Which should be my next card for 1080p gaming at min 60 fps?

[tg2708]

Guy is having issues with his card and people here are pushing brand favoritism towards him. Point is you are not helping and rather than posting none sense give him possible solutions to his problems because clearly the card is not working as it should compared to others of the same chip.

 

[GammaLazer]

Guy is having issues with his card and people here are pushing brand favoritism towards him. Point is you are not helping and rather than posting none sense give him possible solutions to his problems because clearly the card is not working as it should compared to others of the same chip.

Yeah right. Please don't increase my suffering.

Alright I installed windows afresh on SSD and ram the Heaven benchmarks, so here are the results!

 

[tg2708]

Yeah right. Please don't increase my suffering.

Alright I installed windows afresh on SSD and ram the Heaven benchmarks, so here are the results!

Resolution should be 1920x1080 and should be done in fullscreen. As a side note I downloaded heaven to compare but it keeps on crashing so I wont be posting any results as of now.

 

[Black96ws6]

Yeah right. Please don't increase my suffering.

Alright I installed windows afresh on SSD and ram the Heaven benchmarks, so here are the results!

Here are my results at 1290x707, everything maxed, to match your settings so you can compare:

 

[happy medium]

Buy a gtx970, dont buy that hot,power hungry AMD card.

 

[Mondozei]

I would return the card. You probably only got a bad product.

I would also consider buying a used 290 if you can. You get better value for the buck and it's still capable of playing most advanced games at a high frame rate. I would then sell the 290 for 150 bucks or whatever you get for it next year and get one of the 16 nm cards.

Buying a new card this late from previous generations is not a good strategy, regardless if it's NV or AMD.

 

[GammaLazer]

I don't have an option for return now. I can only go for RMA.
Could it be CPU bottlenecking?


Here are the results for 1920 X 1080.

 

[RussianSensation]

1. Make sure your Windows power management is NOT set to Balanced. Change it to High Performance. ExtremeTech even managed to make this mistake in their recent benchmarks and their scores where well below normal until they set Windows priority to High Performance -- always!!

You should NEVER run any PC desktop in the Balanced setting. This setting is made for laptops to preserve battery life and it prevents the CPU and GPU from achieving maximum boost and holding it.

2. Go into the Bios, disable all Intel CPU power saving technologies such as EIST, C1-C3 states.

3. Under Catalyst drivers, move PowerTune to +20%.

Redo all of your benchmarks with 100% CPU and GPU fan speeds.

Run HW Monitor in the background the whole time. What we are looking for are minimum levels of 12V and 3.3V readings from your power supply.

As well as report back maximum CPU and GPU temperatures and % usage (GPU-Z is good for the graphics card).

Also, while you are running the benchmarks, run them all at 1080p or something others can easily compare with you or otherwise we cannot see a point of reference as easily. Also, what are your maximum CPU temperatures?

Your GPU temperatures do not appear high enough to cause GPU clock throttling. That's why you need to co-monitor your GPU clock speeds alongside the GPU usage in MSI AB. My guess is your GPU is purposely throttling its clocks because of a Windows balance on reduced power usage vs. performance and your PowerTune may not be set to maximum. Also, make sure you aren't forcing MSAA through CCC but doing it via the game's menu. Run all benchmarks in full screen mode, not windowed or borderless.

Please get back to us with all of these changes and data.

 

[.vodka]

What I've found strange is that he says the GPU usage isn't hitting 99-100% as it should, a few posts back... but that was before he did a clean windows install.

A 3570k shouldn't bottleneck a 390x much, if at all. GPU frequency should be pegged at the maximum, 1070 MHz in this case, if there isn't a cooling problem... which there isn't. Fluctuating GPU clocks under anything that isn't light on the GPU (CSGO, LOL, etc) is a problem, on a well cooled card like this one and many others.


