- Status
Whoa the 3090 Ti is faster than the 4080 12GB? That’s a pretty hard fail for the 4080.
Im curious what Ada will deliver for laptops, given the 4nm process and efficiency gain with DLSS3, RT and similar TGP to Ampere, maybe the laptops will perform closer to the desktop counterparts this time.
Ideally, I would like them to introduce a SoC like design where they are integrating High End GPUs into a SoC like Apple does with the M chips, that could lead to big efficiency gains further closing the gap from desktop to mobile. Though, of course they'd have to work with Intel and AMD to make this possible. Franky, AMD has a much higher chance of achieving this, given they develop CPU and GPU inhouse.
Ideally, I would like them to introduce a SoC like design where they are integrating High End GPUs into a SoC like Apple does with the M chips, that could lead to big efficiency gains further closing the gap from desktop to mobile. Though, of course they'd have to work with Intel and AMD to make this possible. Franky, AMD has a much higher chance of achieving this, given they develop CPU and GPU inhouse.
davis.anthony
Veteran
Remember the 8800GT that became the 9800GT that became the GTS240?
And the 9800GTX that became the 9800GTX+ that became the GTS250?
That's what they're doing here but without changing the labels on the rtx3000 GPU's to say rtx40xx
They've come full circle, they used to make massive, power hungry GPU's (GTX200 Series and Fermi) and then dropped that and moved to smaller, more efficient GPU's (GTX600 series) and have slowly worked their way back to massive GPU's again.
And the 9800GTX that became the 9800GTX+ that became the GTS250?
That's what they're doing here but without changing the labels on the rtx3000 GPU's to say rtx40xx
They've come full circle, they used to make massive, power hungry GPU's (GTX200 Series and Fermi) and then dropped that and moved to smaller, more efficient GPU's (GTX600 series) and have slowly worked their way back to massive GPU's again.
davis.anthony
Veteran
Does it seems any better than PowerVR's level 4 implementation?
I think calling it a 5 is a stretch imo.
That’s because it’s the 4070
Games that are well optimized for Ampere won't run into scaling issues on Ada which is more or a less a wider Ampere with some shader execution improvements (mostly visible in RT though I suppose).
If you look at results here the scaling between new cards in titles are all over the place.
RE8 does 0.8/1.2/1.7
ACV does 0.9/1.2/1.5
TD2 does 1.0/1.2/1.6
The only card showing the same scaling between these titles is 4080-16 which doesn't make much sense.
Also 4080-16 is more or less the same as 3080Ti but with higher clocks and a different memory system.
Flops gain on it compared to 3080Ti as at ~45% and all of this is basically from higher clocks.
In these three titles 3080Ti to 4080-16 scaling is at +38%, +29% and +26%. These look weird all things considering.
Last edited:
Frenetic Pony
Veteran
Yeah, I was expecting better. And only 3 actual titles, ones they assumedly chose because they make the new cards look good, and that's all the performance they're touting? What's worrying is that they don't mention what resolution these games are running at either.
davis.anthony
Veteran
I have no words for how impressive that tech is!
The amount of games I would love to see using this...... Imagine F.E.A.R with full RT
I really don't mind a smaller leap in raster/fp32 perf where more focus is on RT and other more interesting efficiency optimizations like DLSS. We're seeing it already where high end games use these features and really need them to shine are becoming more commonplace.
The pricing and model naming though is just so Nvidia. Big fuck you to gamers. I hope they sell very few both Ampere and Ada.
The pricing and model naming though is just so Nvidia. Big fuck you to gamers. I hope they sell very few both Ampere and Ada.
They call those x080 now...
davis.anthony
Veteran
Wonder if Daggerfall could be a candidate? Definitely see studios using this to generate additional revenue on old games.
Ugh, after skimming some pages on an enthusiast(!) HW forum, it seems gamers will all do an EVGA this round.
Maybe this wasn't Jensen, but actually his twin from a parallel omniverse, where we have cold fusion for unilimited energy, Musk has a house on Mars, and Apple watches provide life extension.
At least it's big enough so they can sell it as Steam Deck cases, if all else fails.
Sorry. Couldn't resist.
Problem is i'm not so optimistic AMD does better well enough. Slow death of dGPUs confirmed.
Maybe this wasn't Jensen, but actually his twin from a parallel omniverse, where we have cold fusion for unilimited energy, Musk has a house on Mars, and Apple watches provide life extension.
At least it's big enough so they can sell it as Steam Deck cases, if all else fails.
Sorry. Couldn't resist.
Problem is i'm not so optimistic AMD does better well enough. Slow death of dGPUs confirmed.
Flappy Pannus
Veteran
They will be selling them to a market now that actually wants the bulk of them for gaming, though. There actually are limits to the Geforce brand, as it turns out.
No where in my post do I use the word disappointing. From a gaming view, 3x the transistors and “2 node shrinks” for a 50% performance uplift just doesn't seem particularly impressive in terms of architecture.
- Status
Similar threads
