Now my vendor is expecting to get my RTX 3070 on 20/1-2021. I have high hopes but are not expecting anything.
I am putting the CPU and USB upgrade and Reverb G2 install on hold until I get the GPU. I have to pull out my old GTX 970 and move the GTX 1070 a slot and then fit in the RTX 3070. I think that it will be best to do all of that in one go.
There is also the dual PC cable jungle behind the machines and the very narrow space to consider. I don’t really feel like messing with that more than once.
Ahh, today i got a notice that my RTX 3070 have shipped from the vendor in Norway.
I wish for just as speedy trip for it as for the Reverb G2 as it was from the same vendor. The Reverb G2 arrived the day after the shipping notice. The vendor is Komplett.
I think there may be a couple of days ahead, where the 3D printer project goes on hold.
So since vendors have premium prices, I decided to act like a premium customer and bought 2 GPUs:
XFX Merc 319 (RX 6800 XT)
Gigabyte Vision OC (RTX 3080)
My testing so far:
You can fry eggs on both. Incredible how much hotter they get compared to the RTX 2080.
I can not recommend the Vision OC. It quickly ramps up to 2800 RPM and is very loud and has a high pitched whine. This explains why it was available. It sucks.
The Merc handles this much better and is much quieter for two reasons: Less RPM and nicer sound at the same rev. It also has all aluminium covers front and back, so touching it you get a very good idea of how hot it is.
I conclude that the Vision cards cooler and fans are just too small. I’ll send it back. That’s not to say the 3080 chipset is bad. It’s just this particular model, which is the same as the Eagle and Gaming OC models apart from color.
The 3080 has no software apart from thr naked driver and GF experience. You want some functionality, look for 3rd party (MSI afterburner and such).
AMD on the other hand gets you a lot out of the box. Even a mobile app you can use to monitor card metrics while typing messages on the toilet.
I paid 1050€ for the 3080 and 1000€ for the 6800xt. Can’t believe it, but have saved some money like many gamers out there.
Noise aside, what are some quick DCS, MSFS comparisons between the two cards? Interesting as everything else is the same, so curious if stuff like this UserBenchmark: AMD RX 6800-XT vs Nvidia RTX 3080 is actually useful for flight sims (especially in VR )
It’s hard to do a reproducible measurement. So far I looked at rough fps numbers and my feeling. The resolution is not 4k, it’s 1440p on the monitor and 1600x1440 times 2 in VR (Samsung Ody+), so the faster video RAM of the 3080 can’t really flex it’s muscles.
I see mixed results in benchmarks. 3D Mark Timespy is faster on the 6800 but who cares. I mainly used them for artificial load to see how hot and loud the cards get. Furmark does an excellent job at that.
I need more time in MSFS to come to a conclusion. It might have been a little smoother on the 6800xt? Maybe. It does not have more FPS. So hard to say when you can’t really force the game into the exact same situation.
I am already leaning heavily towards the AMD card, but that’s also because of the context. My PC is an old case which I just MacGuyvered to get better airflow, but it still chokes the GPU. New case is inbound.
Leaving the side panel open and revving all fans up to 100% helps, but then my desk situation prevents hot air from going upwards, too. Don’t even think about getting a 3080/6800xt if your case does not have great airflow. It’s all fun and warm feet in winter, but come summer the heat might get annoying.
Is there some sort of standard test I could run in DCS and MSFS to compare against other peoples numbers?
Unfortunately no built-in benchmark tooling for either of them. The nearest to it would be something like a standard DCS quick action mission or MSFS airport like KSFO and a C172 on the deck. MSFS has a ‘Developer’ section in the Options / Graphics that you can use to enable a decent frame-rate counter - that at least tells you a breakdown of CPU vs GPU bottleneck.
Thanks for the comparison so far though. I would think that the advantage of the 3080 is sort of tempered a bit by the 10GB VRAM max vs the 16GB on the 6800 XT. If the O+ was supersampled a bit to higher resolutions it actually gets close to what’s needed for the Reverb, and both DCS and MSFS love having lots of VRAM available.
Thanks for this, especially in DCS. Either would be way faster than my current rig so the ‘other details’ you provide are really helpful (heat, noise, quality, drivers, etc). Thanks again.
None of the cards has gone beyond 8 GB of usage, yet. So this is not a factor at all, actually I completely ignore RAM size. It’s just a number that does not tell me anything.
I did not test DCS yet, though. I’ve heard Syria is RAM-intense.
I have been running “instant replay” on with both cards. They both just don’t care. That’s good, as I can capture the last 8 minutes of gameplay in 1440p 60Hz 50Mbit/s on the push of a button at all times. Old feature, I know. But still awesome.
Ordered another SSD to clone my OS to and later populate my second PC. So for a few days I’ll be able to hop back and forth between 3080 and 6800xt. I need to send one card back in ~7 days. Will be building into a new case Wednesday to Friday, so the Weekend is Benchmark time.
What do you want to see?
My plan:
MSFS High preset (enough Ultra out there)
KSFO(?) with c172 on the runway (glass or old school?) with scattered clouds preset at 3 pm.
The above but drone cam in X feet height
Both of the above in Pancake 1440p and VR in renderscale x (TBD).
