• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

RTX 2070 Gigabyte Windforce

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.


Mar 7, 2008
I got the card in subject as it was one of the low cost Turing cards, available for the same price as a low end 1080. This was £460 with tax, equivalent to $500 without although price conversions don't always work that simply.

I've run some 3DMark family benches on it, with a 6700k at 4.7 GHz so I'm never going to get the highest CPU score. Of course, I also had to overclock it.

Going up core +50 at a time in Afterburner, +100 was ok in FireStrike, but 150 wasn't. Ok, let's stick to 100 then...

Next was mem. I thought this would take forever with FireStrike, so instead I used Heaven. I went up... and up... and up... +925 was the highest I saw that didn't crash, next step to 1000 immediately crashed. I was kinda hoping for some artefacts to tell me to back off. I rounded down to 800, and grabbed a FireStrike run for hwbot.

Ok, FireStrike Ultra next. Blank screen crash. Ok... what is wrong? Dropped mem to +600, still no go. Locked up loading this time. Dropped core to +75, got a run in. Too lazy to put ram back up for rerun so left it for the sub. Repeat on FireStrike Extreme with increased ram again. As I was using my Win7 bench install, no Time Spy for now.

Since I had bought it previously, I did a quick 3DMark11 Performance run too. With that I decided it was bed time and will do VRMark tomorrow.

To translate this into GPU-Z reported MHz, my nominal boost went up 150 MHz to 1770 (at +75 setting in AB). My mem went up 200 to 1950 MHz (+800 setting in AB). During one run of FireStrike, GPU-Z did report a max core clock of over 2000 though. There may be some optimisation possible to fine tune the lower power states, if I can work out how.

I did some tests with FAHBench too (default settings).

135 1080Ti Asus OC
111 2070 Gigabyte Windforce
74 1070 Zotac FE
73 980Ti Gigabyte Windforce
57 RX 580 Gigabyte Windforce

So... it is a good amount over a 1070, and getting close to a 1080Ti. Not bad... I also need to grab power measurements to work out efficiency compared to other cards. For reasons I don't understand yet, reducing power limit actually increased power usage and performance in FAHBench.

The overclock scanner in Gigabyte's software crashed and I didn't bother looking at it again.

Temps were fine throughout the FireStrike and FAHBench testing, going into the 60's. Don't think that's a limiting factor. I didn't monitor it during the other benches.
I'm also receiving this Gigabyte card. Maybe even today, if post delivers fast enough.

I think interesting thing about this is, that it is made to RTX 2080 pcb (source https://www.hardwareluxx.de/index.p...te-geforce-rtx-2070-windforce-8g-im-test.html review).
Instead of other entry level card with 6+2 power phases, this has 8+2 phases. Also 6+8 pin power connectors when FE and many other has only one 8 pin.

So plenty of power resources from pcb side, but artificial power limit from the bios. Isn't the maximum setting +14% -> 210W?
Question is, how to get past of that. Maybe Gigabyte is nice enough to make custom bios with increased power limit, like they have done to 2080 and 2080ti models.

I'm also thinking that 2080 waterblocks might directly fit to this. I'm possible heading to watercooling, especially if higher power limit options comes available with way or other.
I can't remember the exact number but the power limit max is between +10-19%. Do we just need the same old tricks as lowering the resistance of the sense resistors? I think without a voltage mod, this wouldn't be too useful anyway. Like said, on my card I hardly got any more on core using afterburner.
I started overclocking with Gigabyte's own software. Ended up 1840Mhz (+220) setting for core, which showed steady 2055Mhz boost clock value in Black Ops4.
It seems that maybe power limit max value (+14%) isn't that big problem as I feared. I think core can't do much more as +230 / 2070Mhz boost setting crashed after half hour.
While playing, +14% power limit was enough, but on TimeSpy benchmark there was need for more. It was limiting boost to ~2000Mhz.

I'm also very happy about temperatures /noice levels. Fan speed 60% or temperature 68 degrees celsius were maximums i saw while overclocked.
This is big update for me as previous card was GTX680 2GB.
That's interesting... I was seeing actual boosts >2000 even at stock., although the nominal 1st page of GPU-Z was showing a much lower number. I haven't gone back to test it further yet, still on to do list.
That's interesting... I was seeing actual boosts >2000 even at stock., although the nominal 1st page of GPU-Z was showing a much lower number. I haven't gone back to test it further yet, still on to do list.
So, you got bit over 2000 boost values stock and about same when overclocked +75? What was the app used for testing?
I got only about 1850 stock with TimeSpy. In my case max boost value followed overclock value as expected. 1850 stock+220 overclock =2070

Today I will test Afterburner (eg. auto OC) and also how much I can increase memory clock. Gigabyte OC software felt somewhat buggy.
I didn't see auto overclock in Afterburner, and I had latest version at the time I started this thread. Haven't looked since. The scanner in Gigabyte's software crashed the system the one time I tried it, and I haven't tried it again.

When I saw the 2000+ boost was during FireStrike according to my notes in the 1st post. I didn't make detailed notes so can't say beyond that with any certainty.

In case it makes a difference, I am running a benching install of Win7. Haven't tried it in anger on Win10 yet, so still no Time Spy.
I'll let someone else test that first :) I wonder if the 2070 limits are due to the reference board only having a single power connector. I've heard the lower end models are not allowed to be factory OC, and presumably have the same limits as reference boards even if this model has dual connectors.
OK mack this thread is getting boring, where are all the benchmarks?

I'm finally taking my summer vacation from tomorrow, yes, I'm that far behind. I'll be back home later next week and will have another go at it then.

BTW it isn't just me, hwbot is nearly dead of submissions. http://hwbot.org/hardware/videocard/geforce_rtx_2070/
Is everyone playing with the 2080Ti instead? I think I can get a load of gold cups if I just run things. Surprised no one else has submitted even a FireStrike run.

Really should dust off the 5820k for some more CPU points than 6700k in the 3DMark family tests.
Damn, that is some awful background colour Mack, ha ha
Seriously though the 8086k will likely give better CPU score, newer and likely clocks higher
Damn, that is some awful background colour Mack, ha ha
Seriously though the 8086k will likely give better CPU score, newer and likely clocks higher
I haven't checked the exact shade, but it is one of the built in Windows ones and seems to resemble nvidia green :) I just pick a different solid colour on each system to help me tell them apart. Normally I'd use an image, but it makes the file sizes for screenshots go up a LOT.

The 8086k is also my main gaming system. I'm not going to pull it apart every time I bench.
You keep up that crazy talk and I'm going to pull your banner. ha ha

I'm looking to get a 9800X or higher for my test system... :D Probably will have similar issues as seen with 7000X series but moar cores!
I'm looking to get a 9800X or higher for my test system... :D Probably will have similar issues as seen with 7000X series but moar cores!

You are losing a lot because of that CPU score in 3DMarks. Another thing is that 8086K will probably give better CPU/Physics score than Skylake-X at the same or +2 cores. For some reason SL-X is performing not so good in some tests and I guess it's related to redesigned cache. It gets better once CPU has many more threads but this will cost a lot. 7900X results were acceptable for me but CPU/physics score was worse than on 8 core chips from previous generation.
I know you want X series because of some other tests and we were talking about it in some other threads as I remember ;)

Prices on 9900K went down (like by 25% in last 2 weeks) so I ordered one. I have my theory about this CPU vs SL-X and I guess I will check it soon.

Hwbot is quite dead recently. People are pushing the most popular/top series and skip these lower, that won't give high global points or won't count for the best results in competitions. The same is in i7-9700K category or some others.
You can expect that some reviewers will drop results soon but again, some won't bother with these cards and will focus on 2080/2080Ti.
What actual pricing are you seeing? There is only slight movement in UK, 5-10% drops to around £520 if you don't care where you get it from. The 9700k is more interesting, sub £400 now so not much different to what I paid for the 8086k, but the question will then be if 8c8t or 6c12t will do better in a given bench. This weekend I'll be expanding my data on Broadwell vs Haswell now that it is running, and time allowing I might get back to playing with the 2070 as the folding month I'm participating in will be over then.

Edit: if we take for now that my 8086k is in my gaming system and I'm really not going to pull it apart every time I want to bench something, which of my other CPUs would be best for 3DMark GPU tests?
Choice of:
7800X (up to 4.8 GHz, although I'm about to slap a D15 on it instead of water and don't know if that'll keep cool)
5820k (not a great clocker, think I struggled to around 4.5 on my sample)
Others are the usual quad cores, assume they won't be as good as these 6 cores?

One other alternative, I could see if there are any cheap Haswell-E or Broadwell-E 8+ cores around?

Edit 2: Had a poke around ebay and ended up buying a 5930k. Ok, it doesn't buy me any more cores over 5820k but it might help if I want to revisit SLI/Crossfire in the future. Also get another roll of the dice in the silicon lottery as to OC potential.
Last edited: