I'm talking more about the optics. Of course Nvidia can follow suit and drop prices as well.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
NVIDIA ‘Ampere’ 8nm Graphics Cards
- Thread starter LoadsaMoney
- Start date
Soldato
- Joined
- 14 Jul 2005
- Posts
- 9,394
- Location
- Birmingham
Hmm yeah just thinking how best to show it, as it was more a mid gen refresh with a bit more performance at the same dollar release price.
Turning should be all on one messy up and down line too
Soldato
- Joined
- 14 Jul 2005
- Posts
- 9,394
- Location
- Birmingham
Turning should be all on one messy up and down line too
Dont agree, as there is clear distinction between the original dud release and the super range. Its as good as a new gen on its own.
No one with a 2070 was eyeing an upgrade though, same with 2080. They look like a great upgrade to 10 owners, but all 7, 9, 10 are on same lines, so 20 should be too, otherwise your covering for the dud release, the up and down crazy line will show it, it'll highlight the ****ery
Last edited:
Soldato
- Joined
- 14 Jul 2005
- Posts
- 9,394
- Location
- Birmingham
I think showing it this way highlights that Turing 1 was a dud release - its not covering it up at all. Then it shows that they fixed it with Turing 2, which I think is a fair assessment. Yes benefit of hindsight and all that.
Here is my prediction of where the Ampere series will land. No science to it, just lined it up visually by eye.
That would ruin their new lineup in price to performance, if they dropped the 2070S to £329 to compete with a £279 5700XT, then the new 3060 at £499 with a 10% bump over the 2070S looks like at idiot buy, right? I mean who'd get the 3060, £170 less for a loss of 10% no brainer. Optics or not, they can't just drop the prices or they cannibalise there own sales of the new parts, and once those parts run out AMD will mop up the sales or Nvidia need to then drop the 3060 to a sensible price.
The 2060s and 2070s look like what Turing *should have* been at luanch. The 2080s still would have been a pretty weak upgrade along that row. The 2080Ti should have been brought down around the 2080s price point and then Turing would have been decent overall. (Yet late at that point)
Prices for the top end bikes have but a decent usable light bike for general use is relatively cheap and much better specced than 20 years ago.
If an intel motherboard running a Titan RTX @x8 is not saturated then there is no chance @x16 a 3XXX card will be.
The other week I had to upgrade my 24/7 PC, I did look at AMD but in the end went for an intel setup as the AMD CPUs don't overclock that high and would really throttle a couple of Titans.
„Horizon: Zero Dawn on PC shows significant performance difference between 8x and 16x PCIe 3.0”
https://www.reddit.com/r/hardware/comments/i46qtf/horizon_zero_dawn_on_pc_shows_significant/
Soldato
- Joined
- 20 Apr 2004
- Posts
- 4,422
- Location
- Rugby
A shoddy PC port is not a good example of how PCIE bandwidth scales with games
I think showing it this way highlights that Turing 1 was a dud release - its not covering it up at all. Then it shows that they fixed it with Turing 2, which I think is a fair assessment. Yes benefit of hindsight and all that.
Here is my prediction of where the Ampere series will land. No science to it, just lined it up visually by eye.
Just a little note but not that important, perhaps firestrike (ultra) would've been slightly better to use for generation to generation comparison. Timespy uses DX12 which works better on the newest cards whereas the older cards predate DX12 going mainstream so don't perform as well on timespy. But most games are DX12 now so timespy good indicator on real world perf I guess.
Welcome to your new full time job.
Welcome to your new full time job.
Yeah sorry dude haha
Soldato
- Joined
- 14 Jul 2005
- Posts
- 9,394
- Location
- Birmingham
If you can link me to a ranking I'll see what it looks like.
I think, because we're doing a comparative look across generations, it doesn't really matter what benchmark we use we'll see the same overall shape to the chart. The purpose isn't to be tied down to 1% accuracy on a performance increase, just to visually see the shift across time.
Just want you to know I appreciate this!Ive redone my chart showing the Timespy scores this time. The pattern is very similar, so I think this means either benchmark can be used to show/interpret the series progression.
Timespy:
Passmark:
Soldato
- Joined
- 14 Jul 2005
- Posts
- 9,394
- Location
- Birmingham
Thanks. I enjoyed doing it, its helped my own understanding of the progressions massively. I'll be ready to plot on ampere prices and scores once available.
I wouldn’t have thought so. Even something like the 2080Ti doesn’t even saturate pcie 3.0 8x never mind 16x. So the 3090 or whatever it is will need more than twice the bandwidth of the 2080Ti which I think is unlikely.
And if say a 3090 saturated pcie 3.0 16x the only multigpu option is to go threadripper. As pcie 4.0 8x on normal x570 will not be good enough for maximum performance as x570 doesn’t have enough lanes to run two cards at full 16x bandwidth.
Id be interested in seeing if there was i difference but I think by the time it may be beneficial we will have things like ddr5 etc, so most may be wanting to upgrade anyway.
Obviously having pcie 4.0 now is going to be more futureproof. But I can’t see it being all that beneficial just yet. More marketing at the minute.
Last edited:
Unless they produce non-consumer cards electrically wired for 8x PCI-E 4.0 lanes, that is a huge benefit for lower end non-workstation systems that might want more lanes for 2x GPU's or more NVMe storage.
Whut?
If you are running multi gpu on x570 you will be limited to pcie 4.0 8x which is the same bandwidth wise as pcie 3.0 16x...