winjer
Member
Saphire pulse 9060XT.
If you must buy nvidia, then the 5060 Ti 8Gb. But you have problems with lack of vram. Or get the 16GB version for 399.
Nocty
Currently Steaming
he mostly plays older games like Dota, league of legends and fortnite.. so shouldnt be an issue. thx
MMaRsu
Member
Then just put a 30 or 40 series lol
Nocty
Currently Steaming
well I'd like to give him at least 'some' future proofing even if its not a lot. lol. DLSS ftw
PeteBull
Member
NVIDIA GeForce RTX 5070 Specs
NVIDIA GB205, 2512 MHz, 6144 Cores, 192 TMUs, 80 ROPs, 12288 MB GDDR7, 1750 MHz, 192 bit
YeulEmeralda
Linux User
2070 to 3070 = 68% improvement
3070 to 4070 = 30% improvement
4070 to 5070 = 20% improvement
We used to have impressive gen over gen jumps in performance.
Try path tracing in Indy Jones.
roosnam1980
Gold Member
Just received mines
asdasdasdbb
Member
Yeah because the 'shrinkflation'.
2070 = 445 mm2
3070 = 392 mm2 (and a cheap node at that)
4070 = 294 mm2
5070 = 263 mm2
Insane Metal
Member
And it's only going to get worse, unfortunately.
PeteBull
Member
I think we still got 1 good gen of gpu's before us, the 60xx coz it will be on 3nm, problem is, knowing nvidia and its greed, those cards gonna have crazy msrp and even crazier streetprice, but what u gonna do, some of us will be hyped/forced/hungry for dat new shiny gpu piece in early 2027, by then we might already not only know legitimate leaks but official specs of ps6 and xbox pc, which will put additional preasure for pc crowd to think of gpu upgrade.
TLDR: Rip our wallets 2027 onwards
Edit: Example of solid proposition would be for me rtx 6080, 24gigs of vram and at least as strong as rtx 4090(so 15-20% stronger from current rtx 5080, not a crazy jump but good enouh), on top of that- not 450W tdp like 4090, instead 300-320W tdp, and 999msrp, with actual streetprice for good cooling models(not top of the top but solid, aka cool and quiet) around 1200$.
That is still massive progress both price/perf aka value and for many of us almost as important- at tdp too.
Will we get this card? - likely we will, but logics tells me msrp gonna be 1200$ and actual streetprice of models with good cooling closer to 1500$, at least durning first few months of launch window.
Last edited:
T4keD0wN
Member
If a 6070 doesnt perform at the very least like a 4090 at under 300w for like 700 then the GPU market is wrecked beyond repair.
Name ultimately doesnt matter, but calling something that goes toe to toe with a 2 gen old card a 6080 would make Nvidia look like clowns.
Last edited:
PeteBull
Member
Thats impossible unfortunately, 4090 is already on 5nm and eats 450W at really big die size of 609 mm² and its actual streetprice, right before 5090 launch so not even year ago was 2k usd.
60xx series gonna be on 3nm, 6070 in best case scenario could be same price as 5070 aka 550$(not very likely, 700-800$ streetprice seems more realistic, lets forget about msrp coz those wont matter for normal customers, only for reviewers) but 5070 to 4090 is 2x vram pool and +71% avg performance
NVIDIA GeForce RTX 5070 Specs
NVIDIA GB205, 2512 MHz, 6144 Cores, 192 TMUs, 80 ROPs, 12288 MB GDDR7, 1750 MHz, 192 bit
Even gtx 970 3,5/4 GB on 28nm to 16nm pacal gtx 1070 with its 8gigs vram was on avg below 50% performance jump(unless vram starved, then difference is bigger ofc)
NVIDIA GeForce GTX 970 Specs
NVIDIA GM204, 1178 MHz, 1664 Cores, 104 TMUs, 56 ROPs, 4096 MB GDDR5, 1753 MHz, 256 bit
So basically unless we get much bigger jump than from maxwell to pascal(which was fricken goty) u will deem whole rubin generation/ including 6070 model as failure =D
Personally if 6070 is at least 16gigs of vram and at 5080 performance lvl(so almost 50% jump, still 15-20% lower from 4090) at 700-800 streetprice it would be solid offering and sell well, for simplicity it would be like 5080 but at 700-800$ price so 20-30% pricecut with bit better dlss/rt capabilities.
T4keD0wN
Member
I dont think i will be disappointed in all aspects, maybe by their sneaky naming schemes, like having 3060ti (392 mm² GA104) succeeded by a 4060ti (188 mm²AD106) and VRAM shenanigans, granted my oppinion is as low as it gets when it comes to these practices.
But going from 450w to sub-300w for the same performance in 2 iterations sounds like a given even in a pessimistic scenario.
Thats sounds nice in 2025, but it really doesnt look that enticing if you could pretty much buy a 5080 now and already enjoy it for years, albeit for a bit more money by the time that card would release. They cant even attempt another 4x framegen nonsense since theyve already played that card, i think theyll need to deliver something this time or budge on margins and the latter aint happening.
PeteBull
Member
Tech progress slowed down and keeps slowing down more and more, personally i dont even consider 50xx series "new gen" coz its same 5nm process node and very close performance/watt, thats why nvidia was trying to improve dlss/rt capabilities and even introduce multiframegen, coz they cant offer proper raster performance jump from 40xx to 50xx series.
Lets look at previous bis gpu aka rtx 4090 - Die Size 609 mm² at 450W tdp vs current even more expensive bis gpu rtx 5090- Die Size 750 mm² at 575W tdp and thats on avg 31% performance increase.
Same 5nm process node- nvidia couldnt get 3nm process node so they had too make the gpu die that humongous which in turn made gpu eat that much power and be hot af- aka expensive large cooling needed- ofc it made its price even higher vs for example GOTY pascal gpu gtx 1080ti - Die Size 471 mm², only 250W tdp and 699$ both actual msrp and streetprice - even if we account for inflation from 2017- thats 924usd now in late 2025.
Now with new 3nm process node rubin family wont neccecarily be cheaper(especially if we include nvidia's greed here) but at least it will eat visibly less power and will be visibly more performant(within reason, so ~40% jump for similar die size).
Saying all that ofc nvidia isnt out of their mind to put 6070 msrp at 1000 or 1200usd and only then that chip would be big/performant enough to be on pair with 5nm previous bis and current 2nd bis rtx 4090.
Lets look at previous bis gpu aka rtx 4090 - Die Size 609 mm² at 450W tdp vs current even more expensive bis gpu rtx 5090- Die Size 750 mm² at 575W tdp and thats on avg 31% performance increase.
Same 5nm process node- nvidia couldnt get 3nm process node so they had too make the gpu die that humongous which in turn made gpu eat that much power and be hot af- aka expensive large cooling needed- ofc it made its price even higher vs for example GOTY pascal gpu gtx 1080ti - Die Size 471 mm², only 250W tdp and 699$ both actual msrp and streetprice - even if we account for inflation from 2017- thats 924usd now in late 2025.
Now with new 3nm process node rubin family wont neccecarily be cheaper(especially if we include nvidia's greed here) but at least it will eat visibly less power and will be visibly more performant(within reason, so ~40% jump for similar die size).
Saying all that ofc nvidia isnt out of their mind to put 6070 msrp at 1000 or 1200usd and only then that chip would be big/performant enough to be on pair with 5nm previous bis and current 2nd bis rtx 4090.