I'm talking OC. 3080 can't OC worth squat. 2080 Ti can fly given proper cooling (aka any half-decent air cooler), so 3080 goes from 25% faster (stock vs stock) to 15% faster (OC vs OC). At 4K.
-
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.
AMD Navi 23 ‘NVIDIA Killer’ GPU Rumored to Support Hardware Ray Tracing, Coming Next Year
- Thread starter Gregster
- Start date
- Status
I'm talking OC. 3080 can't OC worth squat. 2080 Ti can fly given proper cooling (aka any half-decent air cooler), so 3080 goes from 25% faster (stock vs stock) to 15% faster (OC vs OC). At 4K.
>40% faster when ray-tracing is On:
https://www.techpowerup.com/review/nvidia-geforce-rtx-3080-founders-edition/37.html
6% so far on a locked Vbios. On the Founder's Edition. That makes it about as bad as a Fury X.
Steve @ GN said that maybe partner cards will fix this, but if they revert back to Pascal? that's your lot. The 2080ti is the first card unlocked in years and it was totally against the grain.
I can't see these cards overclocking significantly - they are pushing the boundaries of what you can do without a bunch of EUV layers in there.
The problem is the scaling isn't there, even when they add another 50w with the PL, you can see the gains are abysmal. Let's say you get an unlocked PL card, which is gonna be stupidly expensive but leaving that aside, then what? Gonna run at 500w to reach maybe equivalent difference to the 2080 Ti OCing scaling? Yikes! I already know how this story goes from Vega, there's just chip limits you're gonna run into even IF you manage to cool it (presumably with 2 rads) with pumping so much power.
Samsung 8nm is just awful, that's all. They chose margins over high performance, it's not shocking just disappointing, mostly because I look at the rest of the market and it's still awful for another 2 months and that's why the 3080 still looks good today even when it looks awful.
Why did Nvidia choose Samsung if its process is so bad?
Maybe Ampere as an architecture doesn't uplift the IPC?
Also, the card can potentially have other bottlenecks like memory bandwidth and more thrown power at the problem makes the things much worse.
Samsung 8nm was good enough and provided a significant transistor density improvement from TSMC 12nm. 7nm wasn't needed and might've resulted in lower yields. There is your answer.
AMD might even have production problems with TSMC 7nm, in which case, Samsung 8nm wouldn't seem that bad.
AMD might even have production problems with TSMC 7nm, in which case, Samsung 8nm wouldn't seem that bad.
Nothing awful about it as such - it is a refinement of the 10nm node and about as good as you get while still relying on DUV centric lithography. It is no 7nm but I suspect the major difference there is going to end up mostly power if we look at other 7nm devices in comparison.
Soldato
- Joined
- 6 Jan 2013
- Posts
- 22,263
- Location
- Rollergirl
Have AMD released any news yet, or are they sticking with the master plan of watching everyone buy NV cards for a month before telling us what they've got to offer?
Have to say that the 3080 benchmarks are very very 'meh'. I don't think anyone was really expecting double the performance of a 2080, but most expected more performance that we seem to have been given. Some games are only seeing a 10fps increase at 1440p which is quite honestly trash and completely not worth upgrading for. I am aware that RDR2 is somewhat poorly optimisied, but to see such little gain with a £650 GPU is quite sad. I suppose if you can flog your 2080TI for £550, and only have to put another £100 to obtain the 3080..it's not such a bad deal
I was getting very impatient for a GPU but after seeing these benchmarks would I sell my 5700XT to fund a 3080? NO WAY. I would see an increase of course, but not £650's worth in my opinion.
Waiting for Big Navi is looking the right thing to do.
I was getting very impatient for a GPU but after seeing these benchmarks would I sell my 5700XT to fund a 3080? NO WAY. I would see an increase of course, but not £650's worth in my opinion.
Waiting for Big Navi is looking the right thing to do.
It's not looking good is it?
I wondered why Nvidia had not mentioned overclocking whatsoever so far, given it's one of their marketing tools. And these are supposedly the best GPUs with the best clocks.
I already said why - margins. They get a steep discount from Samsung compared to TSMC so that means they can pocket that money instead of giving you a better card.
Soldato
- Joined
- 6 Jan 2013
- Posts
- 22,263
- Location
- Rollergirl
@matty007 I'm getting 90% increase for less than I paid for my 1080ti, so I'm happy. Obviously for the nutters that bought a 2080ti it's a bit disappointing.
Soldato
- Joined
- 21 Jul 2005
- Posts
- 21,104
- Location
- Officially least sunny location -Ronskistats
Because it was much cheaper. They had a row with TSMC, thought they could use Samsung as a weapon to get cheaper cores and it failed. So here they are.
Suppose it depends on resolution also. I am at 1440p 165hz. There just isn't enough of a jump to make it worthwhile for me at £650.
It's not bad, it's decent. But I was expecting better and it curbes my bad thoughts about getting one tomorrow without waiting for Big Navi. I have come very close.
I saw a claim that Nvidia tried to bluff TSMC for a lower price. They did the arrogant, "we are Nvidia and we will go elsewhere if you don't give us a better price than all your other customers". TSMC basically said "knock yourself out, we can sell all the 7nm wafers we can make". This basically meant Nvidia could not get the 7nm wafers for Ampere and had no choice but to use the inferior Samsung 8nm process.
If it's true then fair play to TSMC not giving in to the bullyboy tactics. It left Nvidia on an inferior node with clear power issues. One thing is 100% accurate and factual... Ampere is on Samsung 8nm. The question is "why"?
- Either Nvidia decided to up their margins by using a cheaper and inferior process Nvidia thought "profit"
- Nvidia felt they had such a lead over AMD they could afford to cheap out and go for a cheaper and inferior node Nvidia thought "profit"
- Or Nvidia tried to strong-arm TSMC for a preferential wafer price but TSMC declined their "generous offer" because Nvidia thought "profit"
What was funny is the AdoredTV "Overvolted" form a few weeks ago where they laughed and declared that Amepre was 100% on TSMC 7nm because "why would they use the crap Samsung process".
Last edited:
Or TSMC is a bad choice, altogether, and AMD is screwed
"
- Pricing. Samsung has the best wafer pricing the industry has ever seen. Being the largest memory manufacturer does have its advantages and wafer pricing is one of them."
I don't think so because Samsung with its low pricing enables the $700 RTX 3080.
With TSMC it could have cost north of $1000, easily.
Last edited:
- Status