AMD VEGA official details

[dalekphalm]

Just now, Bleedingyamato said:

Making NVIDIA continue to improve is good but I hope Vega doesn't compete too well with it.  

 

I'm strongly considering getting a 1080 ti once I start seeing non-reference versions so I'd be annoyed if Vega made getting a 1080 ti somewhat pointless.  ?

 

 

I'm thinking I'd look at EVGA most likely for a 1080 ti.  I'm rather curious to see hoe their new ICX cooler design works since I'd heard Gigabyte 10 series GPU coolers are supposedly louder than other brands.  

 

 

I thought all the high end Ryzen CPUs had already been released?

 

So there hasn't been any idea how Vega will perform vs a 1070, 1080, or 1080 ti?

Correct - there are very vague references, but most are rumour. Realistically, we have no idea where VEGA will land.

 

In my personal opinion, if it can't match the 1080 Ti, that's a disappointment. Why? Because NVIDIA has had 1080 Ti performance via the Titan XP since Pascal launched almost exactly a year ago (May 2016).

 

Now they could still be competitive and interesting if they can match the GTX 1070 and GTX 1080 w/ good pricing. But still, they're in a rut and they need to break out of it.

 

Also: About high end Zen

We've seen the release of "high end mainstream" Zen, in the form of Ryzen 7. AMD has been talking about a 32C/64T Zen based CPU for over a year - but it'll be a Server based chip. They may or may not release a "High end Desktop" version (probably not though).

[Bleedingyamato]

11 minutes ago, dalekphalm said:

Correct - there are very vague references, but most are rumour. Realistically, we have no idea where VEGA will land.

 

In my personal opinion, if it can't match the 1080 Ti, that's a disappointment. Why? Because NVIDIA has had 1080 Ti performance via the Titan XP since Pascal launched almost exactly a year ago (May 2016).

 

Now they could still be competitive and interesting if they can match the GTX 1070 and GTX 1080 w/ good pricing. But still, they're in a rut and they need to break out of it.

 

Also: About high end Zen

We've seen the release of "high end mainstream" Zen, in the form of Ryzen 7. AMD has been talking about a 32C/64T Zen based CPU for over a year - but it'll be a Server based chip. They may or may not release a "High end Desktop" version (probably not though).

They weren't able to match Intel with Ryzen so what makes you think they'll do better with Vega against Nvidia?

 

Competition is good but being realistic I'm not sure there's cause to get your hopes up.  

 

Considering what AMD did with the 400 series GPUs I'm guessing Vega won't match a 1080 ti.  

[dalekphalm]

Just now, Bleedingyamato said:

They weren't able to match Intel with Ryzen so what makes you think they'll do better with Vega against Nvidia?

 

Competition is good but being realistic I'm not sure there's cause to get your hopes up.  

 

Considering what AMD did with the 400 series GPUs I'm guessing Vega won't match a 1080 ti.  

Well, a number of things make it quite possible:

 

1. CPU-side, they were behind for 6 years. Bulldozer was a flop (It was actually an ingenious idea, but totally awful implementation and design)

2. GPU-side, they've been pretty well neck and neck with NVIDIA for years now. Even Fiji was almost as good as Maxwell.

 

GPU-Side AMD isn't really "far behind" in terms of tech. In some areas, yes, they've fallen behind. In other areas, they strove ahead. Basically up until current gen (Polaris and Pascal), they've been trading blows. And even current gen, on the low-to-mid range (480 and down), AMD is still very competitive.

 

Personally, I have no idea if VEGA will be as good as it needs to be. But neither does anyone else, aside from people under NDA's.

[NumLock21]

I added TL:DW into OP

 

TL:DS?

 

99% of the content is a repeat. Can't remember if the new information below are repeat or not.

• AMD Vega is launching soon.
• HBM is 5x faster than GDDR5
• VEGA is designed to be scalable architecture
• Mobile VEGA
• Allows VEGA to be implemented into notebooks and other mobile devices. Letting users continue to experience VR and enjoy AAA games.
• AMD VEGA uses HBM2
• AMD RX VEGA will have 2 vram configurations. It will have 4GB HBM2 and 8GB HBM2

 

I can't have sound now, so I used CC.

VEGA turned to BEGA and HBM turned to HPM and HPMC

High Bandwidth Cache is High Bandwidth Cash.

 

and...

 

 

 

[Armakar]

1 hour ago, Bleedingyamato said:

Making NVIDIA continue to improve is good but I hope Vega doesn't compete too well with it.  

 

I'm strongly considering getting a 1080 ti once I start seeing non-reference versions so I'd be annoyed if Vega made getting a 1080 ti somewhat pointless.  ?

 

 

I'm thinking I'd look at EVGA most likely for a 1080 ti.  I'm rather curious to see hoe their new ICX cooler design works since I'd heard Gigabyte 10 series GPU coolers are supposedly louder than other brands.  

 

 

I thought all the high end Ryzen CPUs had already been released?

 

So there hasn't been any idea how Vega will perform vs a 1070, 1080, or 1080 ti?

.. You'd be annoyed if Vega made a card better then the 1080ti at a (likely) cheaper price?

[DaRk0]

6 hours ago, Princess Cadence said:

I hope Vega competes with the 1080ti solely to force nVidia do even better on Volta [:

 

This. I really want to see what Nvidia will do with Volta. My 1070 will have to stay strong until then.

[Dabombinable]

Ryzen+Vega=fuck off Trinity. Its the only way I'll ever get a notebook which flogs the crap out of my Asus U38N, and replaces my old Phenom II X4 based laptop completely.

[zMeul]

2 hours ago, NumLock21 said:

99% of the content is a repeat. Can't remember if the new information below are repeat or not.

• AMD Vega is launching soon.
• HBM is 5x faster than GDDR5
• VEGA is designed to be scalable architecture
• Mobile VEGA
• Allows VEGA to be implemented into notebooks and other mobile devices. Letting users continue to experience VR and enjoy AAA games.
• AMD VEGA uses HBM2
• AMD RX VEGA will have 2 vram configurations. It will have 4GB HBM2 and 8GB HBM2

• ^TM
• so did HBM, didn't help Fiji a bit
• wtf does that even mean?! I have an aluminium ladder that's scalable
• means you can't mobile Polaris or what?
• see above
• so what?!
• baad idea, we all recall how 4GB of HBM fared on Fiji - and with texture details and resolution going up, the more you need VRAM; no matter how good the compression is

[Dabombinable]

19 minutes ago, zMeul said:

• ^TM
• so did HBM, didn't help Fiji a bit
• wtf does that even mean?!
• means you can't mobile Polaris or what?
• see above
• so what?!
• baad idea, we all recall how 4GB of HBM fared on Fiji - and with texture details and resolution going up, the more you need VRAM; no matter how good the compression is

MY GTX 970 is kind of suffering at 1080p. Space Engineers with the view range above 10KM requires a lot of objects and textures to be rendered (even with them being quite low resolution).

[NumLock21]

25 minutes ago, zMeul said:

• ^TM
• so did HBM, didn't help Fiji a bit
• wtf does that even mean?! I have an aluminium ladder that's scalable
• means you can't mobile Polaris or what?
• see above
• so what?!
• baad idea, we all recall how 4GB of HBM fared on Fiji - and with texture details and resolution going up, the more you need VRAM; no matter how good the compression is

Scalable as in vega is not only for desktops. It can also be in laptops Mobile Vega. 

[zMeul]

Just now, NumLock21 said:

Scalable as in vega is not only for desktops. It can also be in laptops Mobile Vega. 

and Polaris isn't!? I don't get it

[NumLock21]

3 minutes ago, zMeul said:

and Polaris isn't!? I don't get it

Why care about Polaris? Vega can be on mobile, then it's great news.

