AMD VEGA confirmed for 2017 H1

[mahius]

He didn't lie though. To this day Overclockers dream of OC'ing the Fiji cards.

Lol, so true that it's funny. Thanks for making me laugh in these times when we aren't even sure that consumer Vega GPUs will launch next month...

They can't afford to compete with less than a 1080ti if they're going to keep us waiting longer. They must have £400 1080-beaters if they don't want to fail with a H2 launch for consumer Vega.

 

[CAT-THE-FIFTH]

The cost of faulty dies, interposers and memory will just be amortised into the unit cost available to AIBs. The AIBs have far lower risk and less to test on each card during production as a result.

Totally agree with the issue of not having multiple sources of HBM though. Although it is analogous to AMD being the only supplier of radeon GPUs.

Well technically there is Samsung too,but we have only see HBM2 made by them on high end Nvidia commercial cards,which I suspect means it might be quite expensive?? I dunno,but ATM AMD is pretty much dependent on Hynix delivering. Do Samsung and Hynix have purchase costs for their HBM2 anywhere or is down to the packaging,etc??

I still think as I always have that Vega will be better than what people think. My worry is the price. If they over price it will make me wait 6 month's until Volta which will force AMD to lower price and then buy Vega then. lol. Kind of like what others do, but the opposite way around

As with most launches nowadays there tends to be a bit of a premium for early adopters(unless a company is being aggressive in pricing) - I think I will be more interested to see what technical improvements AMD can make in things like CU utilisation and performance/watt improvements,etc.

Plus TBH even if AMD were to beat a GTX1080TI,I suspect people will then say but Volta is out in 3/6/9 and the cycle continues!!

Lol, so true that it's funny. Thanks for making me laugh in these times when we aren't even sure that consumer Vega GPUs will launch next month...

They can't afford to compete with less than a 1080ti if they're going to keep us waiting longer. They must have £400 1080-beaters if they don't want to fail with a H2 launch for consumer Vega.

I am still not getting why Fuji has anything to do with Vega when it will be two generations newer?? Look at the link posted earlier on - its the same idiotic bickering about the same thing in 2016 in the same thread,and it only takes one or two people to wheel out the same thing for the same thing to happen again.

Its like some really old people who have memory difficulties repeating the same old tail again and again and again.

We have waited months for a bit more information about Vega and once we do have some,its back to wheeling out something about a two year old launch.

 

[r7slayer]

Well technically there is Samsung too,but we have only see HBM2 made by them on high end Nvidia commercial cards,which I suspect means it might be quite expensive?? I dunno,but ATM AMD is pretty much dependent on Hynix delivering. Do Samsung and Hynix have purchase costs for their HBM2 anywhere or is down to the packaging,etc??



As with most launches nowadays there tends to be a bit of a premium for early adopters(unless a company is being aggressive in pricing) - I think I will be more interested to see what technical improvements AMD can make in things like CU utilisation and performance/watt improvements,etc.

Plus TBH even if AMD were to beat a GTX1080TI,I suspect people will then say but Volta is out in 3/6/9 and the cycle continues!!



I am still not getting why Fuji has anything to do with Vega when it will be two generations newer?? Look at the link posted earlier on - its the same idiotic bickering about the same thing in 2016 in the same thread,and it only takes one or two people to wheel out the same thing for the same thing to happen again.

Its like some really old people who have memory difficulties repeating the same old tail again and again and again.

We have waited months for a bit more information about Vega and once we do have some,its back to wheeling out something about a two year old launch.

Yea i see that a lot. Vega is pointless, Volta is around the corner bla bla bla. Vega 10 is for Pascal. You think AMD don't have anything for Volta? Ofc they do and we have even seen hints at Vega 20 which most likely will be aimed at Volta. We all know AMD took a generation break from high end to give them time to develop something that can get them back in the game. They have done this now Vega 10 will be competing with Pascal we all know this. We also have Navi coming out sometime 2018. This could be competing with Volta but personally i doubt it as it would be very late. So what's to stop AMD coming along with the hinted Vega 20? Vega is not pointless.

 

[N19h7m4r3]

Yea i see that a lot. Vega is pointless, Volta is around the corner bla bla bla. Vega 10 is for Pascal. You think AMD don't have anything for Volta? Ofc they do and we have even seen hints at Vega 20 which most likely will be aimed at Volta. We all know AMD took a generation break from high end to give them time to develop something that can get them back in the game. They have done this now Vega 10 will be competing with Pascal we all know this. We also have Navi coming out sometime 2018. This could be competing with Volta but personally i doubt it as it would be very late. So what's to stop AMD coming along with the hinted Vega 20? Vega is not pointless.

Raja also clearly stated that by 2015 the company leaders believed consumer PC graphics was dead. It was all supposed to be consoles, and APUs.

It took AMD putting Lisa Su in-charge, head hunting Raja back from Apple, and Jim Killer from Apple as well to start turning things around for them.

So coming from the mindset and R&D budget of "consumer graphics on PC is dead" to turning it around to compete against NVIDIA is no easy task.
Vega is the first step, and Navi will be Raja's main product since his return; so there's still a lot of work for them to do.

 

[D.P.]

Vega isn't pointless but it is undeniably late. Quite obviosuly if it launched close to Pascal last year things would be very different, instead it launching close to Volta so of course people will be wondering if it is worth waiting, especially given the information that Volta is the biggest archietctural update since Fermi.

Vega 1 might also have short life time of its own with Vega 20 coming 9 months later.

 

[shankly1985]

Zzzzzz don't care I just want Upgrade to Vega. Couldn't care less about what nvidia is doing or working on.

Am gaming at 1440p with a R9 290 for God sake with high/ultra settings if Vega can bring me the same life cycle I have got out of 290 am more than happy.

 

[nashathedog]

Same here it's been a great card to live with and although I was worried about how it would perform at 3440x1440 it's done well, It's paid for itself even though I thought it was overpriced at £440 on release.

he talked about the cooling dissipation capability, he talked about pcb build and with robust power phases, all of it about the package, they never claimed the chip itself will overclock, ppl assume what they want to assume, marketing hammers stuff in for it to stick, doesn't mean you should say what they dont have the courage to say, if they said all that and yet they didn't touch on the chip overclockability, that's a give away already, so don't fill in the blank then come complain about something you said, that's your own conclusion, blame yourself.
ps: when i say you, i mean everyone in general.

The exact point is at 13 minutes 28 seconds

Joe Macri said "We built these for the overclockers, for the gamers, We got a 6 phase, 400 amps of power, I mean it's amazing, your gonna be able to overclock this thing like no tomorrow with the cooling solution, This is an overclockers dream".

 

[shankly1985]

The overclockers dream was very silly of amd should never have been label this without it being able to achieve this.

Not has bad nvidia though and lie about vram or GPU dx12 spec not forgotten about that either.

 

[nashathedog]

The overclockers dream was very silly of amd should never have been label this without it being able to achieve this.

Not has bad nvidia though and lie about vram or GPU dx12 spec not forgotten about that either.

Don't forget the claim of a-sync support during the 980ti release. oops you didn't

 

[Gregster]

he talked about the cooling dissipation capability, he talked about pcb build and with robust power phases, all of it about the package, they never claimed the chip itself will overclock, ppl assume what they want to assume, marketing hammers stuff in for it to stick, doesn't mean you should say what they dont have the courage to say, if they said all that and yet they didn't touch on the chip overclockability, that's a give away already, so don't fill in the blank then come complain about something you said, that's your own conclusion, blame yourself.
ps: when i say you, i mean everyone in general.

He most certainly said "An overclockers dream" you are mistaken here bud.

 

[AllBodies]

Vega isn't pointless but it is undeniably late. Quite obviosuly if it launched close to Pascal last year things would be very different, instead it launching close to Volta so of course people will be wondering if it is worth waiting, especially given the information that Volta is the biggest archietctural update since Fermi.

