Blackwell gpus

[Grim5]

pc gamers be like: lol I don't care I just buy the fastest GPU, screw the next generation of gamers who have to buy overpriced junk due to a monopoly

It's just like the housing market. The boomers ruined the housing market and now we're ruining the gaming market, get on the train boys we're going to end of the line

 

[BananaDunka]

I mentioned it before in another thread a while ago, but that's the wrong way to look at it. If nobody bought the previous high priced cards (or current), then the company would be less inclined to R&D into key areas that genuinely advance gaming and benefit everyone down the line, even if that trickle down effect takes some years.

Just like with cars, if nobody buys the highest end models then the entire range takes a hit or gets abandoned as a flop and things don't innovate as quickly.

If nobody bought the higher end RTX cards then would DLSS and oher innovations that improve RT/PT and general rendering QOL gains such as ray reconstruction and RT denoising be what they are right now? Nope.

People need to be buying the expensive stuff for companies to then put money back into the pool for R&D to happen in a timely manner. People often say that AMD just don't have the R&D budget to keep up with Nvidia's speed of advancement with DLSS and the like, why is that has anyone wondered? It's because AMD gave up on the high end and R&D is 2 generations behind for PC gaming as a result.

In the same measure has Nvidia given up on PC gaming in favour of AI? Probably yes and no combined, because the AI advances directly affect PC gaming, since AI has been the key driver for RTX cards from day 1, again, because of the massive R&D put into advancing the deep learning training models to make the features we use today more efficient.

The gamers wanting mid range priced cards that perform at high end ranges need the people who are buying those high end cards to continue doing that, otherwise you don't get anywhere year after year.

 

[Rroff]

As far as we know based on public information, unless a microchip architect or engineer can interject and correct us, using MCM to make a gaming GPU with two full dies on the same silicon substrate requires around 100TB/s of connection bandwidth between the two dies (50TB/s each way) and right now Nvidia's GB Blackwell GPU only has a 10TB/s connection (5TB/s each way). 10TB/s is enough for generative AI, inference and computational workloads because the workloads are not sensitive to latency and don't need to output full image frames within small frametimes, but gaming is sensitive to this and so requires much more bandwidth to work without issues.

And while I do not understand the exact technical reasons for why its hard to scale up the connection bandwidth between the dies; the dumbed down consumer friendly version is that there is lots of little wires that have to run between the two dies to transport the data and if you want to increase bandwidth you have to lay more wires between the two dies and its very very hard fitting all the wires into that tiny space.

Depends how you approach it - MCM isn't just one approach, but yeah if you brute forced it, it would require hideous amounts of bandwidth and low latency interconnects. One of the problems is noise - high speed interconnects tends to become very susceptible to interference and the longer the wire (trace) is the more that becomes an issue.

 

[humbug]

I mentioned it before in another thread a while ago, but that's the wrong way to look at it. If nobody bought the previous high priced cards (or current), then the company would be less inclined to R&D into key areas that genuinely advance gaming and benefit everyone down the line, even if that trickle down effect takes some years.

Just like with cars, if nobody buys the highest end models then the entire range takes a hit or gets abandoned as a flop and things don't innovate as quickly.

If nobody bought the higher end RTX cards then would DLSS and oher innovations that improve RT/PT and general rendering QOL gains such as ray reconstruction and RT denoising be what they are right now? Nope.

People need to be buying the expensive stuff for companies to then put money back into the pool for R&D to happen in a timely manner. People often say that AMD just don't have the R&D budget to keep up with Nvidia's speed of advancement with DLSS and the like, why is that has anyone wondered? It's because AMD gave up on the high end and R&D is 2 generations behind for PC gaming as a result.

In the same measure has Nvidia given up on PC gaming in favour of AI? Probably yes and no combined, because the AI advances directly affect PC gaming, since AI has been the key driver for RTX cards from day 1, again, because of the massive R&D put into advancing the deep learning training models to make the features we use today more efficient.

The gamers wanting mid range priced cards that perform at high end ranges need the people who are buying those high end cards to continue doing that, otherwise you don't get anywhere year after year.

AMD have competed at the high end for as long as they have been designing GPU's.

HD 4870
HD 5870
HD 6970
HD 7990
R9 290X
RX 6950 XT

All these card's were as good as or better than Nvidia's best, all the while they lost market share.
Perhaps AMD aren't doing enough, if enough is 2X as good as 1/2 the price then look at how that turned out for ATI, they ended up bankrupt.

AMD were at their happiest and had the best mindshare when they were making a small selection of low to mid range GPU's, people talk about wanting AMD to compete but it seems the higher end the more criticism AMD seem to attract, i guess no one cares if the most expensive GPU they make is $300.

 

[BananaDunka]

What I mean is that their current focus isn't on high end, as evidenced with the situation currently happening. They used to be, but they no longer are. AI acceleration is driving everything for both AMD and Nvidia and they don't really care much for the high end gaming segment now because the money is in big tech AI.

