NVIDIA GeForce RTX 50-series Blackwell Availability

Causation is reversed, they have low margins because they compete with each other.

It's generally the case for low end hardwares, and that's why most AIB partners are more focused on high end cards, because they are more likely to have higher margins.
For example, on lower end cards such as 3060, the variation of prices between models is smaller, generally something like 10%. However, on a higher end cards such as 4080 or even 4090, you'll see some crazy OC models which are maybe 20% more expensive than other "simple" models. The reason for that is generally when people are buying lower end products they tend to choose the cheaper ones, while when buying higher end products they are more willing to pay more for the bell and whistles.

So there's really no reason for NVIDIA to artificially restrict supply of higher end cards because they are afraid of AIB partners undercutting each other, because they don't want to do that with higher end products. This is especially true this time, because in the past it could be argued that it takes time for existing inventories of last generation products to get cleared, but this time it's already done (it's hard to find any new 4080 or 4090 on the market already).

I'm not saying the supply is not bad. The retailer I mentioned in the first post, only had 28 5090 on the first day (which got sold out, of course), then it's nothing to this day. They sold way more 5080 (419 according to them) but also nothing later. But to say this is because of artificial restriction, I think we need more evidence (right now there's apparently none, only "because it's possible"). There are more other possiblities, such as the rumors last year about some production problems of AI chips eating more wafer allocation, or they really need to fill the AI chip demands so consumer GPU got sacrificed, or maybe it's not the chip but the assemblies of the cards, etc.
 
I mean I can get with this but there is no reason a major GPU launch has single digit unit counts at Microcenter. Nobody is expecting 100% of demand to be fulfilled within the first week but this is a significant step back from the supply situation with the 40 series launch and the 40 super launch.

Why are single Microcenter retail locations special? There's seems to be an inherent bias by framing things that way?

I feel a lot of enthuasists have skewed perceptions. I know from the enthuasist perspective as you frame it it's a major GPU launch. But guess what it is is also? It's going to the lowest selling SKUs over this generation. It's also the launch to the lowest market by volume as well (DIY retail) for those parts. And if we're going to use brick and mortar locations then the likely the lowest channel by volume as well.

As for the 40 series comparisons you should factor in the 40 series was a staggered launch. I'm betting 50 series total shipments were higher than 40 series shipments at launch. And yes there is a reason I framed that way because 40 series launch shipments were the 4090 only, the 4080 launched a month later.

Circling back a bit in terms of enthuasist slant and comparisons, comparing these retail graphics cards to things like the Switch 2 or iphone and etc. are not very analogous as those are really assembled devices. Those devices are really more akin to say OEM prebuilts using the graphics cards. Now Apple and Nintendo only sell the end devices but if you look at something like Qualcomm and Snapdragon SoCs actual phone availability with the high end Snapdragon is something like 6 months after the SoC is announced. The Snapdragon Gen 8 for instance was announced in October last year, I think the first phone with it is scheduled to be available late Feb this year.

I mean again hypothetically to go back to something I mentioned about perspective let's say if Nvidia stock piled for 4 months and that could meet all launch demand is that really a "better" solution? This means in this hypothetical scenario anyone who does manage to buy one now would have been pushed back artificially 4 months. I'll be frank I really hate this perspective, as it's basically saying because I can't enjoy this right now no one else should either until I can. I'm not saying this is your perspective, but this is essentially what stockpiling really is.
 
Why are single Microcenter retail locations special? There's seems to be an inherent bias by framing things that way?
People like buying cards at MC because online stores get scalped during launches and it's a lot harder to scalp MC due to their rules.

I mean again hypothetically to go back to something I mentioned about perspective let's say if Nvidia stock piled for 4 months and that could meet all launch demand is that really a "better" solution?
They don't need to meet all demand, but yeah if they had pushed the launch out a month (and ideally continued making 40 series stuff until we had mid range 50 series products) that would be ideal imo. It just sucks because I'm helping a few guys build and they literally cannot buy graphics cards rn outside of 3060s and 4060s. The 5070 probably isn't going to be much better than a 4070 Super so I'd just recommend they get a 4070S but that's out of stock and the 5070 doesn't exist yet/probably won't be in stock.
 
People like buying cards at MC because online stores get scalped during launches and it's a lot harder to scalp MC due to their rules.

The planet earth is a rather large place. Microcenter has 20 locations.

If we take the 100 or so cards that each Microcenter location had on launch day and multiply that by the number of worldwide computer retailers that amounts to hundreds of thousands of cards.

The world is bigger than the US.
 
I’ve moved this thread to the industry section since it’s primarily a business discussion and not a product or architecture discussion, and tidied it up a little bit.

I also think the thread has run its course unless there’s some new factual information to discuss. It seems to have degenerated into a low-effort tit for tat, which we can probably avoid now.
 
It's fascinating that you could cut 5/6 of the load bearing conductors in a 12V-2x6 (H++) cable and you wouldn't know anything is wrong until you smelled the melting plastic. On a 4090/5090 that is. A 3090Ti would not function if that happened.

der8auer said:
Someone on my German video posted:Nvidia RTX50 with MFG - Multi Flame Generation
:devilish: :devilish:
 
We have no idea what exactly is happening in Der8auer's video.
I can think of a hundred ways to burn a GPU. Doesn't mean that any of them are an actual danger to the cards out there.


The ends are fused, there is no need for any current balancing.
The ends are only fused on the PCB side of the connectors. Not on the wire side. So you have 2 connectors in addition to the wire in each parallel strand. Just do the math what happens when the connectors vary between <1 and 20 mOhm of resistance each, but the wire only adds another fixed 5 mOhm. Remember that the power dissipation on parallel resistors is inverse proportional to their individual resistance, so the lowest resistance path gets the hottest. Total losses are still reasonable for both average and worst case distribution, but it's easily enough to go beyond the rating for the individual wire or pin. If my math isn't totally wrong Der8auer had a minimum of 70W of power dissipation on the hottest wire due to simply unbalanced currents.

Fun fact, the WireView toy of his own brand actually shares that design fault of directly bridging the connector on the PCB side, while also failing to detect either a single wire taking all the load, nor detecting a single pin overheating.
 
Last edited:
So B200 is sold out for all of 2025 and demand for B200 is also lower than expected. I guess both can be true, maybe.

The comment about 40 series being retired is also curious. What does that have to do with B200 demand?
 
Last edited:
Another thing is that generally the lead time of an advanced process node is quite long (can be three months or more). I really don't think it's possible to "switch" the chips just because they somehow found out B200 demand is not high enough.
 
How is B200 not on high demand with all those countries, corporations and hyperscalers spending hundreds of billions on AI? I mean what are they spending it on? H200?

None of the financial institutions and market watchers that closely follow NVIDIA has even hinted that demand of B200 has slowed down. The only thing mentioned is the relative slow ramp of B200 shipping, but that's nothing to worry about considering H100 and H200 suffered the same thing.

NVIDIA's stock even rebounded fast from the DeepSeek dip, and they are back to their usual 3.3 trillion market cap valuation like nothing ever happened. If B200 is not in high demand you can bet your ass their valuation would tank even harder than DeepSeek.
 
So B200 is sold out for all of 2025 and demand for B200 is also lower than expected. I guess both can be true, maybe.

The comment about 40 series being retired is also curious. What does that have to do with B200 demand?
I guess if you follow the guy's logic theoretically the 'unused yield' of the B200 could've also been used for making more 40 series. As that was the same process node and all.

But sadly they apparently have no recourse but to make as many 50 series as they can. And this was decided.. when, supposedly?

I mostly enjoy that this is a direct contradiction of the going conspiracy theory about Nvidia's impending altogether abandonment of gamers for the more sunny uplands of highly profitable AI chips.
Not to mention HUB's awesome stealth price increase thru artificial exclusivity theory.

But really, those make only slightly less sense than this narrative.
 
I'm just going repeat something I posted earlier in that I feel people are just reading too much and overthinking the current supply issue.

It's Q1 the slowest time of year for retail shipments and purchases typically. This is harder to see in a market the size of the US but the issue is more apparent in something like Canada. A lot of of PC component stock is rather low here in January with much less sales (and higher prices by extension) than the Q4 sales season, it's that cycle every year essentially, until things start ramping again into Q3/Q4 and we repeat.

There's also nothing really all that secret about this. If you look at consumer electronics companies in general you can look at their reported financial statements there's always a cycle with Q3/Q4 peaks and Q1/Q2 dips in revenue.

Also not to toot my own horn here but as I mentioned earlier also beaware of how everyone in the supply chain has an agenda in terms of messaging. Don't think those retail channel "leaks" have no self interest as that videocardz article mentions they are doing their own let's just say "supply management."
 
It's Q1 the slowest time of year for retail shipments and purchases typically.
40 Super cards released almost exactly a year ago and there wasn't much of this at all, supply was fairly abundant and after maybe a week or so you could just buy the cards for MSRP. So I don't think this is a seasonal thing.
 
Back
Top