Vega 1 might also have short life time of its own with Vega 20 coming 9 months later.

Isn't Vega 20 just their "Tesla" card though?

7nm refresh for FP64 and deep learning (maybe they'll also do an Nvidia and put tensor cores in it?)

I got the impression consumers won't see a Vega 7nm refresh, we'll just go straight to Navi 2H 2018.

 

[D.P.]

Don't forget the claim of a-sync support during the 980ti release. oops you didn't

The 980ti has Async support, the issue is ts i not easy to use effectively.

 

[nashathedog]

Isn't Vega 20 just their "Tesla" card though?

7nm refresh for FP64 and deep learning (maybe they'll also do an Nvidia and put tensor cores in it?)

I got the impression consumers won't see a Vega 7nm refresh, we'll just go straight to Navi 2H 2018.

Considering how long it took to move from 28nm is the next move really going to be so quick?

The 980ti has Async support, the issue is ts i not easy to use effectively.

I don't know the details but isn't it done differently to AMD's which is hardware while Maxwells is software?

 

[D.P.]

Isn't Vega 20 just their "Tesla" card though?

7nm refresh for FP64 and deep learning (maybe they'll also do an Nvidia and put tensor cores in it?)

I got the impression consumers won't see a Vega 7nm refresh, we'll just go straight to Navi 2H 2018.

I thought Vega 10 was still their consumer card, they don;t really have separated consumer and HPC products like nvidia Otherwise AMD should have put in some proper 1:2 FP64 support, instead they are just using a consumer card for profesional market as they have done with Fiji and Polaris.

I thought Vega 20 was a refresh of vega 10, not on 7nmbut a tweaked architecture on a tweaked process.

 

[D.P.]

Considering how long it took to move from 28nm is the next move really going to be so quick?




I don't know the details but isn't it done differently to AMD's which is hardware whiule Maxwells is software?

It is done different but the async computing on Maxwell is still done in hardware. There are various limitations with Aysnc on Maxwell that generally meant it was never worth the effort to get it working, and poor implementation would destroy performance so Nvidia for a logn time disabled it in the drivers w and only allowed certain exceptions. When used correctly it would give a small performance increase. GCN was generally easier to get a performance boost but it wasn't without its own pitfalls. Ashed of the Singularity ended up disabling Async support on a number of GCN cards because it would actually lower performance and was taking too much effort to maintain.

 

[spoffle]

It is in same niche as Titan Xp.

I'm guessing Frontier belongs to >$500 spot on this picture.

The Titans aren't compute cards. They are for gaming. I really don't understand why people are in denial about this.

 

[N19h7m4r3]

The Titans aren't compute cards. They are for gaming. I really don't understand why people are in detail about this.

Mostly because when they sent out Titan Xps for reviewers, NVIDIA insisted in they brief that it's not a gaming card :/

 

[nashathedog]

It is done different but the async computing on Maxwell is still done in hardware. There are various limitations with Aysnc on Maxwell that generally meant it was never worth the effort to get it working, and poor implementation would destroy performance so Nvidia for a logn time disabled it in the drivers w and only allowed certain exceptions. When used correctly it would give a small performance increase. GCN was generally easier to get a performance boost but it wasn't without its own pitfalls. Ashed of the Singularity ended up disabling Async support on a number of GCN cards because it would actually lower performance and was taking too much effort to maintain.

Presumably that's what they activated not too long ago then.

 

[spoffle]

Mostly because when they sent out Titan Xps for reviewers, NVIDIA insisted in they brief that it's not a gaming card :/

People have been saying it since the original Titan launched though. NVidia very obviously marketed it as a gaming card, said nothing about compute or HPC.

 

[pete910]

The 980ti has Async support, the issue is ts i not easy to use effectively.

Still flogging that horse ah...

It has compute but CAN NOT do it asynchronously due to lack of hardware which is why it's an issue to use effectively. The driver/software has to do the scheduling.

Oh I forgot.it's D.P (Damage Patrol)