Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,873
Nah they've knocked it on the head completely on the lower end stuff (when the entire stack is evaluated ofc, gaming gfx cards are low-end volume biz).
-
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 4000 Series
- Thread starter MrClippy
- Start date
Caporegime
- Joined
- 18 Oct 2002
- Posts
- 30,873
PCIe 4.0 and DP 1.4a ports, livin' the dream!!!
I have been wondering what kind of performance uplift we'll see from a 4080 16GB vs my current 3080Ti
It is wrong and its actally a 4060ti.
A 3090 Cuda = 10240-53%=4812 a 3060ti has 4864
Now take a 4090 with 16384-53%=7700 a 4080 12 has 7680
Soldato
I wrote: "I don't think anyone made the 4070 accusation based on what Nvidia puts in columns next each other."You obviously not read many posts on this forum and other forums, websites, reddit or watching youtube etc.
Yes people wrongfully made overpriced $899 4070 accusations.
Here is Why the NVIDIA RTX 4080 12GB is an Overpriced RTX 4070 Ti | Hardware Times
NVIDIA announced its RTX 40 series GPUs earlier this month to a moderate to lukewarm welcome. The RTX 4090 looks like a proper flagship with the RTX 4080 16GB holding its own, but its 12GB sibling is a bit of an abomination. Featuring the AD103 die, it packs 7,680 shaders (less than the RTX 3080 …www.hardwaretimes.com
Reddit - Dive into anything
www.reddit.com
Reddit - Dive into anything
Why the RTX 4080 12GB feels a lot like a rebranded RTX 4070 | Digital Trends
Nvidia has announced two versions of the RTX 4080. But the more we find out about these two GPUs, the more different they seem.www.digitaltrends.com
The 4080 12GB isn't suspected of being a rebadged 4070 because anyone thought it was listed in the white paper as such. (Maybe 4070 scratched out in crayon and 4080 hand-written in pencil above it?)
The smaller bus-width, and smaller chip size were some of the factors that went into it...but not a white paper.
I personally think all that matters is performance, but we don't know that yet because Nvidia has just given us smoke and mirrors thus far.
It's the price to performance which matters. A 4080 12gb is around 40% more expensive than a 3080 so should offer atleast 40% more performance otherwise you'll be getting less for your money.
Looking at the spec of the card I'd be supprised if it's much more than 10% faster which would again mean it lines up performance wise with a 60ti class as the 3060ti was around 10-15% faster than a 2080 while also being around 40% cheaper.
There's a lot of screeching on both sides.
1. Nvidia has mind-share and free market and Jensen = therefore Nvidia can charge everything and you must pay it because of leather jacket!
2. I'm not paying that, gtfo!
After all the votes and the talk, it'll be interesting to see who does what.
1. Nvidia has mind-share and free market and Jensen = therefore Nvidia can charge everything and you must pay it because of leather jacket!
2. I'm not paying that, gtfo!
After all the votes and the talk, it'll be interesting to see who does what.
Soldato
I wasn't trying to imply that price didn't matter, only that performance matters more than nomenclatures.
Let's ignore the die names, as has been mentioned by others, it doesn't matter that one GPU is called xx102 and another is called xx104. All that matters is the relative size and performance between various model of GPUs throughout the stack, so let's look at that.
* From looking at the data this way, we can see the 4080 16Gb is 59% in both size and performance of the top GPU, the 4090 and the 12GB is 47-49%
Now, looking previous generations, which GPU's roughly fall into this 47% to 59% range?
