• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."

nVidia 3080 reviews thread

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Konan

Senior member
Jul 28, 2017
299
202
86
As expected, the "double 2080 performance" was utter nonsense.
I've seen a few places / Vids mentioning 90%. i.e 2080 vs 3080 same prices 90% better....
But then most seemingly is 70% so yeah not double but still quite good. Plus no FE tax and $100 less in fact.
 
Last edited:
  • Like
Reactions: Tlh97 and guidryp

Wall Street

Senior member
Mar 28, 2012
691
44
91
It just doesn't seem like it's doing as well as it should be though. The # of CUDA cores were doubled on the 3080 from the 2080 TI, which only had an about a 1.2x increase from the 1080 TI, yet we only see about 30% performance gains.
The best way to explain the new cores is that Turing had separate FP32-only and INT32-only cores while Ampere has FP32-only cores and cores which can do either INT32 or FP32 (lets say hybrid). Although this is an oversimplification, It is a little more complicated than this (Turing INT cores can do a very limited subset of FP instructions).

When measuring FP cores or TeraFLOPS in Turing, you only count the FP32 cores. However, when INT workloads come along you have those INT cores available which don't impact the FP32 throughput.

In Ampere, the FP core count and TFLOP count looks at both the FP32-only cores and the hybrid cores. When an INT workload comes along (which can be 20-30% of the work in a game) then one of the cores which is FP32 capable and counted in the FP32 TFLOPs has to stop doing FP work to do the INT work.
 
Last edited:

AtenRa

Lifer
Feb 2, 2009
13,466
2,324
126
Again if we normalise prices like they were before the 2000 series, the RTX 3080 is about 30% faster than RTX 2080ti, which is in line with GTX 1080ti over GTX 980ti and GTX 980ti over GTX 780ti.
GTX980Ti over GTX780Ti was 40% at 1440p/4K

GTX1080Ti over GTX980ti was 75% at 1440p/4K

RTX3080 over RTX2080Ti is just 30% at 4K

This is the worst new gen on a new node after Fermi.
 

AtenRa

Lifer
Feb 2, 2009
13,466
2,324
126
It's not normal. Pascal was an outlier among the best ever releases, this is close to matching Pascal. Putting them both near the pinnacle of generational increases. The last time anything approached this level of gains before Pascal was the Mighty 8800 GTX.

Also it's not a 30% generation upgrade. Generation is NOT 3080 vs 2080 Ti.

Generation would be 3080 vs 2080, and these gains are typically around 70% at 4K. A very Pascal type gain, and very much beyond the typical release.
RTX2080Ti is using the TU102 chip
RTX2080 is using the TU104 chip

RTX3080 is using the GA102 chip
RTX3070 is using the GA104 chip

It is a marketing trick they used and named the card RTX3080 and not RTX3080Ti.
 

davidthemaster3

Senior member
Mar 11, 2011
200
3
81
RTX2080Ti is using the TU102 chip
RTX2080 is using the TU104 chip

RTX3080 is using the GA102 chip
RTX3070 is using the GA104 chip

It is a marketing trick they used and named the card RTX3080 and not RTX3080Ti.
But the RTX 3080 is using a cut down GA102. The unreleased RTX 3090 is the full GA102 chip you're looking for to have a valid RTX 2080 Ti comparaison.
 
  • Like
Reactions: psolord and xpea

crisium

Platinum Member
Aug 19, 2001
2,634
591
136
3080 is more like the 780 - heavily cut big chip. Or 560 Ti Core 448. There is no parallel to Turing, Pascal, or Maxwell.
 

Glo.

Diamond Member
Apr 25, 2015
4,285
2,362
136
Yeah, Nvidia's marketing did miracles here.

RTX 3080 is replacing RTX 2080 Ti, and 3090 is the new Titan. Its a marketing masterpiece, if you ask me. Mask the smallest generational uplift in performance, in history, from Die-tier, and smallest effciency uplift, with "lower" SKUs, or by creating brand new SKUs.

If you think this way, then the new RTX XX80 Ti SKU(RTX 3080) has been just price cut from 999-1200$ to 700$.
 

IntelUser2000

Elite Member
Oct 14, 2003
7,095
1,628
136
Only FP32 are doubled, and games use FP32 and Int both. It is misleading for games. They have equal SMs.
The best way to explain the new cores is that Turing had separate FP32-only and INT32-only cores while Ampere has FP32-only cores and cores which can do either INT32 or FP32 (lets say hybrid).
That's not it. From the Nvidia themselves, for every 100 FP instructions, there are 36 Int instructions.

The TMUs is the same and the ROPs increase is only 10%. Perhaps that's a bottleneck.
This is it. And memory bandwidth did not double either. Actually its only a 23% increase from the RTX 2080 Ti.

TMU/ROP gets boosted by 10%, memory bandwidth by 23%, yet its due to Integer instructions. Even if FP and Integer instructions were equally important you'd only see 25-30% difference.

Riiight.
 
