• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Poll: ** The AMD VEGA Thread **

On or off the hype train?

  • (off) Train has derailed

    Votes: 207 39.2%
  • (on) Overcrowding, standing room only

    Votes: 100 18.9%
  • (never ever got on) Chinese escalator

    Votes: 221 41.9%

  • Total voters
    528
Status
Not open for further replies.
So much argumentative discussion based on an unproven source.
Its also obviously BS anyway because if the Vega 56 was 20% faster than the 1070 then it would be identical to Vega64, there would be absolutely no point AMD in selling the Vega 64.

Vega56 probably is slightly faster than the 1070 and overall is probably more attractive than the Vega64 given cost, power performance. The problem it is not an upgrade from a FuryX.
 
Hoping and praying for some sort of "Sexy beast version" with custom PCB with extra power phases and such.

Didnt somebody qualified take apart the AMD PCB and state its massively over engineered. Opposite to Nvidia, the highest quality make is likely AMD just their cooling is not good. It has 12 phase power on Frontier and is it 10 for Rx, nobody is improving that then ?

Vega 64 is a Furyx on a die shrink
So its not that, it has a lot of features never seen before. If they never get used then it might have been better as a die shrink but we'll never exactly know as its not that.

It does seem waiting for 7nm if you already have a good card is probably the thing to do. As a former two time 2900xt owner, I dont mind fighting back the flames from a hot card in hope it also flies some day :p

I really can't wait for this mining fad to mature into something useful that doesn't involve pouring megajigawatts of electricity into melting GPUs so people can make money speculating on bits of data. Or just die off altogether.

Banks aren't perfect, but at least a big chunk of the money and energy they consume goes into employing people and providing credit which keeps the real economy ticking over.

Crypto isnt likely to die, its a genuine invention. PGP from the 90's was significant and this takes forward progress made and adds further technology advantages. So its unique and its likely required, UK or Germany as some of the worlds richest countries doesnt exactly need it but the world does have a demand for it.

Capitalism and central banking are conflicting principles, theres enough energy wasted then that crypto has some purpose & it seems unlikely to disappear.

Free markets favour efficiency though, Vega doesnt seem a good match in this iteration but its a new industry with inexact demands and not just crypto but AI is distracting AMD from gaming :o


if the Vega 56 was 20% faster than the 1070 then it would be identical to Vega64,
They are presuming overclocking is perfect and cooling is amazing with no throttle. Then also the v64 or 1070 is stock
 
Its also obviously BS anyway because if the Vega 56 was 20% faster than the 1070 then it would be identical to Vega64, there would be absolutely no point AMD in selling the Vega 64.

Vega56 probably is slightly faster than the 1070 and overall is probably more attractive than the Vega64 given cost, power performance. The problem it is not an upgrade from a FuryX.

It probably is BS to be honest. It could be a situation where things are being fed properly in the 64, and the 56 with reduced cores/shaders is at the right level and doesn't lose that much actual performance, but let's be honest.... that is highly unlikely. If anything, a result for a 56 has been posted that has more features enabled in the drivers and now it is faster than the 1070 by a good margin, and the 64 will also gain some good ground on the 1080. However, that also doesn't seem likely given that AMD themselves said the 64 would be trading blows with the 1080.

So in short... I agree with you lol.
 
I still think in my situation Vega 64 is the right choice over a 1080.

Pluses:

Much better support for DX12 and Vulkan. Innovative technology (HBCC, FP16 & DSBR). It's at the start of it's support life cycle.

I own a Ryzen 1700.

Negatives:

Heat, power use. But how much of these do we really care about?

16 months old performance. Volta will crush it when it releases.

-----

But the negatives are also true if I buy a 1080 now. (Except for heat and power issues)

Only pro to buying 1080 now would be I will have a Gysnc monitor so will at least have a monitor that will fully work with all features of Nvida moving forward.

-----

This is assuming when Vega launches that performance is indeed equivalent.
 
$80 a chip? WTF?!?

How much is GDDR5X for the same amount? (obviously there are other costs involved due to needing extra items to feed that GDDR memory compared to HBM, but I am just interested in the comparative chip cost)

The problem with HBM(2) costs is that most people don't have access to it directly at this point - you have to go via a design company who will put your product together on an interposer and the prices are completely different to what a company like AMD or nVidia have access to - $80 is probably what some design company quoted them.

GDDR5 is cheap, GDDR5X a bit more expensive but coming down in price rapidly - IIRC at the volume nVidia buys it is less than 50 cents per GDDR5 chip for them now.

EDIT: To put some perspective on it via my normal supplier it would cost me just over $60 for the GDDR5 on a GTX1070 - it probably costs nVidia less than $15 - I was last quoted that in comparison 4GB of HBM would be $56 and while they couldn't give me a price for HBM2 it would have been over $60 a stack. HBM2 is relatively new and relatively low supply so the cost to AMD won't be hugely lower in bulk than what the design companies would be charging for it to 3rd parties.
 
Last edited:
What ever happened to the Sapphire Vapor cards, they were really nice.
When I asked that question to somebody who works for Sapphire this is the response I got.
"The Vapor X might be dead, though I agree we screwed things by not trying to reinvent it. The whole premise of the Vapor X was making use of a Vapor Chamber for enhanced cooling. However with the Fury and 400 series we have found that there is no need for the use of a Vapor chamber cooling solution. The new heatpipe designs are very efficient and the change in operating temp was within 1 to 2C in most cases, not enough to justify the increased cost of the vapor cooler.

When the entire concept of a gaming lineup was first brought up the two naming conventions I fought for were shot down. The first was to repurpose the name VaporX. Stop the focus on the technology and instead focus on the product name and line up. In my opinion it was one of the best know product names in GPUs and was very well respected. I felt using it would make it easier, we did not need to re-invent the wheel.
 
When I asked that question to somebody who works for Sapphire this is the response I got.
"The Vapor X might be dead, though I agree we screwed things by not trying to reinvent it. The whole premise of the Vapor X was making use of a Vapor Chamber for enhanced cooling. However with the Fury and 400 series we have found that there is no need for the use of a Vapor chamber cooling solution. The new heatpipe designs are very efficient and the change in operating temp was within 1 to 2C in most cases, not enough to justify the increased cost of the vapor cooler.

When the entire concept of a gaming lineup was first brought up the two naming conventions I fought for were shot down. The first was to repurpose the name VaporX. Stop the focus on the technology and instead focus on the product name and line up. In my opinion it was one of the best know product names in GPUs and was very well respected. I felt using it would make it easier, we did not need to re-invent the wheel.

With the increased power usage and heat output of Vega we may see a return of the Vapor X cards.
 
I still think in my situation Vega 64 is the right choice over a 1080.

Pluses:

Much better support for DX12 and Vulkan.
DOes really make sense. Vega does have a higher feature level than Pascal, but Pascal had a higher feature level than Polaris, doesn't really matter since the features don't get used.

Innovative technology (HBCC, FP16 & DSBR). It's at the start of it's support life cycle.
HBCC has no real value in games.
FP16 is not innovative, Nvidia did that 15 years ago.
DSBR is not innovative, nvidia did that with Maxwell, and Imagtech did that 20 years ago.

I own a Ryzen 1700.
not clear that will make any difference.

Negatives:

Heat, power use. But how much of these do we really care about?

16 months old performance. Volta will crush it when it releases.

-----

But the negatives are also true if I buy a 1080 now. (Except for heat and power issues)

Only pro to buying 1080 now would be I will have a Gysnc monitor so will at least have a monitor that will fully work with all features of Nvida.

-----

This is assuming when Vega launches that performance is indeed equivalent.




Power use is irrelevant for a consumser, it is only AMD diehards that say some mindless rant about electricity costs. Its the that that the power produces that is the main issue. Obviously putting more heat into you computer is not great, you can notice the heat build up in your room in the summer, the heat will limit CPU overlooking if not careful, and it is also an issue because he chip has obviosuly been pushed beyond it's sweet spot already so overclocking headroom is restricted. Te extra heat also requires better cooling, but this can often mean faster and louder fans. A good AIB version should largely solve those issues. The power efficiency of a chip is not so much a concern for consumers but it is the best indication of the efficiency of the chip, architecture design and future progress. Quite obviously nvidia could overlcock their chips up to the same power level and gain a massive amount more performance.
 
DSBR is not innovative, nvidia did that with Maxwell, and Imagtech did that 20 years ago.

AMD's implementation of DSBR has some innovations, in some ways it is a much more elegant solution especially in respect to functionality that Maxwell is capable of however from what I can see that comes at a price in that it doesn't have the same ability to be literally flipped on and use brute force to just work like it does on Maxwell requiring more careful implementation at application level to get the best results out of it.
 
With the increased power usage and heat output of Vega we may see a return of the Vapor X cards.
Well he is being very tight lipped atm NDA'S and all that but I think Sapphires Vega cards will have a different cooling solution other than the Nitro series. I loved my 290X Vapor X but when a Palit GTX1080 Jetstream came up for £425 brand new from a different supplier back in May I just could not resist at that price. I am however thinking of buying the Vega RX56 if it performs well.
 
Last edited:
And all this HYPE that derailed after FE got reviewed put me in depression again and had to go back to my doctor for new batch of antidepressants. THIS IS HOW MUCH AMD DISAPPOINTED ME !!!
1 get a dog it helps ..
2 get your missus to shove a finger up your a rse may relax you your way to high strung
3 if all else fails... think it could have been worst you could have been born 100 yrs ago
 
Damn on paper Vega 56 was the card to get. Good specs and reasonable power draw, I bet if you clocked a Vega 56 to Vega 64 levels there would hardly be any notable difference in performance under non a-sync enabled games.
http://www.anandtech.com/show/11690/sk-hynix-customers-willing-to-pay-more-for-hbm2-memory



So much for those AMD diehards who were trying to claim a manufacture's product catalog was wrong.:p Luckily for AMD Samsung was ahead of the curve.
unfortunately this likely means AMD's margins on Vega are going to be quite tight with these memory costs. HBM1 was already largely known to be a lot more expensive than GDDR5, 2.5x that high price has got to hurt.


I guess one "good" thing is that once SK Hynix does get in to mass production their product will likely be better, meaning it will able to hit those 1ghz at 1.2v as opposed to the 945 at 1.35v now. I can see AMD doing a refresh quite soon down the line with a slightly better core speed and Hynix memory clocked higher.
 
Status
Not open for further replies.
Back
Top Bottom