Wednesday, September 27th 2017

GIGABYTE Has No Plans to Release a Custom Radeon RX Vega 64

In what might be shocking news to AMD fans, GIGABYTE has stated that there are no current plans to make a custom Radeon RX Vega 64. This might change in the future. But for now, early Vega 64 adopters have no choice but to settle for the reference design or or custom design cards coming out from other vendors. There is still a light hope for the Vega 56 though, since GIGABYTE didn't discard the possibility of releasing a RX Vega 56 Gaming G1. However, the actual number of units is still unclear considering that GIGABYTE is unable to start production immediately due to various technical difficulties surrounding Vega.

Due to the inconsistency in quality of chips that AMD are providing, AIB partners are having a difficult time establishing a standard GPU frequency for their factory-overclocked cards. Furthermore, temperature reporting is broken. The actual GPU temperature is different from the temperature reported by the GPU which can become a big problem for stability in the long run. And to top it all off, there are three different Vega 10 GPU packages floating around. The molded package consists of the GPU and HBM dies sharing the same height, while there's a 40 μm height difference between them in the unmolded package. Although it seems insignificant, this small difference prevents manufacturers from standardizing a single heatsink design to accommodate all three GPU packages.
Sources: Tom's Hardware (De), Tom's Hardware
Add your own comment

36 Comments on GIGABYTE Has No Plans to Release a Custom Radeon RX Vega 64

#26
Assimilator
Vya DomusSounds 100% like a supposition to me , there's no real evidence that there is some sort of problem with cooling caused by the different package assembly.

As if AMD with Vega were the first ones to use multiple suppliers for components such as memory or by having multiple package versions. To me this sounds like a ridiculous level of nitpicking. Nvidia quietly changed the supplier for memory modules on their 1070s after every reviewer got the ones with Samsung chips that overclocked far better and it pretty much went under the radar , a few angry customers here and there but overall it's chill . The good ol' double standard.

But sure call them penny-pinching assholes , whatever floats your boat or makes you more comfortable with your next purchases , we all do it I suppose.
Except that memory chips are manufactured to an exact size standard to ensure they can be swapped out. To ensure that this exact issue with fitment isn't an issue. Which in NVIDIA's case it wasn't.

AMD honestly has no excuse for this fucking debacle. We're talking about massive, hot chips, in which a difference in tolerances of under a millimetre make a huge difference to the noise/thermal profile. And you can bet that Gigabyte doesn't have the time or inclination to design 3 different coolers for the potential Vega variants, then have to manually bin each Vega chip they receive to ensure it fits. They're a company out to make money, having to bugger around with Vega chip variances cuts into their profit, so they're going to do the logical thing and just concentrate on NVIDIA cards which are much less painful and hence more profitable to deal with.

If AMD wants to ship its GPUs, it's gotta make it easy for partners to design custom cards based on them, and right now it's doing exactly the opposite. That is their fault, nobody else's, and they need to get their damn act together.
Posted on Reply
#27
Xzibit
MrGeniusWhenever. Right now is fine with me. I noticed it weeks ago when it was first reported. And I made my decision about it then. It's BS. The proper cooling of the HBM is the highest priority of any component on the card. As in the HBM is going to overheat before anything else does. They majorly dropped the ball by not making sure that it can be cooled as effectively as possible. Leaving space to be filled by more TIM is NOT a good idea. It can't have anything but a negative impact on cooling the HBM.

AIB partners have taken note of this and have come to the same conclusion I have. It's a fucking nightmare and I don't blame them for not wanting to touch it with a 10' pole. It disgusting. Just thinking about it makes my head hurt and my stomach churn. I was actually considering buying one until I found this out.

Though, like I said, I probably would still buy one if I could know 100% for sure ahead of time it wasn't one with the GPU die/HBM height mismatch. I'm most certainly not having that. And I don't know why anyone would. Given the fact they all aren't like that. And the ones that are don't need to be. AMD could have made sure of that. But they're obviously a bunch of careless, penny-pinching, asshole idiots...frankly. "Let's make them all different so we can't provide the customers with a consistent product, and piss off our AIB partners by making their jobs way more difficult than necessary". Really? FU AMD!
AssimilatorExcept that memory chips are manufactured to an exact size standard to ensure they can be swapped out. To ensure that this exact issue with fitment isn't an issue. Which in NVIDIA's case it wasn't.

AMD honestly has no excuse for this fucking debacle. We're talking about massive, hot chips, in which a difference in tolerances of under a millimetre make a huge difference to the noise/thermal profile. And you can bet that Gigabyte doesn't have the time or inclination to design 3 different coolers for the potential Vega variants, then have to manually bin each Vega chip they receive to ensure it fits. They're a company out to make money, having to bugger around with Vega chip variances cuts into their profit, so they're going to do the logical thing and just concentrate on NVIDIA cards which are much less painful and hence more profitable to deal with.

If AMD wants to ship its GPUs, it's gotta make it easy for partners to design custom cards based on them, and right now it's doing exactly the opposite. That is their fault, nobody else's, and they need to get their damn act together.
People forgot what the original source that started it all said?
Toms Hardware FranceSo we did our tests with the ungraded version of the GPU, on which the HBM2 RAM chips are 40 μm shorter than the GPU. Fortunately, we have not experienced any cooling problems with these HBM2 chips (which give off a lot of heat), as long as the correct thermal paste is used correctly to fill the 40 μm (high viscosity paste).
As usual they just read headlines.
Posted on Reply
#28
Vya Domus
AssimilatorExcept that memory chips are manufactured to an exact size standard to ensure they can be swapped out. To ensure that this exact issue with fitment isn't an issue. Which in NVIDIA's case it wasn't.

It doesn't matter it was the right size or whatever , I simply gave that example of them using different suppliers for components. Funny how when it comes to Nvidia , everything has an explanation and somehow ends up being perfectly fine but AMD definitely fucks up. :rolleyes:
AssimilatorAMD honestly has no excuse for this fucking debacle. We're talking about massive, hot chips, in which a difference in tolerances of under a millimetre make a huge difference to the noise/thermal profile.
Post whatever piece of evidence you found that confirms this , about making "a huge difference to the noise/thermal profile". I am truly astonished how fast this FUD crap spreads and how quickly everyone gobbles it up. Read the post above this one.
Posted on Reply
#29
chief-gunney
Casecutter- Epoxy resin made in Taiwan RX Vega 64 Samsung HBM2 = Apple iMac
- Resinless made in South Korea RX Vega 56 SK Hynix HBM2 = Actual first/new Interposer supplier and I believe they dropped this company.
- Resinless made in Taiwan for engineering boards (probably SK Hynix HBM2) = Reference production cards

As the to HBM2 height not sure if it both Samsung/Hynix, but yes there are picture on the web like guru3d.com and that Asus RX Vega 64. Not sure what and which models had the issue and if what that really did to heat control of the HBM2.
I got a Sapphire Vega 56, resin less from Korea that has Samsung memory ( haven't heard of anyone getting Hynix memory). HBM runs at the same temp as the GPU and overclocks to 1100 like many others. So far in my testing, the GPU boosts to 1660 mhz and others have done better, but I've got an EK water block on it and that's what I think you need to enjoy Vega cos it's hot as hell. Any type of Air cooling can't tame this card and give better performance than the reference IMO.
Posted on Reply
#30
SPLWF
Franzen4RealOut of curiosity, what are they doing different between the AMD and NVidia cards? I think their AORUS GTX line is pretty nice.
I really like the Aorus GTX line as well, wanted a Vega56 version of it to match my current motherboard. All can only hope....
Posted on Reply
#33
TheinsanegamerN
Vya DomusSounds 100% like a supposition to me , there's no real evidence that there is some sort of problem with cooling caused by the different package assembly.

As if AMD with Vega were the first ones to use multiple suppliers for components such as memory or by having multiple package versions. To me this sounds like a ridiculous level of nitpicking. Nvidia quietly changed the supplier for memory modules on their 1070s after every reviewer got the ones with Samsung chips that overclocked far better and it pretty much went under the radar , a few angry customers here and there but overall it's chill . The good ol' double standard.

But sure call them penny-pinching assholes , whatever floats your boat or makes you more comfortable with your next purchases , we all do it I suppose.
You seem to be projection quite hard yourself.

On previous cards, this was a non issue, as RAM was chosen by the vendor, thus they would have physical control of which cards get which components. The RAM also didnt NEED to be cooled like HBM does. if an AIB wanted to make a card with cooled memory, they could specify which type of memory was used on that card irrespective of the manufacturer of the GPU.

With VEGA, the vendors have 0 control on which memory/package they get, and this causes a serious issue when it comes to mass production of PCBs and heatsinks since the memory has inconsistent height, which you would know if you RTFA.
Posted on Reply
#34
Casecutter
chief-gunneyI got a Sapphire Vega 56, resin less from Korea that has Samsung memory ( haven't heard of anyone getting Hynix memory).
That's interesting because that's what Tom's is saying and showing... I would've thought Samsung is for Apple as I thought I heard they are seeing consistent delivery and seen as offering the best part consistency.
Posted on Reply
#35
Kaotik
CasecutterThat's interesting because that's what Tom's is saying and showing... I would've thought Samsung is for Apple as I thought I heard they are seeing consistent delivery and seen as offering the best part consistency.
Samsung is for everyone at this point, Hynix will be used at some point in some chips when Hynix gets their act together
Posted on Reply
#36
Vya Domus
TheinsanegamerNthis causes a serious issue when it comes to mass production of PCBs
Just like in my previous comment , post any information you find that confirms this is indeed an issue manufacturers have at this moment , such as Gigabyte in this case.

Those articles contain no such proof , just suppositions. So get out of here with your RTFA bs , in fact I advise you to take a better look yourself.

There are no brakes on the FUD train.
Posted on Reply
Add your own comment
Nov 18th, 2024 18:49 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts