AMD board partner says Radeon RX 9070 GPU supply will stabilize after April

midian182

Posts: 10,638   +142
Staff member
In brief: Consumers desperate to find a Radeon RX 9070 or RX 9070 XT at retailers at MSRP have been offered a potential timeline from a third-party manufacturer. Chinese AIB partner Yeston wrote that while the situation is currently "unstable," the supply will stabilize after April.

Yeston's RX 9070 cards are currently only available in China, which is experiencing the same availability problems as the US.

Yeston writes that it is restocking the cards every week, but asks customers who can't grab an RX 9070 XT not to be frustrated as the supply is unstable.

It's likely that the stabilizing supply of RDNA 4 desktop cards Yeston mentions will be experienced globally, not just in China, and that it will happen in May.

Last week, an AMD executive called demand for its RX 9070 and RX 9070 XT graphics cards "really unprecedented." Team Red had promised "wide availability" for its latest cards when they were announced, but that's not the phrase that comes to mind when looking for one at retailers such as Newegg, Amazon, and Best Buy.

AMD had said that Radeon RX 9070 and RX 9070 XT cards running at default clocks and without factory overclocking would be available at their respective $549 and $599 MSRPs, but many retailers added up to $130 on top of these prices for some non-overclocked models.

Interestingly, a claim posted to Board Channels – the forums used by AMD, Nvidia, and Intel board partners – earlier today claimed that AMD is using a two-tier system for its GPU allocation. The core partners, consisting of Asus, XFX, PowerColor, Sapphire, and Vastarmor, are said to have received thousands of units in each batch.

The second-tier partners, a group that includes Yeston, are said to be allocated fewer GPUs. Even big names such as Acer, Gigabyte, and Asrock are considered non-core AMD partners, according to the post.

VideoCardz notes that AIB partners in the first group tend to offer four to ten Radeon RX 9070-series cards, while those in the second tier have between two and six models.

Permalink to story:

 
The situation here is dire, at 800 USD XT's are massively overpriced and 5090's are over 4000. We aren't even an afterthought to these firms in this 'AI' race.
 
but many retailers added up to $130 on top of these prices for some non-overclocked models.

This is factually incorrect as the cheapest 9070XT in Germany is +EUR 141 over MSRP. German MSRP for the 9070XT was EUR 689.

See for yourselves https://geizhals.eu/?fs=9070 xt

I imagine in SE Asia, S Asia (Indo-Paki) etc this would be much worse, perhaps +EUR 200 over MSRP or much more.

We are back to 2020 only this time the Leatherman is doing the "mining".
 
the supply will stabilize after April
If "after April" means the period starting May 1, 2025 and continuing to the end of time, then, well, duh. Although I suppose it could still be wrong if supply never stabilizes.

But if it's meant to imply "in May" I have to wonder, given the lead times involved, how much could really change between now and then. Decisions affecting output for this period were made weeks or months ago. It does not seem likely the rate of production could drastically increase unless there was say a big yield problem that was fixed a couple months ago. But I haven't heard any rumblings to that effect.

Maybe they have an internal estimate of how much backed up demand there is from buyers who were ready to buy ASAP at launch, and they believe that initial demand is close to being satisfied. But if so that's another insight I haven't heard anywhere else.

Anyway I'm very curious why they believe this / believe it could be possible.
 
AMD catching up by May is plausible. While there is a time lag between production changes and products on shelves, 1) Nvidia's paper launch in January would have tipped them off that demand could be higher and 2) AMD is finally catching up on 9800x3D demand potentially freeing up additional production capacity for GPUs.

The good news is AMD is financially incentivized to make GPUs as fast as their TSMC capacity allows so it really is a matter of when, not if, AMD's GPUs will be readily available near MSRP. The real question is Nvidia. They could respond relatively quickly to protect market/mindshare or they could continue the trickle of GPUs for another year.
 
Nvidia makes more geforce GPUs then it does AI GPUs. Everyone is limited by TSMC.
ROFL... no... no they don't. If you talk in terms of waffers, they really don't.

As for TSMC, it is not the case because they don't use CoWoS for gaming GPUs and TSMC 4n is now becoming a more mature node.
 
ROFL... no... no they don't. If you talk in terms of waffers, they really don't.
I said the number of GPUs. I didnt say anything about "waffers", as I'm not lookign for snack items while drunk, but if you're talking about TSMC wafers, no, I didnt mention those either.

The geforce division brought in $17 billion last year. The datacenter division $40 bil. Now, given the size and complexity of those datacenter chips, do you REALLY think nvidia is selling those chips at only 2x the proce of a geforce chip, on average?

ROFL....no....not a chance. And unless those giant chips are only being sold at 2x the price, numerically the geforce sales would have to be much higher then the AI chip sales.
As for TSMC, it is not the case because they don't use CoWoS for gaming GPUs and TSMC 4n is now becoming a more mature node.
This is what we call a red herring argument. Neither the lack of CoWoS or being a "mature node" mean there is a surplus of the node. The continued high prices and lack of supply would rather indicate the complete opposite.
 
This article is really reading a lot into a poorly written tweet. Yeston could just as easily mean that their supply will become more predictable or regular in April, not that they'll have any more per month.
 
I said the number of GPUs. I didnt say anything about "waffers", as I'm not lookign for snack items while drunk, but if you're talking about TSMC wafers, no, I didnt mention those either.

The geforce division brought in $17 billion last year. The datacenter division $40 bil. Now, given the size and complexity of those datacenter chips, do you REALLY think nvidia is selling those chips at only 2x the proce of a geforce chip, on average?

ROFL....no....not a chance. And unless those giant chips are only being sold at 2x the price, numerically the geforce sales would have to be much higher then the AI chip sales.

This is what we call a red herring argument. Neither the lack of CoWoS or being a "mature node" mean there is a surplus of the node. The continued high prices and lack of supply would rather indicate the complete opposite.
Those revenue #s are wrong. Data Center to Gaming was like 5-1 in Jan 2024, Now it's like 10 to 1. It's not remotely close.
Yes, we don't matter, or really we just matter much less at the moment. AMD also making much more in data center revenue. Lisa said it herself. "So, we really are a data center-first company." Of course they are going to focus on the AI boom. They're raking it in.
I do like the 9070 series though. Good move to regain market share. Then go back to high end next gen.
 
Last edited:
Back