AMD Says Vega Delays Necessary to Increase Stock for Gamers

[FrustratedGarrett]

I don't buy this one bit. Where was it we read that a couple of 4GB HBM2 chips from Hynix cost somewhere around $150? They don't have enough HBM chips in stock and they can't afford to spend much money on buying these chips. They don't even want to release Vega, because it won't make any money. Vega was intended to be sold for +/$600, but that's not happening because Vega is not good enough, not even close.
That nobody that heads their graphics division and their foreign design teams have failed, miserably.

 

[jabbadap]

If that was the case, Fury has more memory bandwidth

Well yeah fury is quite good in mining, but it has slowly become short of vram.

I didn't know that these were still under NDA. I thought that they were in the hands of tech reviewers.

RX vega will be released 14th of August, and I think NDA for reviews will end in the same day. And yes tech reviewers usually sign that NDA to get the card before hand to make review on release day.

 

[EarthDog]

RX vega will be released 14th of August, and I think NDA for reviews will end in the same day. And yes tech reviewers usually sign that NDA to get the card before hand to make review on release day.

true.

 

[Basard]

That's the equivalent of saying, if it has a steering wheel, then it's a car. Until you realize boats also have steering wheels. And airplanes. And trucks.

There were plenty of architectural changes in vega core that you can't just call it "Fury" or "Fury shrink". Only thing really in common with Fiji is the structural arrangement of 64 CU's, shaders, ROP's and TMU's. And that's about it. Why they aren't getting massive performance gains for it, I don't know. You'd have to ask AMD about that...

I slightly disagree... I'd say its more like sticking a spoiler on a Honda and then calling it a sports car. Or maybe it's more like lipstick on a pig, lol....
And as for the CUs, ROPs, TMUs being the same quantity..... structural arrangement is pretty much synonymous with architecture....
It's a shrunken fury with garnish.... just my opinion....

 

[efikkan]

Rubbish: they could allow reviewers to fully benchmark the cards while waiting for availability but they don't.

Sure, they could easily provide a few hundred cards to reviewers ahead of launch.

As long as miners dont snatch up everything.

Miners can't be blamed for GPU shortages. Sure, they buy a few thousands, but they don't buy millions. Pascal has been in low stocks several times.

Why? No gamer will buy till the driver team gets their heads out of their asses. It's still running with Fiji driver and for some reason incurring a big performance penalty.

No, please stop this BS. It's not running the Fiji driver, that's nonsense.

Stock will be low but didn't the same happened with the Pascal at launch???

Yes, Pascal was in short supply for around two months, but it they sell record volumes. Pascal has been the best selling GPU series in recent years.

 

[Vayra86]

I don't buy this one bit. Where was it we read that a couple of 4GB HBM2 chips from Hynix cost somewhere around $150? They don't have enough HBM chips in stock and they can't afford to spend much money on buying these chips. They don't even want to release Vega, because it won't make any money. Vega was intended to be sold for +/$600, but that's not happening because Vega is not good enough, not even close.
That nobody that heads their graphics division and their foreign design teams have failed, miserably.

But they áre selling the top end RX Vega for 600...

It's just a shame the lower priced versions also carry HBM

 

[MxPhenom 216]

It wont matter. Miners are storing these things in big shipping crates. At Lake Chelan, WA there are miners there that have rigs built in these crates with power hooked up to the crate and distributed inside. Power is cheap there so they load them up with few dozen rigs. All the cards they can get. I talked to a few power engineers there and they were telling ne about it.

The delays wont help availability. These miners are relentless. They will gobble up as mwny cards as possible.

 

[Hood]

They prepared this launch so carefully they didn't even manage to draw the Lyra constellation right in that second picture. They just went with a bunch of lines and dots. So yeah, I'll trust the rest of the marketing surrounding Vega is spot on

Good catch - typical AMD marketing - too lazy to google "Lyra" for the real image. Even when their products are worthwhile, they always screw up the presentation, and usually bork the supply chain as well. The fake "leaks", the bragging about "wins" with cherry-picked benchmarks/games, the ever-receding release dates, the ultimate disappointment upon release - these are all hallmarks of a typical new AMD hardware release. We should all be used to it by now.

 

[Assimilator]

Nobody lies like AMD's marketing department. The real reason Vega is so delayed is obviously because yields and HBM2 availability aren't up to par.

 

[RejZoR]

No, it's a lie. Gamers (sane ones at that) are not going to line up for this garbage. Period.

I might. Out of pure curiosity. It might be a total dud. Or it might turn out to be a gem.

 

[Captain_Tom]

I might. Out of pure curiosity. It might be a total dud. Or it might turn out to be a gem.

That's kinda where I am at honestly.

I know a Vega64 would pay for itself through crypto-mining, and I know it should gain quite a lot of performance over time through refined drivers. Therefore I can justify it at $499, but only at that price (And it better beat the 1080 by 5%+).

 

[RejZoR]

I might even go stupid and buy liquid cooled one. Seeing stores around me are listing them already, it's just a matter of how much they'll crank up the prices because it's new and they want to earn some extra on the release hype...

 

[efikkan]

I know a Vega64 would pay for itself through crypto-mining, and I know it should gain quite a lot of performance over time through refined drivers. Therefore I can justify it at $499, but only at that price (And it better beat the 1080 by 5%+).

The scope of driver improvements are limited, and not even an all-star team of engineers making the best driver ever can make Vega perform greatly.