DCS some quick mission with frame counter and then just record 5 minutes of flying straight or something. Ideas needed. Can anybody record a good track using most recent stable which ticks all boxes or something?
I will have to reduce runs by not going into OC vs silent Bios vs SAM on/off and the like. The test system will have one state.
I’ll preface this with: I’m not an expert…and I only fly in VR (when not testing missions)
I’ve noticed that DCS - on my configuration (see below) - works better when I keep the reported VRAM below the GPU’s 8GB capacity. And it’s easy to get it over 8GB, especially on Syria (takes work on this map keeping it under 8GB).
In Syria I run it at low settings to keep it smooth. Even then the VRAM hovers slightly above 8GB (8175 Mb’s, +/- 150). What I don’t know exactly is the relationship between the settings. I’m assuming it mostly boils down to how efficiently the GPU can push pixels; the fewer pixels the better the speed, the level of geometry being equal.
When I push my VR SS in SteamVR, or pixel density up by even 15%, changing nothing else, the performance goes too far below 45FPS (motion vector ON). This makes sense of course.
My first thought is: this could be a hassle for you - lot of work. But, if you did it I’d say one of the pre-packaged missions with ‘High/Ultra’ settings. We’ll have to extrapolate from there but it’s a start.
I almost never play them so I’m not going to be much help suggesting one; I’m not sure what’s even in there.
Thanks for you time and effort BTW.
i7 6700K @ 4.5 Mhz, 32 GB RAM, nVidia 1080 (non-Ti) w/8GB VRAM, DCS running on an SSD, Odyssey Plus VR with SteamVR SS = 1.5
For VR I figured out that running Steam SS at 160% will basically equal a 4K resolution, while running at 180% will resemble the amount of pixels of a native Reverb G2 resolution.
I also found that DCS is very cumbersome to test with, as you always have to restart the whole game. MSFS can do it all on the fly, which is a real blessing.
For testing DCS I could start the A10 Syria Free Fly mission and then loiter at 2500 feet above the village using the autopilot. That’s not a lot of action, but at least it’s reprodicible and comparable.
Using the Virtual Superresolution feature of AMD I can run a 4K pancake resolution on my 1440p screen. Maybe Nvidia can do this, too.
I picked up the RTX 3070 today and have been to tired to do anything more than having a fast look into the box and closing it up again.
And the table I am going to work on is filled with tools and 3D print drier box parts. I will finish that first.
@Poneybirdscan you run both test using Gryz’s Benchmark missions??
Two missions used for benchmarking DCS graphics settings. Press “A” for autopilot once mission starts and then activate your benchmarking software when the in-mission message tells you to. Deactivate when stop message appears.
Two missions; one has high CPU overhead, the other low.
do you have fpsVR ?
you can get average fps and frametime!
That Gryz mission is on Caucasus and I don’t feel like investing too much time. I decided to rather test with the F/A18 mission in Syria, which starts on the carrier and then there is a demo flight with the blue angels and a turkish demo team over the city. I think this is more what most people want to see the GPU to pull off.
I am happy to say that I am up and running again with my new I7 7700K CPU and RTX 3070.
I am very happy that I didn’t get a larger card as this one barely fits. I had to slightly wedge it to allow the mount thingie (Where you put in the 2 screws) to get inside.
After some swapping around and fideling with my IDE/SATA board, I had to bend the 2 HD LED plugs slightly to be able to fit in the RTX 3070. In the other position I tried for it, a screw on it made it impossible to get the GTX 1070 into the secondary PCI-E slot.
Now I will upgrade the drivers and go to bed.
Tomorrow I will do a thorough cleanup of the cable jungle behind the desk, before I decide on exactly how I am going to plug in all of my screens, plug em in and then I hope I will install and test the HP Reverb G2.
is made on Caucasus on purpose!
that’s the only map that everybody have and can be a common reference!
many users complain the lack of a benchmark i think as a community we should support this bench… @SkateZilla any way to install the mission and have it run at some sort of default settings from your utility collection???
there are 2 versions
cpu heavy
and cpu light (wich should unleash the gpu power and give high fps)
I see the point, but the mission is not very demanding. It does not tell me much if it just flatlines at 90 fps.
If you go buy a GPU for 1000 quid, you want to know how it performs when under load. You want to know if the stutters are gone and how high the resolution can be. It’s not that much about a global ranking of GPUs. I don’t think we can pull that off without a standard benchmark created by the devs.
So I decided to use carrier ops and glass cockpit and Syria and compare 3080 vs 6800 xt.
DCS streams from the HD when you’re flying as well. It’s not just your CPU, RAM, and GPU, because if you have those settings cranked up with a complex mission and a slow HD you will see a performance hit.
The problem is many of these go hand-in-hand and it’s really difficult to make a benchmark that can say “the reason your performance is poor is X” when most often it’s a combination. That’s why you can upgrade one thing and see little overall improvement because all you did was clear one of a number of bottlenecks.
For instance, faster RAM will always help, but only by a small amount and it’s almost never the cause of bad performance. Conversely, more RAM will help if you’re hitting the limit of what you have, but if you’re not then it’s wasted.
Going from 16GB to 32GB will almost always help in DCS, but 32 to 64? Probably negligible.
)
