Wednesday, January 16th 2019

A Sprinkle of Salt: AMD Radeon VII Reported to Only be Available in Reference Design, no Custom Treatment

A report via Tom's Hardware.de says that AMD's plans for the upcoming Radeon VII are somewhat one-dimensional, in that only reference designs will be available for this particular rendition of the Vega architecture. And this doesn't mean"initial availability" only on reference cards, like NVIDIA has been doing with their Founder's editions; the report claims that at no point in time will there actually be a custom-designed Radeon VII. The quantity of Radeon VII GPUs will apparently be "strictly limited" come launch - a likely result of the decision to make use of TSMC's 7 nm process, which will have to serve not only AMD's Ryzen 3000 and Epyc CPUs when those are actually launched, but all of TSMC's other clients.

This is in contrast with AMD CEO Lisa Su's words during her CES keynote, who said that Radeon VII would be available from "several leading add-in board partners plan to offer the cards". According to a Tom's Hardware.de Taiwanese source, "You cannot leak anything that does not exist" in regards to third-party designs. And another Chinese source said "the quantity of Radeon VII is strictly limited… not sure if AMD wants to open AIB to have an own design later".
The saltiness is in the title for a purpose: we'd be very surprised with a decision such as this from AMD's part. Low availability to partners is better than no availability at all for a number of reasons. Let's not forget the damage it would do to AMD's ecosystem to only release a high-performance product - the one that AMD buyers have been waiting for since the original Vega) under their own branding, closing partners out of the profits they'd make on custom designs. It just doesn't strike us as a sensible business decision.
Source: Tom's Hardware.de
Add your own comment

40 Comments on A Sprinkle of Salt: AMD Radeon VII Reported to Only be Available in Reference Design, no Custom Treatment

#28
EarthDog
If you don't believe the horse's mouth on where they want to market/how to define it... who do you believe? o_O
Posted on Reply
#29
notb
LFaWolfAIB partners, not SKUs. Here you go, the usual suspects, the big 3, and the AMD exclusive partners of Sapphire, XFX, and PowerColor.
Nope. One card per manufacturer. It's literally like if they were forced to do it by some agreements with AMD.
The 4 cards mentioned are all custom Vega 64 designs I could find.
Everything else is just the reference model with stickers.
Posted on Reply
#30
LFaWolf
EarthDogIf you don't believe the horse's mouth on where they want to market/how to define it... who do you believe? o_O
I know right. When I saw the CES reveal of the Radeon VII I must admit I was bit excited. $700 is not too bad of a price to compete with RTX 2080. Then I did some research and it is the same chip as the Radeon Instinct, a workstation card. Workstation cards are expensive to make with the HMB2 memory. There are also reports that AMD is only making 5000 of the Radeon VII. I don't think they make much money, if any, on these. The worse part is if nVidia cuts price on the RTX 2080 (but I doubt), they won't be able to match.
www.techpowerup.com/gpu-specs/radeon-vii.c3358
www.techpowerup.com/gpu-specs/radeon-instinct-mi50.c3335
notbNope. One card per manufacturer. It's literally like if they were forced to do it by some agreements with AMD.
The 4 cards mentioned are all custom Vega 64 designs I could find.
Everything else is just the reference model with stickers.
Nope, the person you quoted did not specifically say custom cards, just AIB cards. Reference design cards with AIB brand are still AIB cards.
Posted on Reply
#31
EarthDog
LFaWolfReference design cards with AIB brand are still AIB cards.
Are they? I mean, by sticker, sure. But all the underpinnings are reference AMD. I don't call a sticker swap an AIB card. It needs to have different cooling than reference at minimum (more typically beefier VRMs).
Posted on Reply
#32
LFaWolf
EarthDogAre they? I mean, by sticker, sure. But all the underpinnings are reference AMD. I don't call a sticker swap an AIB card. It needs to have different cooling than reference at minimum (more typically beefier VRMs).
Then it is semantics, a case of different person/different definition. I would call them AIB cards.
Posted on Reply
#33
EarthDog
Technically they are AIB, you are correct. But there is literally one difference though...a sticker.

People arent up in arms for a lack of those cards, lol!! People are concerned over the lack of custom cooled/powered models... a more true representation of an AIB card. You can see the other person thinks in a similar manner. :)
Posted on Reply
#34
Foobario
AMD is doing the right thing by selling this directly to consumers first. By keeping the $100 plus profit margin that the AIB sellers would require they are making these cards profitable from the get go. By foregoing the AIB sales channel they may be limiting the appeal of this card, but they are giving loyal AMD gamers a legitimate 4K card that will keep them in the AMD camp.

If forecasts for a material drop in RAM pricing in the first half of 2019 are correct it seems plausible to expect AMD to have more pricing leverage in a month or so. At that point they can drop the direct to consumer sales price and AIBs can offer their versions of the card closer to the $699 price point.

I doubt the MI60 is selling at extremely high volumes, yet. As it gains more traction it will allow for more failed dies to be diverted to Radeon VII. At the same time the overall revenue per wafer will go up as more MI60s are yielded allowing for more aggressive pricing on Radeon VII. Combine the better pricing from wafer allocation with lower HBM prices and a $599 or less Radeon VII is a realistic possibility by the end of February.

Considering the excessively large die sizes on the 2000 series Nvidia cards and Nvidia desperate to maintain margins while their stock price is in a free fall it seems unlikely they will have much pricing leeway on their RTX cards.

The rumors of 1100 series cards where they pull out the useless tensor cores to reduce the die size seems very plausible. However, I suspect these 1100 series were a recently developed plan B after Nvidia discovered even their brainwashed followers have enough sense to thumb their noses at cards using tech that is still a year or more away from primetime. Perhaps this realization is the reason Jensen Huang was acting more unhinged than usual at CES.
Posted on Reply
#35
Casecutter
Yes I think the initial release will only AMD Branded reference exclusive from AMD directly. Mostly to make sure there isn't the gouging from retail and makes it fair that everyone has an equal chance that first day. The last Vega was a free-for-all and AMD doesn't want that again.

As for AIB's... nobody consider/think existing Vega 64 design don't easily "port-over", in all probability I might even say the interposer are "pin-for-pin" drop-in to the existing PCB that Vega 64 had. MI50 300 TDP while I thought Vega 7 isn't suppose to be that high and Vega 64 was what a 295 TDP, so the power section might need nothing. AIB's have existing PCB and coolers that should not have an issue flipping from their existing Vega 64 customs I can't see AMD not have then build the next batch. I think this was rushed and kept under the radar so most AIB's weren't wrote in, but I don't believe AMD want this to be a Founders Edition kind of thing long term.
Posted on Reply
#36
Ruru
S.T.A.R.S.
eidairaman1Did you have a stroke?
His text is similar in Finnish, can't make anything out of it.
Posted on Reply
#37
RCoon
As a general warning for all members present, before you post something that's basis is utter shite, ask yourself "do I have a source for this information?". If the answer is no, what you're posting is FUD, and against our guidelines.
Posted on Reply
#38
kanecvr
medi01What price war?

And while we are at it, how did price war of 1050Ti vs RX 570 (which is what, 1.5 times faster?) go?


Rumor had it that AMD prepared that mostly for PR purposes, just to be able to tick off "we are present in that perf bracket too" and other than that product wasn't that viable from commercial perspective (with HBM2 alone estimated to cost $200-300 half a year ago).
The RX 570 is 150 to 190% faster then the 1050ti. And like you're implying - due to consumer ignorance and nvidia's strong brand name the inferior 1050ti sold exponentially more.
Posted on Reply
#39
Casecutter
kanecvrdue to consumer ignorance and nvidia's strong brand name the inferior 1050ti sold exponentially more.
Regrettably, the GTX 1050Ti as much as I totally agree with what you say, was the only thing people could find and at not inflated prices while the mining boon was on. The RX 570 was the entrant mining darling and was nowhere to be had. So I'm sure Steam data would've expected a lot of GTX 1050Ti's showing up... There wasn't anything else worth buying even RX 560, while slower had dried up as AMD wasn't about to allocate very many wafer starts at GloFlo for anything but Polaris 20. There's huge amount of the Steam folks who where stuck for whatever reason at the time, who had no choice to ante-up $230-250 for GTX 1050Ti's. Those are now are having a hard time justifying an upgrade, even if at half the price with twice the improvement.
Posted on Reply
Add your own comment
Jan 22nd, 2025 10:47 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts