Speculation and Rumors: Nvidia Blackwell ...

  • Thread starter Deleted member 2197
  • Start date
Hard to imagine big clock increases on such a similar node. I would guess that we either get a small performance jump or Nvidia has finally changed the architecture in more notable ways. I really hope we don’t just get frame gen with more generated frames.
 
If the 5080 is around 35% faster than a 4080 then that’ll make it probably 80% faster than my 3080 which is a nice upgrade depending on the price. Don’t care what the bus width is. If it’s faster it’s faster.

I just tried to watch this hardware unboxed video and off the rip it’s so dumb. Why does anyone care about tiers if the don’t work for the company? You have a product. Look at the alternatives and check if they’re better and if it’s worth the cost. If you don’t like it don’t buy it. Skip a generation if there’s nothing in your price range that’s worth it. Who cares about a product in some price tier you wouldn’t buy anyway. The only relevant comparison is to what you already own.
my 3080 is still playing everythign I throw at it. I may even be able to wait longer
 
my 3080 is still playing everythign I throw at it. I may even be able to wait longer
If price/performance stagnation continues, obviously the 3080 will be fine for a while longer. Games can't require $600+ GPUs for a decent experience.

Only problem you may run into is VRAM capacity (which is why bus width actually does matter until the 3GB GDDR7 modules arrive).
 
I honestly think this stuff is largely driven by FOMO or PCs as status symbols. Say I waited until 60x0 series to upgrade from my 3080. Well, maybe I look at the price and performance of a 6070 and it fits better into my budget and I get a performance boost that I think is acceptable. Why not? I don't care what model it is. I don't know why anyone does, other than personal feelings about not having the best. I'm currently looking at buying new speakers and there's no reality where I'd ever go back and do a historical analysis of price tiers and models to try to figure out if the model number of the speaker has changed relative to other models in a stack. It's just a completely absurd thing to do.
I really have no idea what the point of that video was, other than trying to convince consumers they’re gonna be screwed by products which haven’t even been announced yet. Tim has some strange ideas about how product sales works.
 
Hard to imagine big clock increases on such a similar node. I would guess that we either get a small performance jump or Nvidia has finally changed the architecture in more notable ways. I really hope we don’t just get frame gen with more generated frames.

I hope so too since that would be a travesty.

Let’s not forget the 4090 is “only” 30% faster than the 4080. Would it be so shocking to increase performance by 30% by tweaking clocks and bandwidth and some architecture improvements? Seems doable.

The 4090 is a monster on paper but less so in benchmarks.
 
Devil’s advocate here - an alternative explanation is they do intend to price them high and don’t want sales dented by the last gen cards. Harder to sell a $1300 5080 with plentiful 4080 super supply. Same with 5090.
Wouldn't work, people would just wait for a 5070 - which we all know is coming eventually.
If 5080 would be priced at 4090 levels then they wouldn't have EOLed 4080/S right now. Something like the original 4080 price is possible of course but then EOLing 4080S seems a bit premature.
I take these rumors as an indication that 5080 should land in $1000-1600 range at the very least. 5090 who knows, no ceiling on that one.
 
Wouldn't work, people would just wait for a 5070 - which we all know is coming eventually.
If 5080 would be priced at 4090 levels then they wouldn't have EOLed 4080/S right now. Something like the original 4080 price is possible of course but then EOLing 4080S seems a bit premature.
I take these rumors as an indication that 5080 should land in $1000-1600 range at the very least. 5090 who knows, no ceiling on that one.
I don’t necessarily disagree, but I’m not sure I follow you given the price range you quote - earlier you said the EOL suggests a relatively close price but you quote up to $1,600. If they price the 5080 that high why would anyone buy one if there was still a 4080 Super widely available.

I do disagree that there’s no ceiling on the 5090 price, but I think that’ll boil down to a semantical different rather than a substantive one 😉.
 
I don’t necessarily disagree, but I’m not sure I follow you given the price range you quote - earlier you said the EOL suggests a relatively close price but you quote up to $1,600.
Because there are two scenarios:
1. 5080 and 5090 will be priced at about the same prices as 4080/S and 4090 are. All in all this is a more likely scenario give or take a couple of hundreds of bucks.
2. The range between 4080 to 4090 prices will be assigned to 5080 alone and 5090 will go above that. This is also possible however in this case EOLing 4080S seems a bit premature as I've said as this implies that 5080 will land at about $1300 or so - 4080S at $1000 wouldn't pose any threat to it in this case.

