, a 1070 killer for $300.

I don't think they can get both out of one chip though and there's been very little information about small Vega. Depending on how Vega is binned, I think we realistically see something that slots against the 1080 and 1080 Ti. Now that could mean that the full chip is equal to the Ti and the cut chip is better than the 1080, both are slightly worse performance, both slightly better, etc. but I don't see AMD having an answer to the $300 market with big Vega.

 

[unseenmorbidity]

Volta is an entirely new architecture with substantial changes.

Let's hope they actually properly support DX12 on a hardware level, so they will stop trying to force the industry to remain on DX11 another 2 years.

I don't think they can get both out of one chip though and there's been very little information about small Vega. Depending on how Vega is binned, I think we realistically see something that slots against the 1080 and 1080 Ti. Now that could mean that the full chip is equal to the Ti and the cut chip is better than the 1080, both are slightly worse performance, both slightly better, etc. but I don't see AMD having an answer to the $300 market with big Vega.

I agree. IF, and that's a big IF, they could beat the 1080ti, I don't think they would release it at a heavily reduced cost. The 1080ti is actually priced fairly well. It's in line with the past generations for flagship cards.

 

[Paul98]

This is very much going to depend on what sort of clock speed AMD can get out of this chip. It sounds like it's going to be a good bit different than previous GCN chips.

 

[Mopetar]

I agree. IF, and that's a big IF, they could beat the 1080ti, I don't think they would release it at a heavily reduced cost. The 1080ti is actually priced fairly well. It's in line with the past generations for flagship cards.

Yeah, the 1080 Ti being priced at $700 is probably the best evidence for Vega being a good competitor to NVidia. If NVidia thought AMD wasn't going to flounder again as with Fury, they probably would have sold it at $800 or $900.

I suspect that even if AMD could beat the 1080 Ti by 10% (optimistic estimate) they probably wouldn't price it above $700. If they're closely trading blows with the 1080 Ti or fall just short of it, they probably come in at $600.

 

[Bouowmx]

Cue "Wait for NVIDIA Volta: GeForce GTX 1080 Ti performance for 350-400 USD (GeForce GTX 1170)". Not sure how AMD can answer, given the current schedule execution.

 

[CatMerc]

NVIDIA has no more bullets, full GP102 is as aggressive as they can get until Volta.

Actually, with the density improvements of HBM2 in the PHY department, they can turn GP102 into a 430mm^2 die.

If they go full guns blazing and create a 600mm^2 GP202 or something of the sort, they have quite a bit of headroom left.

 

[crisium]

But if they haven't started the development already, there's nothing they can do. He's right, Volta is the next flagship unless they have a hidden GP101 or something that absolutely no rumour points to.

 

[Qwertilot]

Well, they have started the development, its just that it is Volta

It really wouldn't make much sense for them to do anything else with pascal now. Maybe full GP102 as a normal card if they end up 'having' to.

 

[lobz]

Volta is an entirely new architecture with substantial changes.

That's what she said! ...oh wait...wrong topic

 

[Arachnotronic]

But if they haven't started the development already, there's nothing they can do. He's right, Volta is the next flagship unless they have a hidden GP101 or something that absolutely no rumour points to.

Pascal is done and dusted, has been for a long time.

 

[CatMerc]

But if they haven't started the development already, there's nothing they can do. He's right, Volta is the next flagship unless they have a hidden GP101 or something that absolutely no rumour points to.

Depends on the time table for Volta. A new architecture can take a damn long time, but NVIDIA's cashflow would allow them to fast track a GP101 and have it out within the next 6 months if they need it and Volta isn't too close. Let's assume for a second that Volta is mid 2018 product, would NVIDIA find it acceptable to have an entire year without something that tops the charts?

Remember that NVIDIA does not take kindly to not topping the performance charts. If Vega is indeed faster than Titan Xp, I would not be surprised if NVIDIA has a contingency plan.

 

[railven]

Wow, it has almost been two years since Fury X launched.

It be interesting to see where AMD goes with Vega, but I certainly hope it isn't a race for price / performance. Since that always seems to hurt them more than help in my opinion.