  • Like
Reactions: Tlh97

Aikouka

Lifer
Nov 27, 2001
29,408
426
126
So, I still use a 2560x1440 (WQHD) @ 144Hz monitor. I know it may seem a bit odd to even consider one of these cards given the declaration that they're meant for 4K, but it's really seeing if the card does a better job at ensuring 100+ FPS at WQHD. I was looking at TechPowerUp's review, and it seems like some games do show a decent improvement (2080 Ti vs 3080). For example, Hitman goes from 100 FPS, which is fairly acceptable, to 125 FPS. Although, the only game that I saw that showed real "you should upgrade to this" numbers was Control, which went from 68 FPS to 93 FPS.

Although, I guess one aspect that I may need to find specific reviews on is VR performance. Your basic games like Beat Saber don't require much, but if you start getting into the graphical games or AAA games with VR modes, it can be a lot more demanding.

As for my plans... I'm not sure. I have to consider that any card purchase would include an extra $150+ for a waterblock and backplate. At least based upon TechPowerUp's numbers, it looks like the card ramps up to 1965MHz and then just drops down to around 1815MHz. I'm not sure if that's simply for thermals or because the 1815MHz is a proper figure for the voltage.
 

DJinPrime

Member
Sep 9, 2020
26
18
36
RTX2080Ti is using the TU102 chip
RTX2080 is using the TU104 chip

RTX3080 is using the GA102 chip
RTX3070 is using the GA104 chip

It is a marketing trick they used and named the card RTX3080 and not RTX3080Ti.
That's not logical. So, when the 3080 ti do come out, then what will it compare to? And you know they can release a 3080 ti, because the 3090 shows there's a bigger ampere. Who cares about the code name? Would it have made you happier if they just edged it GA104 on the lid? It's priced like a tradition 80 card and there's even a bigger card. So, no this is not the ampere ti.
 

guidryp

Senior member
Apr 3, 2006
359
164
116
RTX2080Ti is using the TU102 chip
RTX2080 is using the TU104 chip

RTX3080 is using the GA102 chip
RTX3070 is using the GA104 chip

It is a marketing trick they used and named the card RTX3080 and not RTX3080Ti.
Which chip they are using is irrelevant. 3090 is already cut down from full as much as the 1080 Ti was.

There will almost certainly be 3080 Ti several months in the future, that has ~3090 performance. Then you can make your 2080 Ti vs 3080 Ti comparison. You will get a preview with 3090 review.
 

Wall Street

Senior member
Mar 28, 2012
691
44
91
That's not it. From the Nvidia themselves, for every 100 FP instructions, there are 36 Int instructions.
How do you know this isn't impactful? On Ampere, an INT costs you an FP instruction. Turing had dedicated underused INT cores, so INT instructions didn't cost anything (expect power consumption and bandwidth).
 

Dave2150

Senior member
Jan 20, 2015
596
129
116
You can connect your PC to a 4K OLED TV and play from a distance like I do. It's the best PC gaming experience at this point. I do expect high end PC gaming with top hardware (like the 3090) to become more niche though, with few games that really justify it over cheaper hardware.
That's what I'm doing already, I have CX48 hooked up to my PC. 90cm deep desk, and it's doable. We're in the extreme minority though I believe :D
 

IntelUser2000

Elite Member
Oct 14, 2003
7,095
1,628
136
How do you know this isn't impactful? On Ampere, an INT costs you an FP instruction. Turing had dedicated underused INT cores, so INT instructions didn't cost anything (expect power consumption and bandwidth).
Because a balanced architecture will stress compute, texture throughput and memory bandwidth roughly equally.

Hello? Did people forget about memory bandwidth? ROPs? TMUs?

And did you not read my post? For every 100 FP instructions, there are only 36 Int instructions. That's a performance impact of 5-7%.
 

JustMe21

Senior member
Sep 8, 2011
309
35
91
Are there any reviews that show DLSS on only or is it only used when you have RTX on as well? So, far, TechPowerUp shows RTX on, DLSS+RTX, and both on, and RTX off all on the same chart with RTX off having higher FPS, so I keep wondering why I would ever use DLSS+RTX or RTX only competitively in an FPS, when there is added overhead, unless you truly do gain some benefit from having RTX on, such as being able to see an enemy shadow more easily via RTX than via software rendering.
 
Last edited:

Ajay

Diamond Member
Jan 8, 2001
7,260
2,491
136
That's what I'm doing already, I have CX48 hooked up to my PC. 90cm deep desk, and it's doable. We're in the extreme minority though I believe :D
Holy Smokes! I have my 27" (1440p) about 60-70 cm away - you must actually have to move your head to scan the edges of the frame!
 

coercitiv

Diamond Member
Jan 24, 2014
3,858
3,852
136
Macdonal combo meal is $8, pretty sure it was only $6 a few years ago. Have you look at the price of housing? Cell phones? Not to get political, but official inflation number is bs, it's so much more expensive for many things.
I was merely pointing out the massive price difference between the models which signals different product tiers to begin with. But if you want to argue they belong in the same bracket as they are tied by inflation, you better be ready to argue that wages have doubled since 2017.
 
  • Like
Reactions: Tlh97

eek2121

Senior member
Aug 2, 2005
654
596
136
Macdonal combo meal is $8, pretty sure it was only $6 a few years ago. Have you look at the price of housing? Cell phones? Not to get political, but official inflation number is bs, it's so much more expensive for many things.
That dollar menu shows items that cost more than a dollar.
 

ASK THE COMMUNITY