NVIDIA ‘Ampere’ 8nm Graphics Cards

[Sgt_Pepper]

Was just looking at cards on here, cheapest 2080ti Is £1100, if i can get something that matches the performance for £700-£800 i will buy it , nvidia or AMD. Otherwise i will also be considering finding another hobby.

 

[Kelt]

Probably not too far out with that.

They won't be getting my money though.

Same here.

My dosh'll be going towards a new bike.

 

[Richdog]

Indeed, though I can't find any figures on sales numbers to support either "poor sales" or "bucket loads". I can only find hits on sales revenue increase, which is to be expected considering the big increase price.

Here are some sales figures from January 2020... just Germany but still a healthy gaming market. https://www.reddit.com/r/hardware/comments/f46pt1/mindfactory_graphics_cards_sales_report_january/

Nvidia turned a significant profit with Turing GPU's. https://www.forbes.com/sites/greatspeculations/2020/02/25/nvidia-has-a-gpu-business-and-its-big/

Reveue figures speak for themselves (includes all their business) https://www.macrotrends.net/stocks/charts/NVDA/nvidia/revenue

 

[JediFragger]

Yeah, the 3070 at the very least has to be semi-reasonable. Unless they actually do launch a Ti version next to it and whack another $100 on top!

 

[MissChief]

On TVs. Perhaps @MissChief can enlighten us as to what percentage of Sky subscribers subscribe to 4k content?



Aren't they targeting 30 fps?

I have absolutely no idea I’m afraid. There are around 5.5 million Q subscribers though.

 

[FoxEye]

Ok here we are.

Its pretty obvious which generation is the odd one out here. Remove the 1st release of Turing and you get a really consistent set of steps between generations and you get a consistent set of prices between generations.

The 2080 Ti is clearly an odd one out here, almost, I would say, to the extent that it does not belong in this comparison at all.

The super range is clearly where the 1st release should have been, and the equivalent card to the 1080 Ti is not the 2080 Ti, but the 2080 Super.

Look at the 960 -> 1060 -> 2060 S and you can see the xx60 card has *doubled* in $price over 2.5 generations.

That's not really consistent pricing tbh.

 

[danlightbulb]

Look at the 960 -> 1060 -> 2060 S and you can see the xx60 card has *doubled* in $price over 2.5 generations.

That's not really consistent pricing tbh.

Yes but that's not the upgrade path at that price point.

Following the c.$200 banding (+/-$50) you would go 760 --> 960 --> 1060 --> 1660S. The 1060 was actually a bit expensive, and the 1660S has come back down from there.

The stack changes means that the 2060S is now the upgrade path from the 1070 along the $400 price band.

 

[narta]

So the 2080ti to hold its value and price like the 1080ti did? Another pathetic release?

Not really. 2080ti prices in China are ~$1900, there is a premium about 30% for items like gpu. So 3090 is looking at 2080ti launch price

If the 3080 hits near £1k as expected then I think you'll still get 450/500 for the 2080Ti tbf.

No way! 3080 will be at £799 directly replacing the 2080s price segment. 3070 is looking something like £499 with 2080ti perf in ratser and 2x-3x in RT. 2080ti is still a good card but when Ampere will launch second hand market price will tumble for sure

 

[LeMson]

Look at the 960 -> 1060 -> 2060 S and you can see the xx60 card has *doubled* in $price over 2.5 generations.

That's not really consistent pricing tbh.

It's more that the 70 class card is now a 60, the 80 is now 70, ti is now 80. if viewed from a what can I get for £350, or £500, or £750.

So your 960, 1060 is now a 2060( effectively a 2050), 60 becoming a 50 class

 

[danlightbulb]

It's more that the 70 class card is now a 60, the 80 is now 70, ti is now 80. if viewed from a what can I get for £350, or £500, or £750

And if the performance jumps are still there, then naming doesn't really matter. The jumps were NOT there in Turing 1, but they fixed it for the super's.

 

[Illuminist]

There were like 2 people in the entire UK who bought the Ultra.

To quote Anandtech.

"We are all for higher performance, and we don't mind higher prices. But it is ridiculous to charge an exorbitant amount of money for something that doesn't offer any benefit over a product already on the market. $830 isn't the issue. In fact, we would love to see a graphics card worth $830. The 8800 Ultra just isn't it."

Had 2 bfg editions in crossfire...

 

[Journey]

Had 2 bfg editions in crossfire...

Well some poor sucker had to be the one.

 

[FoxEye]

Yes but that's not the upgrade path at that price point.

Following the c.$200 banding (+/-$50) you would go 760 --> 960 --> 1060 --> 1660S. The 1060 was actually a bit expensive, and the 1660S has come back down from there.