[zMeul]

Just now, NumLock21 said:

Why care about Polaris? Vega can be on mobile, then it's great news.

so does every other GPU design

hell .. they even put GPUs in laptops that had no place on mobile platforms

[NumLock21]

Just now, zMeul said:

so does every other GPU design

hell .. they even put GPUs in laptops that had no place on mobile platforms

You ever think maybe vega can be reconfigured to be a mobile gpu.

Not just a desktop gpu crammed into a laptop.

[Notional]

42 minutes ago, zMeul said:

• ^TM
• so did HBM, didn't help Fiji a bit
• wtf does that even mean?! I have an aluminium ladder that's scalable
• means you can't mobile Polaris or what?
• see above
• so what?!
• baad idea, we all recall how 4GB of HBM fared on Fiji - and with texture details and resolution going up, the more you need VRAM; no matter how good the compression is

• Within 3 months
• HBM 1 was half speed of HBM2 and wasn't utilized the same way.
• Good, use that ladder as your GPU, so the rest of us don't have to look at your BS anymore.
• Vega is better than Polaris, and with HBM uses less power too. Having that in a laptop is much preferred to GDDR5.
• No.
• It's much much better in terms of heat, power consumption and performance. So better in every way.
• AMD showcased Deus Ex Mankind Divided running perfectly using just 2GB of HBM2. It would need 6GB of GDDR5 to run like that. High Bandwidth Cache is not used like normal Vram.

[Citadelen]

8 hours ago, Princess Cadence said:

-snip-

You want AMD to force Nvidia to do better with Volta so you can buy Volta? Why not reward the company that gives you competition? If that's what you meant then shame on you, mindsets like that are what creates monopolies.

[Princess Luna]

14 minutes ago, Citadelen said:

You want AMD to force Nvidia to do better with Volta so you can buy Volta? Why not reward the company that gives you competition? If that's what you meant then shame on you, mindsets like that are what creates monopolies.

My brother and I have two 1800x systems at the office and I have no issue what so ever advising AMD products when they are the best alternative whenever I'm helping out people in here, you should really go easy on your preconceptions as mindsets like yours are what makes forums healthy exchanges of experience turn into no better than exchange of sharps after who's correct or not.

 

My personal preference for personal use rig concerns just myself, shame on you.

[Citadelen]

23 minutes ago, Princess Cadence said:

-snip-

I added the disclaimer at the end to try prevent exactly what you just did, guess I shouldn't have bothered. Eh, I stand by what I said, buy what serves you best, I just dislike it when people by identicle Nvidia products over AMD products just because it's a Nvidia card. Should both companies have identicle track records and market share, I wouldn't care, but they don't.

[Princess Luna]

Just now, Citadelen said:

I added the disclaimer at the end to try prevent exactly what you just did, guess I shouldn't have bothered. Eh, I stand by what I said, buy what serves you best, I just dislike it when people by identicle Nvidia products over AMD products just because it's a Nvidia card. Should both companies have identicle track records and market share, I wouldn't care, but they don't.

I don't? besides I said competition is good in order to both brands offer good products, how me being all in favour for competition and better products from both translates in favoring monopoly?

 

Nothing against you but you've made very poor sense in what you said, but I get it, you're an AMD fan it is written on your signature even, you want the company to do good and I have no issue what so ever with it, but buying pointless argues with other people for wanting nVidia or Intel won't do AMD any help.

 

Cheers.

[Citadelen]

4 minutes ago, Princess Cadence said:

-snip-

Well clearly I missunderstood what you meant, it would have been nice if you'd cleared that up in the beginning instead of getting defensive and being slightly condescending. Mistakes happen.

 

Cheers.             <-- Don't hurt me for that I couldn't resist.

 

[TheRandomness]

4GB Vega? Oh please be acceptably cheap. 

