Will GeForce RTX 3070 Stock Be Any Good? Etailer inventory seems low

What's the normal stock amounts for past launches? Do we have info on that? How many 1080's did this etailer get at this point prior to launch? 2080s?

People keep looking at these numbers and saying "it's so low" but I haven't see comparisons to any previous launches at all.

 

January seems a good target for those wanting a 3000 card...

 

@H83 January 2022?

 

Probably a bit of both. Samsung 8nm might be ok for small mobile chips but not for Nvidias monstrous die sizes.

628 mm² for the 3080/3090 and even 393 mm² for the 3070 isn't exactly small.

My 1080 was considered a decent size chip to launch on a new 16nm process with a 314 mm² die size

 

Yep. But it is important to learn from RDNA1. Which outperformed GCN a lot with same transistor investment while eating less energy.
(No changes in game code. It did it thanks to architecture changes.)

GCN had damn bad scheduling. RDNA1 boosted it a lot and another improvement is coming.

And looking at ~200W XSX (wall socket measurement). Factoring PSU power efficiency. VRMs power efficiency. All components on PCB and add-on components... Removing 10 GDDR6 chips... APU itself with 8C/16T Zen2 + RDNA2 (52/56 enabled CUs) eats impressively small amount of energy.
RDNA1 (Navi10) with 40 CUs 8x GDDR6 at lower clock alone eats inside PC more than entire XSX on wall socket at almost same GPU clock.

I expected XSX measurements to be around 275W. We may get to see that 40CUs RDNA1 GPU will eat twice as much as will RDNA2 GPU itself at same clock.

Sure, entire card will have just around advertised 50% improved performance per watt due to static power draw from other components. But when GPU eats fraction of that budget, increasing TBP from 225W by 10% will actually increase GPU's power budget much more.

And that's why I would like to see 320W 3080 limited to same 215W TBP as 2080 uses. I can imagine that its 50~60% performance advantage would drop significantly.

 

No. Not really. Unless 4% is "significantly".
Meaning that even at 200W 3080 would be > 60% faster than 2080 at 4K.

 

Good question. COVID 19 will have caused a host of problems throughout the manufacturing stream. Everything from a shortage of components and materials due to non essential industries being on lockdown for an extended period, to difficulties in shipping everything.

If there is a Covid related reason for gpu shortages then we'll see similar issues with Navi when that launches

 

damn i hope i can get one,i finally saved enough money to move on from my trusty gtx970 lol

 

Not too surprised by this considering on how the 3080 and 3090 went.

To me I would say both COIVID-19 played a part in this as well on top of the other problems.

 

It's like the CNN of video cards. It must be Russia's fault.

 

And you seriously believe it? Then You should call nVidia bunch of morons. (everyone should)
Because if they gained 4% by increasing power draw by 50%, they are nothing else than bunch of morons.

 

Exactly , no way it is true , we need real source for this . maybe Hilbert could do similar test or at least comment on that

 

Guess they are a bunch of morons then, dunno.

Either that or they knew where AMDs cards were going to fall and figured that staying competitive was more important than the blowback from the higher power draw. It's also possible that turning on DLSS/RT makes these undervolts unstable. Not sure.

 

I am pretty sure, that if anyone sane sees:
Card A: 200W and performance X
Card B: 300W and performance X * 1.04

That they'll call card B bad and will buy card A especially since they cost same. There can be many issues with undervolting. If there were no issues, there would be no need to undervolt that much as nV would set lower stable voltages themselves.

Often, games look OK as you do not notice artifact here and there once in a while. But calculations result in defects anyway. And when card uses visual error correction techniques, it is even harder to notice problems.
But running GPUPI while gaming may show true effect of undervolt.

I always tested OC/UV with math based benchmarks where one expects same result every time. Solo and in combination with game workload.

Same way as I test system memory OC by telling steam to check some large game's CRC. Steam knows what result should be, if memory is not stable, there will be error.

 

Except... it's the same card. And it's been like this since forever. Power going exponentially with voltage*freq.
The only thing different with 3080/90 is that Nvidia has provided both cooler and components that allows these cards to go beyond 300W straight out of the factory.


PS
5% for hundred-ish Watts - called it long ago
https://forums.guru3d.com/threads/n...flagship-50-faster.433561/page-3#post-5814436

 

With current state of affairs ,AMD will become new IT bohemut of IT industries

 

It's definitely going in that direction. More cores for less money.

 

I've got to laugh at this. Right now AMD is trying to gain market-share, ofcourse they're going to try to beat Intel on price. However, I give it 2yrs max and you're all going to be moaning about the price of the latest AMD cpu+chipset.

Consumer hopes and desires has little to do with how actual business works.

 

There is just so many things wrong in your post that make me laugh.

 

All the bickering over nothing.

Optimum Tech simply has a "Golden Sample" GPU that works well at lower voltage.
It doesn't mean that ALL 3080's work as well at that lower voltage. Most will probably crash as soon as game is launched when lowering the voltage that much.

Today computing chips don't differ much in the clocks they can reach, but in how much voltage is needed to reach a certain clock.
This is why "undervolting" is possible and widely used. It is no different than overclocking. Silicon lottery, nothing else.