 

[bug]

I might. Out of pure curiosity. It might be a total dud. Or it might turn out to be a gem.

Realistically speaking, what can turn a 300W, $500 card with the performance of a GTX1080 into a gem at this point?

 

[Basard]

After all the griping I'll just add that we saw some pretty good gains on Fiji from later driver releases.... There's always hope that with this chip, and all the little garnishes of transistors they sprinkled around the Fiji core will hopefully translate into even better performance gain than Fiji got over the months following it's release.
There should be absolutely no reason though for them delay it, other than the drivers just suck so bad that the reviews will be tainted like with Fury. I say do a paper launch with a WORKING system/drivers before releasing another card that just gets crappy reviews. Stock quantities don't even matter as long as you get that part right.
The hardware specs are all right there, it should be an awesome card...
I don't even remember reading more than two reviews of Fury after the first week of it's release.... But there were all these charts popping up everywhere showing gains.... The reviews that people read during release are the ones they remember--that's me anyways. Get it right AMD!!!

 

[Captain_Tom]

I might even go stupid and buy liquid cooled one. Seeing stores around me are listing them already, it's just a matter of how much they'll crank up the prices because it's new and they want to earn some extra on the release hype...

At the moment $499 just seems like the sweet spot in terms of what I would consider Vega's maximum value.

We will have to see though. If Vega64air can comfortable beat the 1080 by 10% or more, then the Liquid version should easily come very close to the 1080 Ti - and that would be the same situation as the Fury X vs 980 Ti: more energy usage for similar performance, late to the party, but in a nicer package. The difference is that Vega actually has enough VRAM, and it is now $100 cheaper (Fury X should have been $600 imo).

 

[erocker]

Thread cleansed, stay on topic.

Thanks

 

[cadaveca]

Realistically speaking, what can turn a 300W, $500 card with the performance of a GTX1080 into a gem at this point?

Well, with 4K adoption going relatively slowly, GTX1080-level performance is still kind of high-end. I mean, that's perfect for 2560x1440, and as such, I don't see Vega's performance as a problem at all, especially given the MSRP.

Sure, it might not be perfect for 4K, but the majority of people aren't going to care one bit about 4K anyway, so to me AMD is actually doing far better than some people may want to suggest.

As to the idea of them delaying to build stock... we'll see. That fact there are "Limited Edition" cards kind makes this idea even weirder in my books. If I can't get a limited edition card wit han AIO one month after launch, to me, AMD missed the mark on the number of cards they needed for release, but that in no way reflects on Vega's performance, or how well-suited to the market it really is.

 

[evernessince]

That's a lie AMD, you screwed up with Design (HBM, GPU both), accept it and work for another day.

EDIT: HBM has costed AMD twice now. The time to market the GPU had to be within the cycle of 1 year but both FURYX and VEGA got delayed due to HBM and GPU being a turd in VEGA.

Yep, because random internet guy always knows more than the company that makes the product. I would agree with you if it was obvious. In this case, it is not, especially seeing as keeping product in stock has been an issue recently. Also, point to HBM as the problem when you have no proof, yeah. It could be the HBM or it could be that AMD has had extremely limited funds for some time now. If you haven't noticed, AMD has not redesigned it's GPU architecture since GCN and the 7970.

 

[Prima.Vera]

If you haven't noticed, AMD has not redesigned it's GPU architecture since GCN and the 7970.

Bingo. And people are still surprised by the mediocre performance...

 

[FrustratedGarrett]

Bingo. And people are still surprised by the mediocre performance...

After GCN was finalized back in 2012 by their entirely Canadian design team, AMD has outsourced pretty much all of their graphics IP design to China. No wonder their graphics chips haven't improved in any noticeable way since GCN1.

 

[RejZoR]

Oh my God, can people stop with the GCN bullshit already? By that logic, GTX 1080 is nothing more than tweaked GeForce 6800...

 

[bug]

Well, with 4K adoption going relatively slowly, GTX1080-level performance is still kind of high-end. I mean, that's perfect for 2560x1440, and as such, I don't see Vega's performance as a problem at all, especially given the MSRP.

Sure, it might not be perfect for 4K, but the majority of people aren't going to care one bit about 4K anyway, so to me AMD is actually doing far better than some people may want to suggest.

As to the idea of them delaying to build stock... we'll see. That fact there are "Limited Edition" cards kind makes this idea even weirder in my books. If I can't get a limited edition card wit han AIO one month after launch, to me, AMD missed the mark on the number of cards they needed for release, but that in no way reflects on Vega's performance, or how well-suited to the market it really is.

That makes in an ok/passable card, not a gem.

Oh my God, can people stop with the GCN bullshit already? By that logic, GTX 1080 is nothing more than tweaked GeForce 6800...

If you say so. https://en.wikipedia.org/wiki/Graphics_Core_Next#GCN_5th_Generation_.28Vega.29

 

[jigar2speed]

If you haven't noticed, AMD has not redesigned it's GPU architecture since GCN and the 7970.

So what delayed the GPU if they have not redesigned its arch ? Are you telling me they have been stock piling since 2 years ? Its commonsense, the culprit is HBM. AMD bet on HBM and it costed them time to market. Lets be realistic here, please.

 

[efikkan]

After GCN was finalized back in 2012 by their entirely Canadian design team, AMD has outsourced pretty much all of their graphics IP design to China. No wonder their graphics chips haven't improved in any noticeable way since GCN1.

And people are surprised why they have stagnated…
Saving money by outsourcing development rarely pays off in the long run.