Did I just downgrade to a 3070?

[porina]

 

 

Ok, annoying title in place. I would write some background here but I'm sure everyone will skip to the results first anyway so might as well do that now. 

 

In the green corner we have a 2080 Ti, specifically an EVGA Black Edition. In the other green corner, we have the 3070, this time MSI Gaming Trio X.

 

  

 

In my quick testing I picked a small sample of graphical benchmarks at 3840x2160, as well as two related projects at PrimeGrid, called GFN16 and GFN17. More on that later.

 

Test system:

• Gigabyte Z490 Elite AC
• i5-10600k
• Noctua D15
• Kingston HyperX RGB 4000@3600 2x8GB
• EVGA Supernova G2L 850W
• Corsair 230T
• Crucial P1 1TB + MX500 1TB
• LG OLED55B9PLA

 

"4k" Gaming

 

 

For gaming benchmarks I only ran two titles: Shadow Of The Tomb Raider and Final Fantasy XIV: Shadowbringers. Both are available as a stand alone benchmark that is (mostly) easy to run.

 

SOTTR was tested with "Highest" preset, DX12 enabled, exclusive fullscreen, and testing was repeated twice with and without DLSS.

FFXIV: Shadowbringers was tested with "Maximum" preset and exclusive fullscreen.

 

In SOTTR the 3070 produces a lower fps than the 2080Ti, by 6% with DLSS off, and 7% with DLSS on. FFXIV:ShB the 3070 is only 0.4% slower, so we can call that effectively unchanged. The difference in SOTTR is not insignificant in that it is clearly measurable, but at the same time it isn't felt in use.

 

 

3DMark

 

 

I decided to run only Time Spy and Port Royal here, as representing the newer technologies. Time Spy was their DX12 benchmark, and here we also see the 3070 score slightly below the 2080 Ti in the ball park of 2% for overall and graphics score. The CPU score was essentially unchanged at -0.6%. Port Royal is their ray tracing benchmark. Here we see the 3070 4.4% down on the 2080 Ti.

 

2080 Ti Time Spy http://www.3dmark.com/spy/18570958

2080 Ti Port Royal http://www.3dmark.com/pr/902246

3070 Time Spy http://www.3dmark.com/spy/18587862

3070 Port Royal http://www.3dmark.com/pr/903919

 

PrimeGrid

 

 

PrimeGrid is a project run on BOINC that, as it names hints towards, is looking for prime numbers. For my personal interest, I'm mostly interested in finding prime numbers big enough to be reported to the top 5000 list. For GPU projects, the best chance of finding such primes are in the GFN16 and GFN17 subprojects. These look for prime numbers of a specific format and there is efficient GPU code to run to find it faster than CPUs. For the purposes of this forum, GFN17 numbers are bigger than GFN16, so they take longer to run. It also means they require more work to do, which can make it easier to scale on a GPU. We possibly see this effect happening here. For GFN16 the 3070 takes 11.5% more time, but for GFN17 it is just over 1% faster. To prove this, I could also have tested GFN18, which you can guess, is bigger again. But for now my motivation doesn't stretch that far.

 

I also took GPU-Z sensor screenshots to give a snapshot of what was happening while the tasks were running.

 

GFN16

 

  

 

It is interesting that the 2080 Ti seems to be more heavily loaded. So whatever differences there are between the two generations and models, this particular code seems to perform better on the 2080 Ti. However, the 3070 while only a bit slower, take about 1/3 less power than the 2080 Ti. So it is more efficient overall in this running state.

 

Note advanced users running this project for maximum throughput will modify client settings so that two tasks are run at the same time on one GPU. This is able to use up the remaining execution potential and increase throughput, at the cost that each task is individually slower.

 

 

GFN17

 

  

 

GPU usage here is higher and the two cards perform more similarly, both in performance and power consumption.

 

 

 

So Porina, we knew 3070 was 2080 Ti performance class since launch, did you expect it to be faster?

 

No, and that isn't the reason I got it! I just thought if Linus can do bad titles, so can I. The real reason is to enable beyond 60 Hz at 4k on my LG B9 OLED TV. The TV supports HDMI 2.1 which allows 4k120. The 2080 Ti, being an older card, only supports HDMI 2.0 and 4k60. Some report being able to overclock it a bit beyond that, but I had no luck myself. I actually wanted a 3080, but those are still a higher grade of unobtainium and will have to wait. A 3070 is a sidegrade but was still interesting to get the higher refresh rate.

 

Ok, so how did that go? Badly. I literally dropped the 3070 in place of the 2080 Ti in the system next to the TV, reinstalled the driver to be safe and... it worked! I set 120 Hz in SOTTR and it gave it to me. I didn't run it at the time while I fiddled with settings, and in the process I dropped out of 120 Hz. No problem, I'll go back there. TV says no. But it was working a moment ago? Ok, the game also offered me 100 Hz, that worked. Maybe 120 Hz was a bit borderline for my cable. Well, so was 100 Hz, as that dropped out after a while. I was back to 60 Hz.

 

I was prepared, because I went through the same pain when I originally got the TV. My random HDMI cables were apparently no good for 4k60. Some of them seemed to work for a time but picture could drop out at any time. Not good. I bought a certified high speed cable, and all was good. So when I was looking at the 3070, I thought, do I need another cable upgrade? Short answer was yes, there was now a newer "ultra" HDMI certification to support 4k120. I ordered one of those separately from the 3070, and it arrived not long after I had the above problems with the old cable. Simply replace the cable and I was basking in 4k120 glory!

