• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

R9 290 and 290X Overclocking - Is pushing for max always worth it?

Soldato
Joined
30 Dec 2011
Posts
7,277
Location
Belfast
Hi all,

I have been meaning to definitively find out if pushing my R9 290X overclocks really made a difference. Most people assume simply pushing until you get artifacts, then back off slightly is the best way to find the best 24/7 overclock. What I actually wanted to see was is there a substantial difference between pushing for max OC and a moderate OC.

I used Tomb Raider benchmark for my testing and used max IQ settings in Radeon Pro. I also used SweetFX (via RadeonPro) to enable image sharpening and SMAA. So no driver optimisations were used at all and image enhancements were set in SweetFX. Finally, I set my own personal preference for the game graphics settings as follows.





Using the above settings I set about testing various core clock and VRAM speeds. My results are recorded in the table below.


As can be seen the max I could reach without artifacting was 1200/1600. I was able to complete a benchmark at 1230/1650 OC and have included the result for reference. I found the following intersting.

  • In order to achieve best results, it is necessary to increase both core and VRAM clocks.
  • Increasing one while ignoring the other will marginalise the performance increase slightly.
  • 1100/1600 compared to 1200/1600 should be noticeable on paper, but it truly does make zero difference in game-play.
  • By contrast the increase from 1000/1600 to 1100/1600 showed a definite and noticeable difference even during game-play.

I have concluded there is essentially a diminishing return in truly pushing a GPU to its limits. Pushing a GPU until it artifacts, then backing off slightly is not always beneficial. For example the fan speed at 1100/1600 was noticeably quieter than what was needed at 1200/1600. So the very marginal ~3% gain in performance was not worth the extra noise.

I know most here will already know this, but I felt it might be useful for new comers to see.
 
My 290 is a terrible clocker, it'll only go up to 1125 on the core, any higher and it'll start artifacting.

I've also noticed that upping the memory literally does nothing in terms of FPS gains in games/synthetic benchmarks. I benched Heaven at 1125/1300 then 1125/1550, the 250Mhz jump in memory speed gave me precisely 1FPS haha, literally pointless.
 
Mine runs at 1200/1500 artefact free, its starts artefacting at 1225/1500, i haven't taken it any higher than that. (Using Sapphire Trix +200Mv)

I run it at 1100/1400 +75mv 24/7, i don't see the point in running it higher than that, running it mid way through highest artefact free overclocks is a good balance between higher customised performance and piece of mind for the longterm safety of the GPU.

That extra 5 to 10% pushing it to the limit day in day out makes no sense to me just for the sake of that little bit more.
 
Is there any way of somehow upping the core without it artifacting or is that it for me? :(

Yes, Stop using MSI AB for Benching.

MSI AB only allows +100mv (1.3v) i can't get mine much past 1150 without artefacting on those volts.

Hawaii GPU's are capable of running 1.4v easily

Sapphire Trix with +200mv and all the artefacting is gone right up to 1225Mhz.
 
Really like AB though. Tried Trix before and it seems a lot more bloated than AB too, do you know if the new AB beta fixes the 100mv limit?

I don't know, ask Matt or 8Pack^^^^. but i have to say i find Trix more stable than AB for benching.

For gaming i just use AB with +75mv and its happy, as i say, i don't run it 1200Mhz 24/7.
 
I would say that Tomb Raider isn't a particularly representative game to test this with as it shows much larger increases from memory overcloks than most other titles.

I keep my 290X at 1150 at 50mv. I can get slightly higher with higher volts but it isn't worth it. I've had no luck with memory clocks though, some games will run at 1400 but others with black-screen. I have ****** VRAM :)
 
I would say that Tomb Raider isn't a particularly representative game to test this with as it shows much larger increases from memory overcloks than most other titles.

I keep my 290X at 1150 at 50mv. I can get slightly higher with higher volts but it isn't worth it. I've had no luck with memory clocks though, some games will run at 1400 but others with black-screen. I have ****** VRAM :)

Turn your vCore up from 50mv, Memory volts and vCore are tied together on these GPU's. :)
 
My 290x still struggle to overclock beyond 1375GHz stable. Any tips of how to clock it higher and stable or is it "you got what? Elpida memory? Might as well give up"? :p


Read above you :p mine also has Elpida, i struggle to get it past 1400Mhz unless i turn vCore volts up.

For whatever reason Memory and core volts are now tied together, turning your core volts up also turns the memory volts up.
 
Both my Elida's do 1400mhz on stock voltage, 1500 with voltage bump of +25 - +50mv and will both easily do 1625 at +100mv. Marine you may have got unlucky because your card has unlocked to a 290X. Maybe as a result of that there is less headroom on the memory. Or... you just got a stinker. That's what you get for keeping that 5850 for so long. :p
 
Last edited:
i wonder if AMD are playing silly ###### here?

The original 7970 was run at 1250/1350Mhz out of the box on the memory, so what everyone did is turn up the volts and had them running @ 1800Mhz in no time.

It made the 7970 GE they released later look pointless with its 1500Mhz memory out of the box.

So this time round they are stopping that by setting lower memory volts and tieng those volts overclocking into the core, so you can't crank up the volts and run the memory at silly-Mhz now to give AMD's next GPU more of a reason to exist with higher Memory clocks.
 
i wonder if AMD are playing silly ###### here?

The original 7970 was run at 1250/1350Mhz out of the box on the memory, so what everyone did is turn up the volts and had them running @ 1800Mhz in no time.

It made the 7970 GE they released later look pointless with its 1500Mhz memory out of the box.

So this time round they are stopping that by setting lower memory volts and tieng those volts overclocking into the core, so you can't crank up the volts and run the memory at silly-Mhz now to give AMD's next GPU more of a reason to exist with higher Memory clocks.
Yea. I do recall people mentioning that on the 7970 they can increase the voltage for the memory directly or something, but can't do that for the 290...

Have tried upping the voltage to +150mv, but doesn't help...
 
Last edited:
Turn your vCore up from 50mv, Memory volts and vCore are tied together on these GPU's. :)
It didn't make much difference to the possible memory overclock. Given the minor difference the memory overclock makes in most games I just gave up to avoid having annoying black screens.
 
i wonder if AMD are playing silly ###### here?

The original 7970 was run at 1250/1350Mhz out of the box on the memory, so what everyone did is turn up the volts and had them running @ 1800Mhz in no time.

It made the 7970 GE they released later look pointless with its 1500Mhz memory out of the box.

So this time round they are stopping that by setting lower memory volts and tieng those volts overclocking into the core, so you can't crank up the volts and run the memory at silly-Mhz now to give AMD's next GPU more of a reason to exist with higher Memory clocks.

I read somewhere it was the memory bus controller that was giving less OC potential (512 vs 384).
 
Back
Top Bottom