Sapphire RX 5700 XT Pulse OC results

Anomnomnomaly · January 30, 2020

I've had my 5700XT since around Sept last year, was rather impressed with it's performance but aware that the drivers were a little buggy... I didn't mess around with any overclocking because to be honest.. I was more than happy with the performance @ 1440p on my then 75hz monitors.

But I picked up a 144hz gaming monitor in the black friday sales... so I've decided it was time to try and push it harder.

I've done some testing using the built in tuning with AMD's own radeon software, and with afterburner just to see if there was a difference... and I can't find any.

So far my testing shows that I can push the 1750mhz core to 2139mhz and the 1700mhz memory to it's max limit of 1900mhz.

I'm kinda impressed and disappointed... I wasn't expecting that much of a boost on the core but I was expecting a little more from the memory... swings and roundabouts and all that... We all know ram clocking gets better gains than core. But I'm kinda pleased so far and junction temps are solid in the 70's under load, hitting low 80's if ambient is higher (like when the heating is on).

I'll take that as win... considering I paid £325 for the card when it was supposed to be £399 and is normally closer to 420 or more... I got the XT for non XT money and given the 5600XT is around £350 in the UK at the moment (if you can even find one)... I'm laughing all the way to the bank.

Jurrunio · January 30, 2020

7 minutes ago, Anomnomnomaly said:

I can push the 1750mhz core to 2139mhz

but does it actually run that high? All you've done is to let it go that high, not that it would do so.

Anomnomnomaly · January 30, 2020

5 minutes ago, Jurrunio said:

but does it actually run that high? All you've done is to let it go that high, not that it would do so.

So far I've only run unigine heaven, for an extended amount of time (about 40 mins) with everything turned up, ultra settings and Just ran a fresh benchmark with it to make sure... and noticed a tiny flicker on the other 2 monitors every now and again as I run 3x 27s. So I've just dialled it back 50hz on the mem and taken it back to 2050 on the core and will test it further.

I did find a quirk in the amd software vs afterburner though. If I let the amd auto tune the card, it takes the core to 2100, but when it sets the vram auto clock to 1900 it gives a pale single colour screen on monitor 1 that you have to switch main monitors to another and back... but manually setting it doesn't... Still some bugs in the code obviously... and you can only auto tune one or the other it seems.. not both at the same time. But It does you a baseline for manual tuning in afterburner.

On the slightly lower clocks, temps are a lot lower, maxed out 75ºC and the heating was on... with a radiator just 6ft away from the system.

Obviously a lot more tweaking and testing will going on... and all of this is on air in a case that has plenty of decent fans but not fantastic front intakes so those corsair ML 140mm fans can get a little noisy... Got some mods for the front panel planned this weekend.

Arrogath · January 30, 2020

9 minutes ago, Jurrunio said:

but does it actually run that high? All you've done is to let it go that high, not that it would do so.

When my clock is set to 2150 @ 1200mv and +50% power limit I only hit around 2100 to 2110 under water just for reference. I've played with undervolting but my performance drops off, same with slight clock speed target reductions. Just because you've dialed in a target doesn't mean that's the speed it's running, you'll need to monitor that in real time

Anomnomnomaly · February 6, 2020

I've been doing some more benchmarking and have the following results.

I'd been running into some issues when trying to COP the vram... my main monitor would suddenly go to a single colour, luckily as I run a multi monitor setup, just changing the default to another and back resolves this. I thought I was pushing the vram too hard... but it turns out that there's a conflict between the AMD software and MSI afterburner and I'd left them both running at the same time... D'oh.

All temps checked with HWINFO64 v 6.04

Shadow of the Tomb Raider - Stock settings in Blue, VRAM OC in Green, Core OC in Orange and Combined OC in Red

All tests done on the same settings in the same room, at roughly the same time/ambient settings when the heating wasn't running.

1440p, 144hz, FXAA, DX12, everything on high, tessellation ON, Screen Space Reflections ON, Specular Reflection Normal, Dynamic foilage Medium, Bloom and vignette On, motion blur on, hair on... lens flare, screen effects and film grain off.

Stock 116fps Core temp 54ºC Junction Temp (max) 76ºC

Vram OC (1900mhz) 124fps Core temp 66ºC Junction temp (max) 87ºC

Core OC (Max 2004mhz, Min 1954Mhz) 125fps Core temp 72ºC Junction temp (max) 101ºC

Combined OC (1900Mhz Vram - 1978mhz Max - 1946 Min) 126fps Core temp 74ºC Junction temp (max) 97ºC

So here we run into a rounding error really... doesn't seem to matter in this game if you OC the memory or the core or both together... you hit the same rough level... But the Vram OC gives the far better temps.

I did some Far Cry New Dawn tests, but messed up the results by forgetting to change a settings... But the stock was 74fps rising to 89fps with everything cranked up to 1440p ultra, but AA, motion blur off... 2 of the tests were done with the core and vram clocked as I forgot to reset the vram before running again with the OC... and in retrospect I should have lowered the settings to something similar to tomb raider which I'd left on high as the last time I played it was on my old RX580 and decided to leave the settings there, rather than rerun everything on that too.