As someone who considers himself still neutral (while I own a G-Sync monitor, I have the means to easily buy a FreeSync if I choose to) I don't see myself vendor lock. But I also haven't seen AMD do much to win me back as a buyer. I personally doubt Vega will be what I want, but I can be wrong.

I don't feel Vega is DOA, but I do feel it will face the same hardships Fury did. ignored by its targeted audience and herald as the price / performance king when AMD eventually has to price cut it to move units. None of which, in my opinion, benefits AMD.

 

[Mopetar]

Probably better for them to just cut the price from the start. AMD probably could have charged more for Ryzen considering how it performs relative to what Intel was charging, but they priced very aggressively. Even if they're close, but not quite Titan X levels of performance, they could gain a lot of traction by charging $500. I think they almost have to do that no matter what because so many people who were in the market for something like Vega already bought 1080 Ti's because those seemed like such a bargain, almost a 50% discount over the previous asking price for a Titan X with similar performance.

I guess I have the luxury of not being locked to either yet as I'm waiting for a good 4K HDR monitor (at a reasonable price) to come out before putting together a new box and might even hold out for Volta.

 

[Paul98]

Fury X was an unbalanced design, it felt like a 290x with more shaders and HBM, not much else. I expect Vega to be far more balanced with fewer bottlenecks. One of the main things keeping NVidia ahead right now is it's clock speed advantage, if AMD can close that gap they should have a good chance to compete with the top end NVidia cards. Take that and architecture improvements, and there is a chance that we could see something very nice.

Now if AMD is unable to increase clock speed much, and the architecture improvements won't be shown in current games it will be an issue.

But right now we just have to wait for more information.

 

[SolMiester]

Guess you forgot about their "execution" that almost killed the company.

The FX,?? thats like 12+ years ago!...LOL

 

[SolMiester]

RX Vega will be better than the 1080 Ti. Volta is a 2018 product, and will probably bring no big innovations (they may use TSMC's new 12nm FinFET rebrand of 16nm FinFET). IMHO NVidia is making the same mistake Intel made.

WOW, that's a lot of confidence in a company that historically doesn't execute well and has failed to find the mark on their last 2 generations

 

[tential]

Wow, it has almost been two years since Fury X launched.

It be interesting to see where AMD goes with Vega, but I certainly hope it isn't a race for price / performance. Since that always seems to hurt them more than help in my opinion.

As someone who considers himself still neutral (while I own a G-Sync monitor, I have the means to easily buy a FreeSync if I choose to) I don't see myself vendor lock. But I also haven't seen AMD do much to win me back as a buyer. I personally doubt Vega will be what I want, but I can be wrong.

I don't feel Vega is DOA, but I do feel it will face the same hardships Fury did. ignored by its targeted audience and herald as the price / performance king when AMD eventually has to price cut it to move units. None of which, in my opinion, benefits AMD.

Amd faces the same problem it always faces of being second to market. Nothing new.

So the only new data point left is the performance. Without that, we have no real idea what to say.

But even if it was fury x level of performance, there will be buyers. I will buy no matter what. I need working freesync and the fastest amd gpu possible to make that work for 4k gaming.

There is a market that isn't going to jump to Nvidia gsync, because the price is just too much for SOME people. So there will be a market for Vega.

This is really more of a fan boy style discussion of will Vega best the 1080ti or titan xp and that really doesn't matter to me as someone who wants Vega. It's whether the price and performance and of reasonable measure, and if Vega fits my needs. Then it's just I don't have a cheaper alternative and adaptive sync is adaptive sync.... I'm not paying 1k+ more for the Nvidia version of it.

 

[sm625]

I don't think they can get both out of one chip though and there's been very little information about small Vega. Depending on how Vega is binned, I think we realistically see something that slots against the 1080 and 1080 Ti. Now that could mean that the full chip is equal to the Ti and the cut chip is better than the 1080, both are slightly worse performance, both slightly better, etc. but I don't see AMD having an answer to the $300 market with big Vega.