[Bleedingyamato]

8 hours ago, dalekphalm said:

Well, a number of things make it quite possible:

 

1. CPU-side, they were behind for 6 years. Bulldozer was a flop (It was actually an ingenious idea, but totally awful implementation and design)

2. GPU-side, they've been pretty well neck and neck with NVIDIA for years now. Even Fiji was almost as good as Maxwell.

 

GPU-Side AMD isn't really "far behind" in terms of tech. In some areas, yes, they've fallen behind. In other areas, they strove ahead. Basically up until current gen (Polaris and Pascal), they've been trading blows. And even current gen, on the low-to-mid range (480 and down), AMD is still very competitive.

 

Personally, I have no idea if VEGA will be as good as it needs to be. But neither does anyone else, aside from people under NDA's.

They're still behind now unfortunately for them.  They made up some ground it seems but my impression is Ryzen isn't proving to be the slap in Intel's face that AMD was hoping for.  ?

 

Perhaps AMD will surprise people with Vega then?    Fingers crossed I suppose.   

 

 

7 hours ago, Armakar said:

.. You'd be annoyed if Vega made a card better then the 1080ti at a (likely) cheaper price?

Since I'd already have a 1080 ti yes I would be just a little.  ?

 

 

4 hours ago, Notional said:

• Within 3 months
• HBM 1 was half speed of HBM2 and wasn't utilized the same way.
• Good, use that ladder as your GPU, so the rest of us don't have to look at your BS anymore.
• Vega is better than Polaris, and with HBM uses less power too. Having that in a laptop is much preferred to GDDR5.
• No.
• It's much much better in terms of heat, power consumption and performance. So better in every way.
• AMD showcased Deus Ex Mankind Divided running perfectly using just 2GB of HBM2. It would need 6GB of GDDR5 to run like that. High Bandwidth Cache is not used like normal Vram.

So this cache thing is serving a similar but not identical function to the pool of VRAM on other GPUs?  

 

They're just calling it something different and it's being utilized in a different way?

 

3 hours ago, Citadelen said:

Well clearly I missunderstood what you meant, it would have been nice if you'd cleared that up in the beginning instead of getting defensive and being slightly condescending. Mistakes happen.

 

Cheers.             <-- Don't hurt me for that I couldn't resist.

 

To be fair you were being a bit harsh with her...    

 

3 hours ago, TheRandomness said:

4GB Vega? Oh please be acceptably cheap. 

You wouldn't be happy if it only was 4GB would you?    

 

4GB doesn't cut it nowadays if you want to max out (or come reasonably close) graphics settings in newer games.  

 

RE7 I believe being a good recent example.  It can use more than 4GB at higher settings.  About 4.5-5.5GB I think.  

 

 

Sorry if I seem harsh.  A cheaper Vega card would be very nice if that's what you're hoping for.  I just mesn to say that I think 6GB would be a more suitable minimum at this point.   

[themctipers]

instant regret, buying my gtx 1070

 

regret

regret

regret

REGRET

[Carclis]

4 hours ago, TheRandomness said:

4GB Vega? Oh please be acceptably cheap. 

Is that what he meant though? I know he fumbled his words at certain parts of the presentation, but at that point he was talking about giving AiB partners the option when it came to HBM2 stack capacity. In the context of the slide it almost sounded like he was referring to capacity/stack.

 

 

[Notional]

39 minutes ago, Bleedingyamato said:

So this cache thing is serving a similar but not identical function to the pool of VRAM on other GPUs?  

 

They're just calling it something different and it's being utilized in a different way?

Well it's essentially acting as another caching layer, that can pull data from a lot of sources, like SSD's, HDD's, system Ram and so on. How it will work in practice if difficult to say. You can see if you can make any sense out of PCPer's article on the matter:

https://www.pcper.com/reviews/Graphics-Cards/AMD-Vega-GPU-Architecture-Preview-Redesigned-Memory-Architecture