Nvidia's 3000 Series RTX GPU [3090s with different memory capacity]

Malo said:
There is no need to be rude, I like to consider this a mature forum. It's all speculation of course since numbers are very hard to come by and I'm merely expressing my thoughts on why I think miners are getting the lion's share.
Click to expand...

After looking at the post above from me do you still feel lion's share is going to miners? I tried to do some math on numbers we know and made some guesses. I find it hard to believe most cards would be going to miners. If you feel so, maybe you could fudge the numbers to direction you feel are more correct.

edit. My numbers are probably pessimistic. I forgot to add component estimating how much of hash rate increase is due to gamers mining when not gaming. My math assumes no gamer mines which is not true.
 
Malo said:
There is no need to be rude, I like to consider this a mature forum. It's all speculation of course since numbers are very hard to come by and I'm merely expressing my thoughts on why I think miners are getting the lion's share.
Click to expand...

You found that rude?
That I did not give any value to pure speculation with no data?
Erhmm...what?
 
manux said:
I forgot to add component estimating how much of hash rate increase is due to gamers mining when not gaming. My math assumes no gamer mines which is not true.
Click to expand...
Yeah that's probably a significant factor. Anecdotally there are gamers here and other forums that have their Amperes sitting there mining instead of gaming, expressing that they are specifically not gaming at the moment due to mining craze. "Miners" includes both organized companies and groups, and the home gamer. The latter we have a new GPU in the hands of a gamer, which is good, but for your calculation purposes it does factor into it I guess.

I'm not sure how you would estimate that though.
 
Malo said:
Yeah that's probably a significant factor. Anecdotally there are gamers here and other forums that have their Amperes sitting there mining instead of gaming, expressing that they are specifically not gaming at the moment due to mining craze. "Miners" includes both organized companies and groups, and the home gamer. The latter we have a new GPU in the hands of a gamer, which is good, but for your calculation purposes it does factor into it I guess.

I'm not sure how you would estimate that though.
Click to expand...

I think we can just throw numbers around to get a range of sales going to miners. Something from super pessimistic to super optimistic. The concrete numbers we know are the nvidia revenue for q3,q4 and estimate for q1. We also know eth hash rate increase. Hash rate increase gives us about 3.1 million 3070 level gpu's being added to mining capacity during q1 of 2021.

The most pessimistic view I can see is is following. Nvidia makes 400$ revenue per 3070. 2 billion out of nvidia's 2.56 billion dollar gaming revenue comes from selling geforce cards. This results to 5 million cards sold, 3.1 million cards went to mining. This assumes all eth hash rate increase comes from new ampere cards. No amd, no used, no nothing else. So miners took 3.1million, gamers got 1.9million. I see no way of the situation being worse than this from gamer POV.

One could assume some gamers mine and some capacity comes from amd and also used cards. There also is the nvidia mining cards that are already being sold and used(turing based for now, ampere based mining cards are coming out q2). Perhaps more than 2 billion of gaming revenue came out of geforce card sales and perhaps 400$ is too high estimate of revenue per card. As eth price goes up it becomes even more profitable to use also ancient cards to mine so people might be digging up any gpu they can find and mine with them.
 
Last edited:
The closest I have come to seeing BOM price was this posted on a forum a few years back:
vkRDt.jpg


No clue if it is accurate though, but prices has not gone down on BOM since, especially on the GPU itself.
 
Putas said:
Remove one zero and I might believe it.
Click to expand...

Removing one zero makes the crypto effect to supply 10x less. 10x more gpu's sold while crypto capacity increase will still sit at 3.1million gpu's.

Though the memory alone is probably more than 40$. I believe nvidia sells gpu and memory together.
 
HLJ said:
You found that rude?
That I did not give any value to pure speculation with no data?
Erhmm...what?
Click to expand...

We already had this argument 2-3 weeks ago ? Miners mine with pure willpower ? We know that most AIBs sell directly to big mining groups. We know it because thoses mining groups showed it. We know nVidia do too (a big contract was in the news few weeks ago).

Spéculations about this are not baseless. Believe what you want but it's not because you're out of touch with the mining world that you can dismiss this cause. And you know that hard numbers from AIBs about this will never be public...
 
Rootax said:
We already had this argument 2-3 weeks ago ? Miners mine with pure willpower ? We know that most AIBs sell directly to big mining groups. We know it because thoses mining groups showed it. We know nVidia do too (a big contract was in the news few weeks ago).

Spéculations about this are not baseless. Believe what you want but it's not because you're out of touch with the mining world that you can dismiss this cause. And you know that hard numbers from AIBs about this will never be public...
Click to expand...

That nvidia contract was specific to mining gpu's that are based on turing. Those are supposedly made from defunct chips that are not useful for gaming.
 
So now "supposedly" is enough ;) You're right (but the point was miners =easy money for gpu sellers, and thet know it).

Anyway, maybe it's posts for the crypto topic...

BTW, have we numbers about the samsung 8nm process ? Things like density, etc, vs TSMC 7nm ?
 
DegustatoR said:
GA102 vs N21 is 45MTr/mm^2 vs 51.1MTr/mm^2.
GA100 is 79.4.
"Average/peak" process density is rarely relevant for end products as it can vary greatly between different designs according to their targets.
Click to expand...
Often, the amount of on-die SRAM is driving X-tors/mm² up as well. Another important factor is the design for fast-paths.
 
You must log in or register to reply here.

Similar threads

S
Nvidia Pulsar - variable refresh rate with backlight strobing
Replies
9
Views
1K
xz321zx
X
T
GPU Stability testing for Nvidia cards - PowerSweep
Replies
0
Views
468
T2098
T
DmitryKo
Folding@home Beyond3D_Team - Summer 2023 Call for NVidia RTX users
3 4 5
Replies
80
Views
10K
Davros
D
D
Intel Max GPU series for HPC, AI and Data Centers
Replies
3
Views
2K
trinibwoy
T
D
Paperweight GPU back in business
Replies
4
Views
858
Davros
D
Back
Top