* From the RTX 3000 series, the RTX3070 does at 56%-57% of the RTX3090
* From the RTX 2000 series, the RTX2060 and RTX2070 does at 46%-54% of the RTX2080ti
* From the RTX 1000 series, the RTX1070 does at 54%-64% of the RTX1080ti
So now, we can see that that the RTX4080 (both versions), fall into the same size and performance category that the prior 3 generations reserved for its xx60 and xx70 series
| SM Count | FP32 Tflop Perf | SM Relative/TOP GPU | Tflops Relative/TOP GPU | |
| RTX4090 | 128 | 82 | ||
| RTX4080 16GB | 76 | 48 | 59% | 59% |
| RTX4080 12GB | 60 | 40 | 47% | 49% |
| RTX3090 | 82 | 35 | ||
| RTX3080 | 68 | 29 | 83% | 83% |
| RTX3070 | 46 | 20 | 56% | 57% |
| RTX3060 | 28 | 12 | 34% | 34% |
| RTX2080ti | 68 | 26 | ||
| RTX2080 | 46 | 20 | 68% | 77% |
| RTX2070 | 36 | 14 | 53% | 54% |
| RTX2060 | 30 | 12 | 44% | 46% |
| GTX1080ti | 28 | 11 | ||
| GTX1080 | 20 | 9 | 71% | 82% |
| GTX1070 | 15 | 7 | 54% | 64% |
| GTX1060 | 10 | 4 | 36% | 36% |
* From looking at the data this way, we can see the 4080 16Gb is 59% in both size and performance of the top GPU, the 4090 and the 12GB is 47-49%
Now, looking previous generations, which GPU's roughly fall into this 47% to 59% range?
* From the RTX 3000 series, the RTX3070 does at 56%-57% of the RTX3090
* From the RTX 2000 series, the RTX2060 and RTX2070 does at 46%-54% of the RTX2080ti
* From the RTX 1000 series, the RTX1070 does at 54%-64% of the RTX1080ti
So now, we can see that that the RTX4080 (both versions), fall into the same size and performance category that the prior 3 generations reserved for its xx60 and xx70 series
Last edited:
Associate
but the above table could also be interpreted as 11tf jump over 3080.. and btw this table is kinda stuck in the past because it looks like nvidia isnt going down the raster route, they have released a whitepaper on ada and raster performance has been relegated to a footnote or maybe not mentioned at all.. nvidia is showing a clear divergence from a raster led product strategy, its a strategic bet.. nvidia taking its chances.. possibly another decade of technology leadership if successful
you can rant about price and performance all you want but die size, tiers and GPU names are entirely irrelevant. as previously said, if you really want to start comparing GPUs then transistor count is a much better starting point
The major flaw in this logic is using the top GPU as some kind of benchmark. If the 4090 was smaller and less performant would that suddenly make the 4080 better value and a "true 4080"?
Because it is the benchmark and always was.
Also, if you compare 4080 vs 3080, you will see that there is almost no improvement gen on gen, thats unprecedented, especially considering this is 1.5 node shrink!(from garbage 7nm class(Samsung 8nm) to top of the line 5nm class(TSMC N4)
Look at Pascal vs Maxwell gains...
From this gen only 4090 is worth anything, rest is just a disappointment.
Why wouldn't you use the top GPU as the benchmark especially when it's priced not much higher?. If the 90 was smaller then the 4080s would look better but not in terms of price as a 4080 12gb is still going to be barely any faster than a 3080 12gb.
Last edited:
So if we make a GPU where 25% of the transistors only get used 5% of the time that's a better starting point than a GPU with 5% fewer transistors that get used all the time?
Last edited:
About 100 cheaper with inferior raytracing capability and no DLSS
AMD hardly did us much in terms of favours.
and the availability sucked.
Yea, AMD really dropped the ball there. They should've priced comparable cards at least £200+ less than Nvidia counterpart if only to demonstrate how much more people are paying for RT. I would guess the majority of people are willing to pay £100 more for a similar card with RT but if you had to pay the equivalent of the next tier up just to get RT i suspect most people would balk at that.
If a 7900 performs roughly the same outside of RT as 4090 but is priced similar to the 4080 that's a much easier sell.
If a 7900 performs roughly the same outside of RT as 4090 but is priced similar to the 4080 that's a much easier sell.
Last edited:
Soldato
- Joined
- 28 Oct 2011
- Posts
- 8,556
I cant believe people start to justify 4080 12gb
They'll justify anything on here.
#Nvidiastrong