The chip is approx 2.3 x polaris 10, and I think AMD is only getting $80 apiece for the cut down dies. That puts a cut down vega with 4GB at maybe $350.

 

[Carfax83]

Could you list any AMD flagship GPU you have purchased ever ? I don't know how old you are but let me know if you bought any of these - Radeon 9700 Pro , Radeon HD 4870, Radeon HD 5870, Radeon R9 290X. If you have bought one of these I can believe you are neutral otherwise I have to say you are on the green camp.

I had a 9700 Pro back in the day, which was awesome, or even legendary. That's what got me to accept the notion of buying AMD. But I also bought two Radeon HD 4870 Asus Dark Knight and put them in crossfire, and they gave me all sorts of problems; including, one GPU not idling on the desktop, flickering textures in the original Witcher and Crysis, texture corruption in NWN2, plus blurry texture filtering compared to my GTX 285 at the time.. Taking one card out solved most of those problems, but then the performance was significantly less than my GTX 285.

All those problems made me return them and also put a bad taste in my mouth for AMD GPUs. Although I do understand that AMD has vastly improved since then, but my overall point remains the same. I'm not running a charity here, so if AMD wants my money, they need to exceed what NVidia is doing.

 

[Carfax83]

Everything in your post made complete sense until you dropped that bomb. Near flawless execution?! Bumpgate alone should help you refrain from such marketing inspired remarks.

Nvidia is clearly in a better place than AMD in terms of consistency and execution, but let's not build shrines of perfection, there is no GPU god to bow to.

What part of "near" don't you understand? Of course NVidia has had their fair share of screw ups, but compared to AMD, they are basically top notch. NVidia might screw up every now and then, but with AMD, it seems there's a pattern of screw ups, one after the other to include some truly devastating ones.

 

[wanderica]

I'm sort of with the OP on this one. For me at least, Vega missed the boat. It may be everything we've been hoping for and more, but the 1080 Ti came out first at $700, so Nvidia got my money this round . . . again. I suppose there's the off chance Vega ends up beating the 1080 Ti by 20% or so, however unlikely that scenario may be, in which case I'll gladly upgrade again. Who doesn't love new toys? In the end, I just want AMD to be competitive again. As a consumer, I love having a choice. Right now, I have only one option.

 

[coercitiv]

What part of "near" don't you understand?

Do you understand what flawless means in real terms, and not marketing speech? You putting "near" next to flawless does not magically turn wood screws into finished products.

Of course NVidia has had their fair share of screw ups

Then stop talking like it hasn't. Talk about their really good execution lately, since they deserve praise for that and is also one of the reasons Vega is facing a difficult launch (besides AMDs own faults), but don't make up some imaginary "near flawless" track record for a company you do not represent.

 

[tamz_msc]

Taking one card out solved most of those problems, but then the performance was significantly less than my GTX 285

The GTX 285 was not 'significantly' faster than the HD 4870, considering its price. 10-20% is not significant, especially when the HD 4870 was costing half as much during the launch of the GTX 285 and the debacle that was the GTX 280.

Want a fair comparison? Compare the HD 4890. The HD 4890 and the GTX 275 offered 95% of the GTX 285 performance for 100$ less. The HD 4890 even beat the GTX 285 at lower resolutions in some cases.
http://www.anandtech.com/show/2745

The HD 4000/5000 series were hands down the better lineup than the GTX 200/400 series. Oh and good job equating Crossfire issues with general bad driver support from AMD. I don't deny that the gap between NVIDIA and AMD drivers were wider at the time than they are today, but the way you worded it had a more negative connotation than the reality.

 

[Dribble]

The biggest problem for Vega is Rizen. Simply AMD lives or dies depending on how well Rizen can do, not Vega. Hence resources will all go the Rizen - and while sure some stuff is done by separate teams all the shared stuff will be prioritising Rizen hence pushing back the Vega release.

The second problem is they've made Vega so expensive - huge chip, HBM memory, etc. That really doesn't work unless you can demand premium prices, which won't happen as they are late and chasing the competition. Hence they are unlikely to make much money, which gives it even less importance within AMD.