Quick Impressions: Gigabyte RX 5700 XT Gaming OC

jhatfie

jhatfie

[H]ard|Gawd
Joined
Mar 19, 2003
Messages
1,636
Just got my Gigabyte RX 5700 XT Gaming OC delivered today from NewEgg. Messed around about an hour to get a quick and dirty OC and did a few real quick comparisons to my Radeon VII which was lightly overclocked.

First observations. 5700XT is getting random flickering with two monitors running at the same time, 1 monitor is fine. Did not have same problem with Radeon VII.

+50% power limit does seem to allow a slight boost in clocks, but junction temps start to get high. +30% runs a bit cooler and not much drop in performance compared to 50. 950mhz is max I can go on memory in AB or wattman....curious how much higher it can go. No voltage adjustments either. Would love to see how undervolting does.

I run a custom fan curve, with 55% fan set for when temps hit 70c. 90% fan at 85c. Noise is actually very similar to my Radeon VII at the same fan speed settings. Not bad, but not all that quiet either. Not surprised as both cards have what appears to be the same fan sizes.

OC Settings:
5700 XT : 2070 core | 950 memory +30% power limit
Radeon VII: 1840 core | 1100 memory (this was my 24/7 daily settings with slight undervolt which gave more stable clocks)

Tests were just the canned benchmarks in the game except Control, which I tested over a 5 minute time and used FRAPS. All settings mostly at all highest possible settings (few exceptions though like volumetric clouds and shadows usually a click below max) and 1440p.

Radeon VII - Min | Avg
Division 2: NA | 74
AC Odyssey: 26 | 71
Far Cry NewDawn: 78 | 99
Metro Exodus: 33 | 57
Forza Horizon 4: 91 | 105
Control: 47 | 62
TimeSPY: 8877 (graphics score) Although to be fair my best score was 9560 at borderline unstable OC but was on a different build.

5700XT - Min | Avg
Division 2: NA | 68
AC Odyssey: 27 | 67
Far Cry NewDawn: 77 | 96
Metro Exodus: 33 | 57
Forza Horizon 4: 94 | 108
Control: 50 | 62
TimeSPY: 9533 (graphics score) Stock score was 8585

So with those games anyway, really pretty dang close performance. Still have some messing around to do, but I doubt there is much more there for performance unless voltage and memory controls become available for me.

20190903_190402.jpg
 
Last edited:
Crazy how close 5700 XT gets to the VII at 1440p. I'd expected at 4k, however, the VII would pull ahead some.

As for the flickering, can you try switching the cables and port connections e.g. try running a single monitor out of the second port and see if you're stick getting flickering. You could also switch your secondary monitor to your primary in Windows and see if the issue persists: if it persists on the primary monitor, it's likely a hardware issue (port or cord); if the filickering occurs on the secondary monitor, it might be software or drivers.
 
Thanks, great info. Any chance you got any power #'s between the two? Let us know if you get to undervolt, I'm still running my fury OCed and undervolted, much more efficient than factory :).
 
Ready4Dis said:
Thanks, great info. Any chance you got any power #'s between the two? Let us know if you get to undervolt, I'm still running my fury OCed and undervolted, much more efficient than factory :).
Click to expand...

I did not get any accurate power numbers, but if afterburner is accurate at all, I saw a peak of 278W with averages maybe in the 220-240W range (seems similar if I recall to my undervolted VII). Again, just eyeballed. Temps peaked at 70C and were generally 65-68C. Memory peaked at 72C, VRM 64C. Clocks jumped around, but generally averaged over 2000Mhz through most all the testing, depending on the game I was testing I'd say the average was 2020-2040mhz.

I'd need to get more testing in though to say for sure, as above info was just from limited testing.
 
Last edited:
jhatfie said:
I did not get any accurate power numbers, but if afterburner is accurate at all, I saw a peak of 278W with averages maybe in the 220-240W range (seems similar if I recall to my undervolted VII). Again, just eyeballed. Temps peaked at 70C and were generally 65-68C. Memory peaked at 72C, VRM 64C. Clocks jumped around, but generally averaged over 2000Mhz through most all the testing, depending on the game I was testing I'd say the average was 2020-2040mhz.