 

I then did some more testing. First up was a quick return to the SOTTR benchmark. It looks better when running above 60 Hz as tearing is now gone. Even before, below 60 Hz was ok since that was in G-sync range. How about some real games? I went into Yakuza Like a Dragon. Now this wasn't going to be the best test for beyond 60 fps testing as even the 2080 Ti didn't really manage that at 4k Ultra settings. On the 3070, depending on where I was I was getting between 50fps and 65fps. I think it was like for like slightly lower than the 2080 Ti, but in less graphically intensive areas it did manage to break above 60fps. So, this isn't the best demonstration for it. Ok, into FFXIV then, which is one of my major timesinks. The benchmark previously gave me 77fps average. In game, it varied more. In housing and city areas I was getting as low as 65fps, but it did go up to high 70's and even occasionally 80's when I went into an instanced area. This game doesn't really need high fps, but going over 60fps still helps it a bit with the sweet spot, in my opinion, in the 70's. On my desktop system with 144 Hz display, I force 1/2 rate limit (72fps) since that is all I need and it helps to keep things running cooler than if I let it run flat out.

 

A quick comment on the LG OLED TVs. When the 30 series first came out there was an incompatibility leading to problems when G-sync was used. For a gaming TV, that was serious. They quickly released a firmware update for the newer CX series TV which resolves that. But I was on last years B series. When would that get the firmware? I kept checking but I never saw the G-sync fix mention in the firmware update notes. Someone on this forum, sorry, I don't recall who, said the fix was in. Ok, the firmware updates do give a vague "minor bug fixes" description. Was it hidden in there? I guess so. At least in my limited time so far, I have not experienced any problems with running G-sync and the TV on latest firmware (at time of writing).

 

Overall, it is doing what I expected of it. I'd like a bit more GPU to help with getting the 4k framerates up, but for now it provides the feature add I was looking for. Performance can wait. I didn't mention noise, this 3070 is so much better in that respect. It does have a beefy 3 fan cooler that barely fits inside the case as it rests on the 2.5" drive bays. The GPU came with an anti-sag bar, but no need for that if the case is doing it for me. The 2080 Ti had a basic two fan cooler. It deserves more. I bought it used and the previous owner ran it under water. After hearing it, I'm almost tempted to go that way myself. Anyway, I now have a "spare" 2080 Ti I'm debating what to do with...

 

 

[Fasauceome]

If all else fails, the 3070 can be used to upgrade your bank account

[porina]

1 minute ago, Fasauceome said:

If all else fails, the 3070 can be used to upgrade your bank account

The 2080 Ti can also be similarly used  Used price for it now is more than I paid for it.

[Moonzy]

4 minutes ago, porina said:

Anyway, I now have a "spare" 2080 Ti I'm debating what to do with...

sell it for the price of a future 3080 (or even 3090, for my location)

[TetraSky]

7 minutes ago, porina said:

I just thought if Linus can do bad titles, so can I.

Oh, prepare yourself to the onslaught of people coming over and telling you "yeah no shit, you should've looked at benchmarks, the 3070 is basically a 2080 ti", for that title without reading further (to be fair, it is a massive wall of text... like jeez...)

 

Good to know the 3070 fits your purpose and it needed a new cable for it to work right. Just sell the 2080 ti, you might be able to even get a profit out of it from what you initially paid for it depending on the state of the used market and GPU availability in your region... It's ridiculous.

[porina]

5 minutes ago, TetraSky said:

Just sell the 2080 ti, you might be able to even get a profit out of it from what you initially paid for it... It's ridiculous.

I bought it used for £620 with both original cooler and also an EK waterblock at end of last year when it became obvious my chances of getting a 3080 any time soon were slim. If I look on ebay now, the same model (without waterblock) is going for anywhere between £650 and £999. Basically I could probably sell it now without waterblock and get at least my money back on it.

 

I am half debating if I want to keep it though and upgrade a different system. The 3070 is in my TV gaming system. I'm running a 2070 in my desktop gaming system, and a 1080 Ti in my streaming system. #firstworldproblems

[DildorTheDecent]

31 minutes ago, porina said:

Anyway, I now have a "spare" 2080 Ti I'm debating what to do with...

Since it's a non-A card, shunt mod it and send it to the moon.

 

Or just settle for the Palit 310W BIOS. It's not that bad tbf.

[porina]

5 minutes ago, DildorTheDecent said:

Since it's a non-A card, shunt mod it and send it to the moon.

 

Or just settle for the Palit 310W BIOS. It's not that bad tbf.

If you're saying overclock it, pass. The air cooler is barely adequate as it is, and I'm not looking to put the waterblock back on. I'll either reuse it as is, or sell it at some profit keeping in mind I got it used myself.

[TVwazhere]

46 minutes ago, porina said:

Anyway, I now have a "spare" 2080 Ti I'm debating what to do with...

[porina]

1 minute ago, TVwazhere said:

Hmm... now, what is that shape doing? Bending? No, folding in on itself. I see what you did there. Mining it is!  

[TVwazhere]

3 minutes ago, porina said:

I see what you did there. Mining it is!  

Gamers be like: