5090 Production Halted

Dan_D said:
Not true. I think within certain brands, the flagship is what sets the tone for the rest of the lineup. People might correctly, or incorrectly assume that because NVIDIA has the fastest GPU's at the top of the stack, that holds true for the rest of the lineup. When most people think of Chevrolet, it's not the spark, or the Volt they think of. It's the Corvette or at least a high end Silverado or something. Not the Equinox. I think of Ford I think of F-150's and Mustangs. Not the crappy Mach-E or other BS they make.

In any event, whether its entirely relevant or not, there are people who don't look at mid-range or lower end offerings for their needs. Price be (mostly) dammed, I only ever really look at the top two GPU's in the NVIDIA or AMD lineup. I couldn't care less about the rest of them. The only time I do, is if I suddenly need one for another system or my next laptop or something.
Click to expand...
Your perspective isn't most people's perspective. By your own accord you just buy a top card, price be damned. Most people don't do that. Most people aren't judging by the best card that flat out exists. They're judging by the best card they can afford. That's the mid range or entry level for most people. If you look outside the enthusiast spaces where normies are, they ask questions like, "What's the best card I can get for $300-$400? Should I consider used"?

Nothing wrong with high end cards existing or you getting excited by them. They are awesome. I wish I had the disposable income to justify one. However, they aren't influencing most people's decisions though.
 

“GALAX Prepares GeForce RTX 5090 HOF with Swarovski Crystals​

by AleksandarK Today, 11:12 Discuss (7 Comments)
GALAX has officially teased its GeForce RTX 5090 Hall of Fame Edition GPU, featuring unique details with Swarovski crystals. What the company originally showcased at Computex 2026 now appears to be a product scheduled for launch on October 22. The GeForce RTX 5090 HOF maintains the distinctive white aesthetic that GALAX GPUs are known for. However, this time, the design is enhanced with Swarovski crystals embedded across the GPU shroud, adding a nuanced touch. This should help the GPU stand out in all-white builds, especially when RGB lighting hits the crystals, reflecting and bursting in multiple directions.

Another interesting aspect is that the card is expected to feature dual 12V-2×6 power connectors, providing significant power headroom. GALAX has already offered its flagship GeForce RTX 5090D HOF Edition GPUs with dual 12V-2x6 power connectors, but this will be the first time an overclocking-focused graphics card receives such a luxurious finish with crystals. Initially, the GPU's appearance at Computex 2026 was thought to be a preview for the GeForce RTX 6090. However, it is now clear that the item will be more of a collectible, as GeForce RTX 5090 GPUs have become extremely expensive on their own.“
 
erek said:

“GALAX Prepares GeForce RTX 5090 HOF with Swarovski Crystals​

by AleksandarK Today, 11:12 Discuss (7 Comments)
GALAX has officially teased its GeForce RTX 5090 Hall of Fame Edition GPU, featuring unique details with Swarovski crystals. What the company originally showcased at Computex 2026 now appears to be a product scheduled for launch on October 22. The GeForce RTX 5090 HOF maintains the distinctive white aesthetic that GALAX GPUs are known for. However, this time, the design is enhanced with Swarovski crystals embedded across the GPU shroud, adding a nuanced touch. This should help the GPU stand out in all-white builds, especially when RGB lighting hits the crystals, reflecting and bursting in multiple directions.

Another interesting aspect is that the card is expected to feature dual 12V-2×6 power connectors, providing significant power headroom. GALAX has already offered its flagship GeForce RTX 5090D HOF Edition GPUs with dual 12V-2x6 power connectors, but this will be the first time an overclocking-focused graphics card receives such a luxurious finish with crystals. Initially, the GPU's appearance at Computex 2026 was thought to be a preview for the GeForce RTX 6090. However, it is now clear that the item will be more of a collectible, as GeForce RTX 5090 GPUs have become extremely expensive on their own.“
Click to expand...
I guess when AICs are getting so few dies, they need to figure out a way to justify selling them for $10k each.
 
djstarfox said:
I think Intel left the GPU market at the worst possible time.... NVIDIA price gouging everyone... Means even a 3rd place company, selling at the right price becomes very competitive. I almost bought a B580 until I found a lightly-used 5070.
Click to expand...
TSMC charges more for their GPU nodes now than ever before.
Unless Intel manages to get their GPU’s made with their own Fabs they would have had to raise their prices too high to remain competitive.
 
I really hope Intel gets back into the game in a serious fashion. Nvidia will eventually abandon consumer or greatly reduce it's output for consumer in favor of enterprise. AMD once it gets good translation layers for CUDA will also start moving away from consumer. Silicon is just too hard to come by and AI is looking like the bubble is more sturdy than people had hoped.
 
OFaceSIG said:
I really hope Intel gets back into the game in a serious fashion. Nvidia will eventually abandon consumer or greatly reduce it's output for consumer in favor of enterprise. AMD once it gets good translation layers for CUDA will also start moving away from consumer. Silicon is just too hard to come by and AI is looking like the bubble is more sturdy than people had hoped.
Click to expand...
If intel compute foundry get going and good enough for GPUs, the memory bottleneck will still be the same for them, it is way more memory the issue than die silicon, the AI industry do not consume that much of the world compute wafer capacity versus say the 900 millions smarthphone being sold in 2026, it was just a tiny bit, a giant gigawatt is not even 900,000 GPUs or 0.9 million gpu in it anymore, but come with giant networking and memory needs.

Compute wafer is still vastly spent on consumer because of phones, one way to look at it an 3-4 millions / 120kw Nvidia NVL72 rack consumer in compute wafer (the TSMC 4NX) only the equivalent of 1,180 Iphone SOC, virtually nothing, but 20-25,000 iPhones in memory wafer, i.e. the impact on memory is about 10-20 time the one on compute, that why CPU price did not suddenly jump that much, right now they are not able to be limited by TSMC capacity for AI, packaging-networking-memory limit how much they can deploy to very small quantity of compute all things considered.