I'd need to get more testing in though to say for sure, as above info was just from limited testing.
Click to expand...
No problems, thanks for the additional info.
 
The flickering is a conflict with MSI Afterburner. I have the same problem.

In any case, you can do mostly the same things in Wattman, so just close/disable Afterburner and you should be okay until they fix it.
 
cybereality said:
The flickering is a conflict with MSI Afterburner. I have the same problem.

In any case, you can do mostly the same things in Wattman, so just close/disable Afterburner and you should be okay until they fix it.
Click to expand...

I was going to mention this, figured it out yesterday. Thanks for sharing though!

Did a quick comparison with overclocked/undervolt to stock running the Metro Exodus Benchmark. First OC setting was stock power limit, +30Mhz core, 950mhz memory and 1.15v.
OC+ setting was +30% power limit, +30mhz core, 950mhz memory, voltage 1.15v.

Stock
Low fps | Avg fps | Edge | Hotspot | Peak Core Speed | Peak Watts | Fan speed
31.97 53.91 72C 98C 1916Mhz 195W (Auto Fan) <---- Right around 50-55%

OC
Low fps | Avg fps | Edge | Hotspot | Peak Core Speed | Peak Watts | Fan speed
32.77 54.95 64C 87C 1960Mhz 196W (60% peak fan)

OC+
Low fps | Avg fps | Edge | Hotspot | Peak Core Speed | Peak Watts | Fan speed
33.84 57.78 67C 98C 2012Mhz 227W (60% peak fan)



OC was about 2% improvement over stock in terms of average fps
OC+ was about 7% improvement over stock in terms of average fps.

Increasing power limit showed power draw go up but not no real gains in fps with this specific test.
 
that's pretty much the story of the game for all amd products. There's nothing really left on the table from the rating sold. Just tons of heat produced for no significant gain. But that's not bad, the products being sold are performing very well.

I'll have some stuff to say about how this behaves in debian probably by mid next week. I use only the open source amd drivers so it should be interesting to compare stability / performance from my current rx 480.
 
Here are some teardown pic's. Looks like they did a pretty decent job. Thermal paste was not put on crazy thick and everything made decent contact. Temps did not improve when reassembling with GC Extreme. I was hoping that hot spot temps would improve a little. It is interesting how certain games heat the gpu more than others, for instance Control hammers it and gives me the highest temps, then games like Forza, even with 100% gpu are much cooler running.

20190906_110830.jpg 20190906_110821.jpg
 
Another observation, for some reason this specific card's idle temps are also way higher than I'd expect. Even if I manually put the fan to 30% minimum fan, when idle and doing nothing, the temp always hangs out at around 44-45C in a 21-22C room. A good amount higher than any other card I've owned. My Radeon VII was mid 30's at idle with the same fan curve. Reseating the cooler with new paste did nothing to improve the idle temps either.
 
Last edited:
I too am surprised at how close the 5700 XT is to the Radeon VII. Here I thought there was no way it would be faster than Vega 64 before they were released. Good information here. Thanks for sharing your experience so far.
 
  • Like
Reactions: N4CR
like this
Yeah the XT was a nice surprise. Especially at 1080p the performance is pretty comparable to the Radeon VII.

1440p is still good, but once you go to 4K, the VII pulls ahead. Way better than Vega 64, though.
 
Darth Ender said:
your temps in general seem a bit on the high side. Perhaps poor contact or maybe a bad heatpipe? 21C room temp, https://www.gamersnexus.net/hwreviews/3509-gigabyte-rx-5700-xt-gaming-oc-review-benchmark
and these are temps with the fans at a given noise level, not even at full tilt.

Heatpipes can fail, a non-functioning heatpipe could massively impact heatsink perf. just a thought.
Click to expand...