The stack changes means that the 2060S is now the upgrade path from the 1070 along the $400 price band.

960 -> 1060 was in your graph is a bit more than +4000 score
1060 -> 1660 S is a bit more than +2000.

1060 -> 2060 S is about +6000

1060 -> 1660 S might be "an upgrade in it's price segment" but it's an incredibly mean upgrade. And not one many will be interested in.

How do you think people are going to feel if you tell them that "the upgrade for the 2080 is the 3070. For the same money you get +10% perf! It's an upgrade!"

 

[Illuminist]

Well some poor sucker had to be the one.

It was at that time of life when you can start to get loans and are in higher education, I was a fool

Q6600 I think back then whole thing was 3k...eek

 

[danlightbulb]

960 -> 1060 was in your graph is a bit more than +4000 score
1060 -> 1660 S is a bit more than +2000.

1060 -> 2060 S is about +6000

1060 -> 1660 S might be "an upgrade in it's price segment" but it's an incredibly mean upgrade. And not one many will be interested in.

How do you think people are going to feel if you tell them that "the upgrade for the 2080 is the 3070. For the same money you get +10% perf! It's an upgrade!"

I think what's clear from these charts is its far better to upgrade every 2nd generation.

If you went 770 to 1070, or 970 to 2060S, in both cases you're nearly doubling performance for around the same price.

Same if you go from 780 to 1080, or 980 to 2070S. And same if you go 760 to 1060, or 960 to 1660S.

It also works up in the Ti's until you get to the latest gen then it fails, but that's because of where they stuck the 2080 Ti. I think what happened was they weren't willing to undermine the 2080 Ti and so they gimped the 2080 S when it came out in the 2nd Turing release. That would explain why the pattern breaks down in that specific spot.

 

[Things change I've changed]

Put me down for a fiver on

3090 £2000-2500
3080ti £1500-2000
3080 £1000
3070 £750

If these prices are anywhere close to reality then I'll be staying a generation behind and probably going down secondhand route on principle.

 

[MissChief]

If these prices are anywhere close to reality then I'll be staying a generation behind and probably going down secondhand route on principle.

Have to say, as a 2070 Super owner I’d be skipping the next gen as well if those prices are anywhere close to being accurate. I don’t fancy £750 for a 20-25% boost. I’d rather go to a 2080 Super for half that second hand if I do need something faster. (Although looking at benchmarks there’s little in it.)

 

[Quartz]

I have absolutely no idea I’m afraid. There are around 5.5 million Q subscribers though.

Sky Q is 4k isn't it? So there are around 5 million households in the UK (I'm allowing part of that 5.5 million to be commercial premises) that have 4k displays. There are 23 million dwellings in the UK. And not everyone with a 4k TV will have Sky Q. So that's between 1 in 4 and 1 in 5 households with Sky Q. Add in Netflix, Amazon, etc, and the figure is probably more like 1 in 3.

 

[danlightbulb]

Have to say, as a 2070 Super owner I’d be skipping the next gen as well if those prices are anywhere close to being accurate. I don’t fancy £750 for a 20-25% boost. I’d rather go to a 2080 Super for half that second hand if I do need something faster.

2070S --> 2080S is only a marginal upgrade because the 2080S didn't have the expected performance jump that it should have done. The 2070S did though, and is IMO the best value card in Nvidia's current line up, clear as day. Therefore it wouldn't/shouldn't be that surprising if you find you don't have a value for money upgrade path in the next series. Having now studied these upgrade paths, I don't necessarily think that is a bad thing. Clearly upgrading every two generations offers far more value, if coming at it from an entirely value for money perspective. Of course, if you choose to upgrade every gen, that's your choice and your money and you're paying a premium for it.

 

[FoxEye]

I think what's clear from these charts is its far better to upgrade every 2nd generation.

If you went 770 to 1070, or 970 to 2060S, in both cases you're nearly doubling performance for around the same price.

Same if you go from 780 to 1080, or 980 to 2070S. And same if you go 760 to 1060, or 960 to 1660S.

It also works up in the Ti's until you get to the latest gen then it fails, but that's because of where they stuck the 2080 Ti. I think what happened was they weren't willing to undermine the 2080 Ti and so they gimped the 2080 S when it came out in the 2nd Turing release. That would explain why the pattern breaks down in that specific spot.

If Turing becomes the "new normal", then the upgrade for the 2080 (at the same price point in two gens time) will be the 4060, offering +10% performance. Or the 5050, at the same price point, offering +10% performance

If rumours are true, Turing isn't an anomaly, but the new gold standard in price hikes.