If they price the 5080 that high why would anyone buy one if there was still a 4080 Super widely available.
Weird question. Performance and features, as usual.

I do disagree that there’s no ceiling on the 5090 price, but I think that’ll boil down to a semantical different rather than a substantive one 😉.
Well strictly speaking there is no ceiling on how high a GPU can be priced. Whether anyone would buy it at a very high price is a different matter.
 
Hard to imagine big clock increases on such a similar node. I would guess that we either get a small performance jump or Nvidia has finally changed the architecture in more notable ways. I really hope we don’t just get frame gen with more generated frames.
Rumors have said boosts are past 3Ghz, and panzerlied claims 5090 base clock is “nearly 2.9Ghz.” We don’t know what meaningful difference there is between 4N and 4NP, but it’s unlikely to account for that big of an increase (if true).
 
Weird question. Performance and features, as usual.
I didn’t mean that comment literally, but I really think they’d have a hard time moving a $1600 5080 if there was still lots of 4080 Super stock. Unless DLSS 4 has a truly killer feature that’s Blackwell exclusive.

To be clear though, I was genuinely playing devil’s advocate in my earlier post. My initial reaction to that news was the same - though this generation is atypical in that the previous one is sharing node capacity. I know it’s not unheard of. But I’m also wondering whether this EOL is as much about capacity as it is price.
 
I didn’t mean that literally, but I really think they’d have a hard time moving a $1600 5080 if there was still lots of 4080 Super stock.

I think the line of reasoning here is that there would be no reason to shut down 4090 sales unless the 5080/90 makes it a bad deal at $1600.

Two ways that can happen.

The 5090 is much faster at ~$1600 or the 5080 provides similar/better performance at <= $1600.

I suppose there’s another option where Nvidia thinks the 4090 is too good of a deal and the 50 series will be worse bang for the buck.
 
I suppose there’s another option where Nvidia thinks the 4090 is too good of a deal and the 50 series will be worse bang for the buck.
That’s my concern - I’m not saying I think that’s likely what the EOL means, just that I kinda see it as a possibility.
 
Well strictly speaking there is no ceiling on how high a GPU can be priced. Whether anyone would buy it at a very high price is a different matter.
True, though the 5090 is not unique in that respect 😉. Appreciate your responses to my earlier comments, thanks.
 
If price/performance stagnation continues, obviously the 3080 will be fine for a while longer. Games can't require $600+ GPUs for a decent experience.

Only problem you may run into is VRAM capacity (which is why bus width actually does matter until the 3GB GDDR7 modules arrive).

VRAM is my biggest concern regarding this pricing 'tier' discussion and the main point I can actually get behind in HUBs historical x80 spec/price analysis video.

I really would like to see more product tiers at the high end to address this. I long for the simple days of Geforce 4 for an obscure and dated example: 3 tiers of cards with the same VRAM occupying the highest tier, 2nd and 3rd. I understand that modern process methods might make selling more of a binned die part uneconomical and a bad case for business, but with 3GB GDDR7 chips not being available now, cutting down the second tier chip by half of what the top tier is leaves it with same VRAM as my current card from 5 years ago. History has taught be that this is a bad upgrade and that VRAM will very likely hamper the longevity of this purchase. So what should I do, stretch my budget and pay a lot more? Historically it wasn't always that way...

I would really hope to see a proper second tier card with a somewhat generous amount of VRAM thought out from the design phase (if binning the top tier to obtain a wider memory bus isn't feasible). My expectations are that after 2 generations of the top tier having 24GB VRAM, that the upcoming 2nd tier should now have that amount of VRAM - simple as that. We might end up getting that with 3GB GDDR 7 chips, but if the wait is 6-12 months from the original part then that's not ideal either because many will buy the initial part unaware of another option coming - or buying that part makes the value worse because a newer replacement is closer on the horizon...
 
If price/performance stagnation continues, obviously the 3080 will be fine for a while longer. Games can't require $600+ GPUs for a decent experience.

Only problem you may run into is VRAM capacity (which is why bus width actually does matter until the 3GB GDDR7 modules arrive).
yea I may get a 5070 or amd equivalent just to bring down the card size and power requirements.
 
yea I may get a 5070 or amd equivalent just to bring down the card size and power requirements.
I'll definitely wait at least until an 18GB 5070 Super or something. Not getting another 12GB card. It's already been a bit of a problem in a couple games I've played.
 
Back
Top