The side effect of this is that because AI accelerator cards are so powerful, it just so happens that they run games great too. That's the gist of it.

This doesn't sideline the fact that people need to be buying the high priced models so what is said above is still accurate.

There was a good post about it on reddit a few months ago re: AMD

https://www.reddit.com/r/AMD_Stock/comments/15iyepk/with_rumored_no_high_end_on_consumer_roadmap_is/

AMD are not doing enough, and have not been doing enough for a long time. Which is why they are at least 2 generations behind in tech innovation.

 

[humbug]

What I mean is that their current focus isn't on high end, as evidenced with the situation currently happening. They used to be, but they no longer are. AI acceleration is driving everything for both AMD and Nvidia and they don't really care much for the high end gaming segment now because the money is in big tech AI.

The side effect of this is that because AI accelerator cards are so powerful, it just so happens that they run games great too. That's the gist of it.

This doesn't sideline the fact that people need to be buying the high priced models so what is said above is still accurate.

There was a good post about it on reddit a few months ago re: AMD

https://www.reddit.com/r/AMD_Stock/comments/15iyepk/with_rumored_no_high_end_on_consumer_roadmap_is/

AMD are not doing enough, and have not been doing enough for a long time. Which is why they are at least 2 generations behind in tech innovation.

When people say "they are not doing enough" as if everything that Nvidia is doing is fanatically in reach of a company 10X smaller, with that it can only be that AMD are incompetent or lazy.

There isn't much of an argument there, its a pretty circular opinion, it makes it so that AMD can only ever be the villains and that's where i have a serious problem with it, it conveniently ignores the monopoly.

 

[humbug]

It allows people like Steve Walton to say Nvidia GPU's are worth 30% more than AMD for reason X and then say the monopoly and overpricing is AMD's fault for not trying hard enough.

In fact people like him are part of the problem, but that reasoning allows him to use AMD as a ascape goat. Its why they make such reasoning.

You see if you advocate giving the monopoly more money than the underdog them your routing for the monopoly, and he knows it. These are reasons i question where their shares are.

 

[humbug]

I want AMD to stop chasing Nvidia's high end, i want them to stop spending R&D on making huge GPU's smaller so they can fit it inside a budget people are willing to pay for AMD.

No, scrap all of that, go back to making midrange GPU's that don't need expensive complexity to bring them down to a budget, spend what you save on making RT and upscaling better.

Which it seems they are doing now, good, Nvidia are going to end up charging $2500, $3000, $4000.... for their halo card and bring everything up the pricing tiers to match, let them.

Oh and ignore the people who will say for that reason AMD are not doing enough, they only want cheaper Nvidia GPU's, that's their problem to solve, not yours.

 

[TNA]

I mentioned it before in another thread a while ago, but that's the wrong way to look at it. If nobody bought the previous high priced cards (or current), then the company would be less inclined to R&D into key areas that genuinely advance gaming and benefit everyone down the line, even if that trickle down effect takes some years.

Just like with cars, if nobody buys the highest end models then the entire range takes a hit or gets abandoned as a flop and things don't innovate as quickly.

If nobody bought the higher end RTX cards then would DLSS and oher innovations that improve RT/PT and general rendering QOL gains such as ray reconstruction and RT denoising be what they are right now? Nope.

People need to be buying the expensive stuff for companies to then put money back into the pool for R&D to happen in a timely manner. People often say that AMD just don't have the R&D budget to keep up with Nvidia's speed of advancement with DLSS and the like, why is that has anyone wondered? It's because AMD gave up on the high end and R&D is 2 generations behind for PC gaming as a result.

In the same measure has Nvidia given up on PC gaming in favour of AI? Probably yes and no combined, because the AI advances directly affect PC gaming, since AI has been the key driver for RTX cards from day 1, again, because of the massive R&D put into advancing the deep learning training models to make the features we use today more efficient.

The gamers wanting mid range priced cards that perform at high end ranges need the people who are buying those high end cards to continue doing that, otherwise you don't get anywhere year after year.

You changed your mind and ordering a 5090 for £1999 on release is it?

 

[Calin Banc]

Competition... L O L. What competition when AMD was doing what nVIDIA was doing (including in the CPU market)? If there would be competition then something like r290/x series would happen much often. When prices are similar or worse when you look at performance and features there is no competition, no reason to bother with the lesser product. That's on the undersog, too. No point in washing it clean.

 

[humbug]

Competition... L O L. What competition when AMD was doing what nVIDIA was doing (including in the CPU market)? If there would be competition then something like r290/x series would happen much often. When prices are similar or worse when you look at performance and features there is no competition, no reason to bother with the lesser product. That's on the undersog, too. No point in washing it clean.

I don't really know what that means? What was so special about the 290X that made it the standout generation?
It was good.... i had an R9 290, i also had an HD 6950, that was good, the HD 4870 was also a very good GPU.

Is the 7900 GRE not a good card? i would argue its a better card than the RTX 4070.

