• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Discussion Ada/'Lovelace'? Next gen Nvidia gaming architecture speculation

Page 23 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
igor_kavinski said:
The ARM acquisition could also have been Nvidia's ticket to low power ARM GPU cores, allowing them to ditch their power hungry cores and use multiple power sipping modified Mali cores developed exclusively for themselves. They could still do that through licensing but it wouldn't be the same as having the brilliant ARM engineers in-house to do their bidding.
Click to expand...
I don't know if it's sarcastic but the ARM acquisition was to ditch MALI and license GeForce GPUs instead and develop new CPUs cores for HPC and datacenter markets
 
xpea said:
19k for Time spy extrm is Kimi sandbagging on cut down low frequency AD102. Full die (SM and all L2 cache activated) equivalent to AIB OC models will be above 25k at less than 600W (can't be more specific for obvious reasons)
Click to expand...

By "low frequency" probably still means a 2.7-2.8 effective clock. Unless there's no point because it's that bandwidth limited. If a presumed 10% L2 cut for the 4090 makes that much of a difference, gotta figure that the issue is only going to be worse with AD103 and below since they have less memory bandwidth than the equivalent? Ampere model.

Samsung btw promised 27 gbps GDDR6 and also talked about 32 gbps GDDR7 but seems like only 24 GDDR6 is coming. And the 4090 is probally using 21 GDDR6X.
 
igor_kavinski said:
Did Nvidia mention that? Getting MALI to emulate CUDA etc. would be easier than to get their Geforce cores to be power efficient, unless you want something like the GPU used in Nintendo Switch.
Click to expand...
Nothing was really disclosed until the deal was approved but yes, the goal was to replace Mali with GeForce.
Regarding Nvidia graphics IP, it scales very well at low power. In fact, the GPU part of NV SoCs were never a problem on the Tegra line. On the contrary.
Finally, a small story about the Switch, Nintendo refused a deal given by Jensen himself of a custom SoC with Pascal, instead of Maxwell, for only $1 more. With Pascal Tiled based rendering, this custom SoC was planed to be much more efficient and much faster. Still Nintendo bean counter execs didn't see the point and they went for the standard old SoC that we all know and complain about. Lost opportunity...
 
xpea said:
Nintendo refused a deal given by Jensen himself of a custom SoC with Pascal, instead of Maxwell, for only $1 more.
Click to expand...
I would blame this more on Jensen. He gave them the deal figuring they would go with the more recent architecture and that would get him an extra $1 per SoC. The cost to Nvidia for the better SoC wouldn't have been substantially more. Jensen, as usual, did his part depriving gamers of something good.
 
igor_kavinski said:
I would blame this more on Jensen. He gave them the deal figuring they would go with the more recent architecture and that would get him an extra $1 per SoC. The cost to Nvidia for the better SoC wouldn't have been substantially more. Jensen, as usual, did his part depriving gamers of something good.
Click to expand...
That's one way to think about it. But the die was bigger and more important Nvidia was taking care of all R&D cost for this custom SoC, including tape out. It was a good deal in my opinion but you are free to disagree.
 
DooKey said:
Not sure it will be $2K. If the rumors are true and NV only releases the 4090 this year, then there's no way it's $2K. They can't take the kind of hit that would cause on the bottom line.
Click to expand...

They're not going to change what they intend to charge for the 4090 because they are releasing the 4080 and 4070 3 months after the 4090's launch instead of 1-2.

And they are not releasing the 4080 and 4070 this year because they need GA102's inventory to clear first. That is potientally an issue for nVidia but releasing new products would only make things worse. Unless you are charging more for it.
 
It seems to me that Nvidia wants to ask 1500 for the 4090. That's why they said that the reduction of the 3090 MSRP from 1500 to 1300 is only temporary. The $2k part would be the 4090 Ti.

Of course, whether they actually can get away with it is a different question.
 
DooKey said:
Not sure it will be $2K. If the rumors are true and NV only releases the 4090 this year, then there's no way it's $2K. They can't take the kind of hit that would cause on the bottom line.
Click to expand...

Given how well the 3080/3090s sold for well over $2000 or even $3000 last year, I don't think they will have any trouble selling at that price. And with $10%-ish inflation that will be even more when these cards come out.

If I'm honest, I would probably get it myself at that price, even though it's a complete ripoff. 😛
 
A bit less than twice as fast. Just drop in the new 24gbps GDDR6X (I wonder if it costs more than Samsung's equivalent GDDR6 ram, and Nvidia just has to buy it due to contract issues...) and re-architecting the cache heirarchy towards a similar efficiency gain as AMD did with RDNA2 and... yeah you get that exact number. It'll probably also be around 450 watts, the practical limit for air coolers, which Nvidia and AMD (and Intel) want to keep around as whos going to trust a minimal profit margin AIB to make a reliable (and thus expensive) water cooler? No of them so long as they might in any way be liable or appear liable for any damage when one fails.

So the score seems really plausible. Hooray plausibility!
 
biostud said:
Combined with Corona shortages.
Click to expand...
Not really, the nadir of graphics card sales (in units) was Q2 2019, while COVID was not even a thing. It is true that Nvidia couldn't scale up production as hard as they did during the previous mining boom, but their production numbers were quite similar to 2014-2016. So without the immense demand from miners, the gap between demand and supply would be way, way smaller and thus the prices far lower.

See: https://www.tomshardware.com/news/gpu-analysis-jpr-q4-2021
 
If rumored perf is true, 4090 will easily sell for $2000 *initially* and in limited volumes as a new product. Once its 'new product' allure wears off after several weeks, then I can see it going down lower in price (if they manage to clear out all 3090 inventory).
 
Aapje said:
That's why they said that the reduction of the 3090 MSRP from 1500 to 1300 is only temporary.
Click to expand...

It's probably just as simple as the price cut is aimed at reducing inventory. If they get ahead of schedule, they will end the rebates or whatever they are doing.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top