Yeah I have reinstalled the cooler several times and temps are consistent. Even at default NOT OC'd settings with default fan curve, temps seems higher to me than I would expect. Playing Control (What seems to be one of the most gpu stressing games I have) I am seeing 81-82C core and 103-105C hot spot average. Default fan at those temps seems to be 46-47%. Cranking the fan up helps, but it starts getting noisy at 60% fan so it is the max I prefer to use for daily gaming. With default settings, except using 60% fan I am still seeing 74C core and 96C hot spot. Those temps are after at least 30 minutes of gaming compared to some of my tests earlier which were short duration and did not give enough time for full temps to be reached as it was just a few loops.
 
well, debian unstable still doesn't have support for the 5700. even building mesa git with llvm 10 and running a built 5.3-rc8 kernel still results in a failure of xorg to use the amdgpu driver and have hardware accel. Trying to use the debian built mesa 19.2-rc1 pkgs results in just deadlocks when xorg tries to use the amdgpu driver. Xorg logs dont really show much info on why, since as usual with those kind of errors, it doesn't write to the log.

trying to use amdgpu-pro similarly failed, since, as usual, the proprietary drivers use older kernels and I'm not running an older kernel just for the driver.

So, stuck in xorg with non-accelled vesa type drivers. Supposedly 19.2-rc2 will work, since rc1 was compiled with the wrong version of llvm but that probably wont get pushed to the server for a few days

On the windows front, the amd drivers also suck in that OS as well. Half the games I have in that OS failed to work. Elite Dangerous just sits there with a blank screen. No Man Sky gets to the initializing screen after loading it's shaders and locks. Both of those aren't in steam.

Tried a steam game (Conan Exiles), that worked fine. Guess i'll be crossing my fingers that Borderlands3 works well.


All in all, i guess it's a good reminder as to why I was still running an rx480. AMD graphics drivers truly are garbage for a good 3-6 months after hardware release.

And while Nvidia always ...and repeatedly, deserves a huge middle finger. What the hell amd. Hire some competent software engineers
 
You should have no problems in Windows 10. I tried at least 10 or 15 games with no issue.

But do make sure your are not using MSI Afterburner, RIva Tuner, etc. There is a conflict right now.

I'd recommend a DDU safe mode uninstall, make sure to remove both AMD and Nvidia drivers in case there are left overs.

You can also unplug the internet before the DDU uninstall to make sure Windows doesn't auto-install an old driver. Then install AMD latest before plugging back in.
 
i dont use tuning tools in windows, nor do i install motherboard tools or 3rd party card drivers etc. Only amd drivers from amd before this new card and now.

reinstalling the drivers wouldn't solve these issues since i'm not experiencing an inability to play any game that uses the same graphics api / features. Just some games are triggering lockups / freezes or are crashing when trying to render.

It seems on the windows front, AMD has an issue with navi misbehaving when there are multiple applications attempting to use hardware accelerated calls (like chrome and a game or a game and some video player or a game and a video encoder or a game and say the microsoft gaming overlay)

Lots of reports of issues involving that, which maybe what is occuring on my particular set of games.
 
I have had my Gigabyte Gaming 5700 XT for a few days now. I have experienced a BSOD at least once a night, almost entirely while browsing websites with embedded video.. usually youtube. I have had great performance in games and no instability at all while gaming, so i'm content for now and hoping the drivers get cleaned up to cut down on the BSOD issue.

I have not done any driver cleanup using DDU but I think I might give that a shot. The fact that it has not had any issues while 5+ hours of gaming makes me think driver cleanup won't help very much, but you never know.
 
DDU can absolutely still help, it takes 10 minutes and it's worth a shot in case driver files were corrupted or left over from a previous install.
 
So a couple games stopped crashing once I manually deleted the game's graphics/display settings file.

One downside to modern games behavior in no longer having a 2d ui to config graphics settings prior to getting into the actual game I guess.

Should be able to test Linux support by this weekend. Debian hasn't yet rebuilt Mesa and pushed to the servers.

Power usage currently in Linux vs windows is significantly higher as it seems in Linux the system is not dropping to the lowest pstates or sleeping cores as much as windows seems to
 
cybereality said:
DDU can absolutely still help, it takes 10 minutes and it's worth a shot in case driver files were corrupted or left over from a previous install.
Click to expand...