And soon high power giant reticle need/giant dies datacenter wafer node will probably diverge from consumer one (and packaging need between the 2 will continue to be different would be a reasonable guess), if the others bottlenecks ever clear out to make compute wafer something AMD-Nvidia ever has to do between dgpu and datacenter gpu, it will stop to be possible to make.

SpaceX achieving the impossible and make memory in that giga-factory sized foundry could be more a game changer or just CXMT-others scaling up.

If GDDR or large bandwith lpddr supply become good enough for intel, it should for AMD-Nvidia has well, only if AI accelerator/training get different enough from gaming-blender-Adobe use case for AMD-Nvidia and they do not make a division for those, maybe some of Apple-Intel-Qualcomm-Samsung-Mediatek-Sony-Nintendo and others could become those making gaming GPUs instead, Intel is particularly well suited to win this if they want.

I think a gaming GPU and a generative AI silicon will become quite similar for a large part of the die, a bit of compute ground truth used in input of generative models and AMD-Nvidia will be the best at making them and the one doing so is more likely than Intel being the main high end gaming providers of GPU for the sony-nintendo in the near and mid future.
 

“NVIDIA Has January Discontinuation Dates Set for RTX 5090 and RTX 5080 16 GB, Claims Leaker​

by DrUsmanSaleem Today, 18:57 Discuss (3 Comments)
Earlier this month, leakers claimed NVIDIA had stopped allocating GB202 silicon to the GeForce RTX 5090, effectively discontinuing the Blackwell flagship. NVIDIA denied it a few days later, with SVP Jeff Fisher saying the RTX 5090 hasn't been discontinued and will keep shipping. Now, Moore's Law Is Dead (MLID) says the plans are still in place, and the RTX 5080 is part of them too. In the latest Broken Silicon episode, MLID read out a message from what he calls one of his oldest sources, apparently a major retailer, saying NVIDIA has set discontinuation dates for both cards in the first half of January 2027.

According to that source, the last RTX 5090 shipments arrived in September, while the final RTX 5080 batch should land by the end of October and sell through Thanksgiving. We recently reported that Micro Center has already marked its RTX 5090 listings as "Not Carried," which could be further evidence of the RTX 5090 leaving shelves. The retailer from MLID's video also says its system already has a placeholder SKU for an RTX 5080 24 GB, which aligns with BenchLife's earlier report of a Q1 2027 launch.“
 
Sycraft said:
Well, I'm actually inclined to believe nVidia in this case rather than the leakers as it doesn't make a lot of sense for them to make a statement like this if it isn't true. They could just say nothing if they are discontinuing them, they don't even have to confirm it. Wouldn't make sense to flatly deny it and to be lying as there really is no gain.
Click to expand...
"Supply contraints" seems like code for we're using them for higher priced SKUs. Technically true, but more in line with the discontinuing statement.
 
Sycraft said:
Well, I'm actually inclined to believe nVidia in this case rather than the leakers as it doesn't make a lot of sense for them to make a statement like this if it isn't true. They could just say nothing if they are discontinuing them, they don't even have to confirm it. Wouldn't make sense to flatly deny it and to be lying as there really is no gain.
Click to expand...
Just a farce to keep up the image that they still care about gamers.
 
The 4090 has the bw (mem+compute) to run dlss5 with the ai iq stuff... maybe not the precise architecture for it but it's got the power.. and they could driver that shit up so fast. maybe off que for this conversation but got me thinking about it.. again.. they're planning on doing (edit: FULL) dlss5 for 4090/4080 I believe still.. no date set... or so I read. (later in the fall of 2026...)... with the memory thing we might win this one.. they're going software with the capable previous line... this is the weirdest timeline
 
Last edited:
Sycraft said:
I think we will in the long run. We are in the middle of a bubble right now, companies are borrowing insane amounts of money, a large number of the new GPUs haven't even been installed, they are just sitting in warehouses, this is not sustainable. It will chill out and when it does, the consumer market will be more interesting again. This isn't happening tomorrow, maybe not even next year, but this idea that the AI companies ahve that they will just grow to infinity and beyond is not realistic. They aren't making any money at their current prices and they are running out of other people's money to borrow.

Particularly the datacenter builders. Oracle is just one step above junk bond status, and is paying 8% and up, Coreweave is rated junk (and losing a billion a year), banks are getting skittish and starting to look to sell the loans they underwrote. At some point, you run out of other people's money. That doesn't mean it is over and AI goes away, but it means the feeding frenzy dies down.
Click to expand...
But the prices will never ever come back down to what they were. They will be 20-25% higher and will be the new baseline. Because reasons.
 
Eivind68 said:
But the prices will never ever come back down to what they were. They will be 20-25% higher and will be the new baseline. Because reasons.
Click to expand...
No, they won't. Well unless someone else decides to start a price war and that isn't very likely. However I do think there will be more care about the consumer market after the DC boom wears off. They will still almost certainly make more on datacenters than consumers, but the ratio won't be as much as it is now.
 
This card was as low as $8000 just six months ago https://shorturl.at/IMhbC. I remember talking to my spouse and toying with the idea of actually getting one. Yeah, nope...

1792610034473.png
 
Eivind68 said:
But the prices will never ever come back down to what they were. They will be 20-25% higher and will be the new baseline. Because reasons.
Click to expand...
They won't especially if there is no real competition. $1000 x070 class cards will be normal even if the AI bubble pops.
 
You must log in or register to reply here.
Back
Top