 

[Dirk Diggler]

I don't really know what that means? What was so special about the 290X that made it the standout generation?
It was good.... i had an R9 290, i also had an HD 6950, that was good, the HD 4870 was also a very good GPU.

Is the 7900 GRE not a good card? i would argue its a better card than the RTX 4070.

I went for the 7900XT and the equivalent IIRC was the 4080 at £250 more, so the AMD option definitely worked for me.

 

[Quartz]

I want AMD to stop chasing Nvidia's high end, i want them to stop spending R&D on making huge GPU's smaller so they can fit it inside a budget people are willing to pay for AMD.

This is called the console market, which AMD dominate.

 

[TNA]

Name of thread Blackwell GPU's. Content in thread the last page is AMD GPU's and why are people not buying them more wah wah. Underdog please support wah wah. No matter what people won't buy AMD wah wah.

Oh God leave it out. AMD need to do a lot better. They need to address FSR and improve price for performance significantly. Can't be behind in features and just give small cuts in price here or there and hope to improve their market share in a meaningful way.

Either get ahead technically or offer good price for performance. 50 dollars here 100 dollars there just won't cut it when Nvidia have priced their cards so stupidly to begin with.

You have @ICDP saying my brand new 4070 Ti was not a good deal at £575. So if that isn't a good deal and a bad one then everything AMD has on offer is complete tripe too as far as I am concerned.

This is all coming from someone who has purchased lots of AMD GPU's and CPU's. What do you think people who hardly ever purchase them think? That's how far behind AMD is. They need to up their game.

 

[Grim5]

at GTC Micron announced that next Gen GPUs using gddr7 would be for sale by the end of this year. They did not say if it's Nvidia, Intel or AMD GPU

Next-gen GPUs likely arriving in late 2024 with GDDR7 memory — Samsung and SK hynix chips showed chips at GTC

GDDR7 16Gb chips in production, 24Gb coming next year.

www.tomshardware.com

 

[humbug]

at GTC Micron announced that next Gen GPUs using gddr7 would be for sale by the end of this year. They did not say if it's Nvidia, Intel or AMD GPU

Next-gen GPUs likely arriving in late 2024 with GDDR7 memory — Samsung and SK hynix chips showed chips at GTC

GDDR7 16Gb chips in production, 24Gb coming next year.

www.tomshardware.com

I would think all of them.

 

[Rroff]

at GTC Micron announced that next Gen GPUs using gddr7 would be for sale by the end of this year. They did not say if it's Nvidia, Intel or AMD GPU

Next-gen GPUs likely arriving in late 2024 with GDDR7 memory — Samsung and SK hynix chips showed chips at GTC

GDDR7 16Gb chips in production, 24Gb coming next year.

www.tomshardware.com

Given semi-conductor production timelines and what details are known of the nVidia order book end of the year (September onwards) for mass volume is highly likely, though doesn't preclude a paper launch or limited availability earlier - though Apple has most of the early production tied up for bleeding edge nodes through to the start of summer.

 

[Dicehunter]

pc gamers be like: lol I don't care I just buy the fastest GPU, screw the next generation of gamers who have to buy overpriced junk due to a monopoly

It's just like the housing market. The boomers ruined the housing market and now we're ruining the gaming market, get on the train boys we're going to end of the line

As someone also mentioned on Discord the other night, I can see Nvidia eventually down the line exiting the game GPU market IF AI keeps going.

It makes no sense to sell a £1000 GPU here and there when you can instead prioritize that same silicon for AI chips that sell for a starting price of £10,000, Also much less of a hassle.

In the short term though I reckon we are exceptionally close to £1000 MSRP xx70 tier cards.

 

[Grim5]

As someone also mentioned on Discord the other night, I can see Nvidia eventually down the line exiting the game GPU market IF AI keeps going.

It makes no sense to sell a £1000 GPU here and there when you can instead prioritize that same silicon for AI chips that sell for a starting price of £10,000, Also much less of a hassle.

In the short term though I reckon we are exceptionally close to £1000 MSRP xx70 tier cards.

I think Nvidia would only completely drop gaming if AI demand was so high that the wafer supply could not keep up and so all production could be allocated to AI high margin products. But as long as AI demand is less than TSMC wafer supply it makes sense to make extra products so have a diverse portfolio. It also depends on hard it is to modify the AI architecture for gaming, if it's easy and you have the wafers and the performance is competitive then you may as well do it if people buy it. The gaming versions of Nvidia architecture is also used in other industries, for example to power the entertain system computers in cars - Mediatek just signed a new deal to use Blackwell chips in its next gen car entertainment system products and these are not Blackwell AI chips they are basically gaming architecture chips that need to compute and output motion graphics

 

[Rroff]

I think Nvidia would only completely drop gaming if AI demand was so high that the wafer supply could not keep up and so all production could be allocated to AI high margin products.

Unless they had no choice (ultimately they have a responsibility to shareholders) I can't see them exiting the gaming market - they have significant gaming teams/culture and people at high levels who are enthusiastic gamers.

EDIT: Though maybe they might subsidiary/split it off or something in the longer run.