Ok I ran DDU in safemode and ran it twice to clean out any nvidia drivers and amd. Unplugged ethernet cable then installed the latest 9/10/19 drivers from AMD. I'll report back in a few days if I notice an improvement in BSODs.
 
dx12 badass quality setting 1080p ....benchmark in borderlands 3 seems to average out around 74fps. Though I'm also using triple buffering to reduce frame tearing since I dont have a freesync monitor.

gpu temps stay in the mid 70's according to the wattman overlay. fan speed is in the high 1900 rpms. ambient temp was 76F.

Pretty happy with the performance so far.

Though, i think the dx12 driver is glitching in certain lighting glow effects - more of an issue with the beta quality of that rendering engine than the card though.
 
So I still got BSOD at least once a night with my Gigabyte 5700xt, even after cleaning everything out and running DDU in safemode. So I saw a few nights later 19.9.2 came out that said it addresses the BSOD issues with video playback, so I installed them. Glad to report that it's been at least 3 days and not a single crash since the update. Anybody else have luck with the 19.9.2 drivers?
 
I only had issues with lockups during some games where they configure graphics settings within the 3d ui of the game and do not auto reset when the previous card is no longer detected. Never any bsod's. Deleting the config files manually to reset the default fixed those.

Also, I was already using an amd card before. I never clean install. Been fine for years, still fine now with the current drivers.

My computer is always on, so I'd notice if it decided to crash on its own just sitting in Windows or Linux.
 
I was borrowing my buddy's GIgabyte 5700 XT OC last week, played with it for a few days. Compared to my RTX card, (2070S), it was hard to notice a difference... and since my monitor is the PG279Q, I was also use to G-Sync. I ran BF-V on Ultra, 1440P, R6 Siege on Ultra, and Ghost Recon Wildlands. About my only complaint with the 5700 XT card, and maybe this is Gigabyte's fault, was that it ran hot! The 2070S I use is pretty much the same card, 3 fan, OC version, etc... but while the 2070S get's warm during game play, the 5700 XT get's egg cook'in hot. Otherwise sweet card.
 
Make sure you're not comparing junction temperature (reported on AMD cards) to edge temperature (on Nvidia).

Junction (or hotspot) temps will usually be around 15 to 20C hotter than edge, even though both chips are technically at the same temperature.
 
cybereality said:
Make sure you're not comparing junction temperature (reported on AMD cards) to edge temperature (on Nvidia).

Junction (or hotspot) temps will usually be around 15 to 20C hotter than edge, even though both chips are technically at the same temperature.
Click to expand...

possibly...? My 2070S run's about 65-70c when gaming. The 5700 XT, runs about the same, 65-75c... the difference is, the RTX I can keep my finger on the back of the card while gaming, mean while the 5700 XT would give me 3rd degree burns.
 
dave343 said:
possibly...? My 2070S run's about 65-70c when gaming. The 5700 XT, runs about the same, 65-75c... the difference is, the RTX I can keep my finger on the back of the card while gaming, mean while the 5700 XT would give me 3rd degree burns.
Click to expand...
could just be the back side of the 5700 makes better contact with the hot spots and thats why? im sure your trying for max airflow to cool it off!
 
primetime said:
could just be the back side of the 5700 makes better contact with the hot spots and thats why? im sure your trying for max airflow to cool it off!
Click to expand...

3 top mounted Noctua Redux 1300rpm fans exhausting, 1x 140mm Fractal exhausting on the back, 3 120mm fans attached to a h150i for Intake, and 1x 120mm Noctua Redux on the bottom floor for intake. Definitely have good airflow going, case is an Fractal R6. I also keep the front door off the case, so the AIO intake fans can breath properly. The back plate on the 5700 XT having more contact could definitely be the case... because they seem to run about the same temp.
 
Last edited:
dave343 said:
the difference is, the RTX I can keep my finger on the back of the card while gaming, mean while the 5700 XT would give me 3rd degree burns.
Click to expand...

The GPU die, memory modules, and VRMs generate heat -- if anything else is hot, it's because it is drawing heat away from those three things.

It's not a good or bad thing, really.
 
You must log in or register to reply here.
Back
Top