He might be being throttled by powertune if he's running one of those useless +100-200mV overvolts for an extra 50-75MHz that don't have much if at all extra performance in comparison to the heat output. Most of the 300 series cards I've seen so far, can even be undervolted and remain stable at stock clocks (which are already high for Hawaii, around 1100MHz is what you'd usually do on a 290/290x without getting too voltage happy, and the 300 series cards run at around these clocks).

He might also be being throttled if this particular 390x comes with some stupid factory overvolt unnecessary for 1070 MHz, and the card's BIOS has the power limit set at the usual 208-216w (with powertune at 0%, which means that same limit)... but still shouldn't incur in such a performance loss, I think.

 

[RussianSensation]

Dp

 

[RussianSensation]

Great points. When diagnosing a card problem, you should always test the actual videocard at stock clocks and volts and make sure all power saving settings in Windows are off while PowerTune is turned up to the max. We want to isolate all OS, GPU driver or VGA Bios TDP limits as possible variables. Only after that we can start to accurately assess if the GPU is thermal throttling or something else is wrong with it.

Also, for maximum performance + overvolt + OC, PowerTune set to 0% could easily result in exactly the behaviors described by the OP.

 

[casiofx]

I don't have an option for return now. I can only go for RMA.
Could it be CPU bottlenecking?


Here are the results for 1920 X 1080.

This is the proper score for 390

My score for my 1000Mhz core 1300Mhz 290 Tri-X is around 1350.

My maxed 1230Mhz/1600Mhz 290 is 1559

If you don't believe the screenshot above, this is what I got on 1200/1500 on previous run:

If you want to know if Powertune is throttling your GPU or not. Increase the Power Limit to 50% and see if the score is increased or not. If it does not then there was no throttling previously. You had to be careful with the temps if you're putting 50% Power Limit, especially on those cards with high default voltage in the BIOS.

 

[TXBPNT]

My MSi 390X running at 1150/1700

 

[tential]

My tri x doesn't hold exactly steady at 1000. Fluctuates between 997-1000. Guess I need to add volts

 

[GammaLazer]

First of all I would like to convey my heartfelt thanks to all of you and this AnandTech community.

Actually I sometimes believe that as @Seba had suggested that it could be due to badly applied TIM or some other hardware issue as the card which I got was manufactured in November 2014 and was lying unused in the warehouse of the distributor for around year. That's also the reason i got it for cheap, around 130$ less than market price. That period might have had some effect on card or TIM, I don't know.

Also I had not made and changes in clock speeds and anything, I was running the card on factory settings. Its factory oc'd t0 1070 MHz.


So Followed the advices of all of you and ran the tests once again and the method I followed is as below:

First of all I changed the power plan to high performance.

Then went to BIOS settings and changes minimum cpu fan speed 100% and disabled C, C1E, C3 and C6:

Then increased the power limit in CCC by 20%.

Changed the GPU fan speed to 100% in Afterburner.
And finally ran the Heaven Benchmark.

CPU Details while benchmarking:

GPU Details while benchmarking:

 

[.vodka]

Good, now your 390x is putting out expected results for that particular benchmark.

I'm concerned that the GPU gets to 85°C at 100% fan speed. That can't be good, even with Asus' heatsink that only has 3 heatpipes out of 5 making contact with the GPU... maybe the thermal paste job was botched at the assembly line.

That temperature isn't enough to make the card thermal throttle (that happens at 95°C, like the reference 290/x) but still is concerning... or maybe that's just how the heatsink works in this case. Asus' 290x used a similar heatsink that was designed for the 780Ti and had similar results, too..


Anyway, now put the fan back to auto mode, dial powertune back to 0% and test again. If you see around 58 FPS again, the card is working fine (please note the temperatures, if you get to the throttle point at 95°C performance will suffer and clocks will fluctuate instead of being at 1070MHz). After that, enable c-states again in the BIOS so your CPU can idle properly. If you still get around 58FPS again, everything is working as it should, more or less.

Have GPUz running in the background so you can check for clock speed fluctuations and all the other parameters as they happen and also to have a history of them, hwinfo only shows min/high/actual values.

 

[GammaLazer]

Good, now your 390x is putting out expected results for that particular benchmark.

I'm concerned that the GPU gets to 85°C at 100% fan speed. That can't be good, even with Asus' heatsink that only has 3 heatpipes out of 5 making contact with the GPU... maybe the thermal paste job was botched at the assembly line.

That temperature isn't enough to make the card thermal throttle (that happens at 95°C, like the reference 290/x) but still is concerning... or maybe that's just how the heatsink works in this case. Asus' 290x used a similar heatsink that was designed for the 780Ti and had similar results, too..


Anyway, now put the fan back to auto mode, dial powertune back to 0% and test again. If you see around 58 FPS again, the card is working fine (please note the temperatures, if you get to the throttle point at 95°C performance will suffer and clocks will fluctuate instead of being at 1070MHz). After that, enable c-states again in the BIOS so your CPU can idle properly. If you still get around 58FPS again, everything is working as it should, more or less.

Have GPUz running in the background so you can check for clock speed fluctuations and all the other parameters as they happen and also to have a history of them, hwinfo only shows min/high/actual values.

Okay so here are the results after fan speed set to auto and powertune back to 0%.

GPU Z Readings:

MSI Afterburner Readings

 

[Seba]

Good, now your 390x is putting out expected results for that particular benchmark.

Even the minumum fps (10.8 fps)?

 

[Seba]

GPU Z Readings:

In the drop down menus, select to display the maximum values.

 

[GammaLazer]

In the drop down menus, select to display the maximum values.

I have closed the GPU Z now, but MSI AB reading show the maximum, which I have posted above.

 

[.vodka]

Even the minumum fps (10.8 fps)?

Do the benchmark a second time and you won't see these minimums, since everything is now cached. It's standard behavior for this benchmark.

-----------------------------------------------------------

As far as performance is expected, it's putting out what it's supposed to be, 57-58 FPS. Core speed is fixed at 1070MHz during the run and GPU load is at 100%, no powertune throttling there at 0%, as it should be. So, your 390x is working fine, performance wise. Reinstalling windows was the right thing to do.

Now, 92°C maximum GPU temps at auto fan speed? Damn. Well, the GPU is made to withstand those temperatures without a single issue. If this is your card, it's loading at 87°C there, so ambient temps do play a part here... elevated temperatures compared to other aftermarket Hawaii cards, but well, it's within specs. Asus' heatsink isn't the best for this particular GPU, it's the second time they decide to use heatpipes without a heatspreader, it shows bad results again.

Nothing you can do about that, but again, it's standard for your card. Enjoy!

 

[Seba]

He also got a minimum of 42 fps in Skyrim 1920x1080 Ultra, with official HD textures pack, but without mods. The purpose of this buy was to avoid exactly that (drops bellow 60 fps in games at 1920x1080).

 

[.vodka]

Those results were before the windows reinstall where he was seeing about 80% GPU usage for some bizarre reason. He should check now how Skyrim performs.


And I think I've found what was wrong.

Hey sorry evryone for reviving my old thread again, but i am in need for some help.

I am getting low FPS on most games with this card, please help me out.

GTA V (Very High - Ultra - MSAA 2X)
MAX FPS: 65
MIN FPS: 27
AVG FPS: 49
CPU Usage: 70-100 %
GPU Usage: ~90%

Skyrim (Ultra - No Mods - Pure Vanila - High Res Texture Pack from Bethesda)
MAX FPS: 59.9
MIN FPS: 42
AVG FPS: 51
CPU Usage: 25-30 %
GPU Usage: ~80%

Metro LL (Ultra - PhysX off)
MAX FPS: 58
MIN FPS: 32
AVG FPS: 44
CPU Usage: 50-55%%
GPU Usage: ~85%

Tested only these so far.
Even my 7970HD used to give much better FPS on all these games.

60 FPS max on these two games? That is vsync's fault, that also introduces a truckload of input lag, making everything irritably unplayable if the FPS are below the limit. That also explains the 80% GPU usage number.

GTAV is another matter, he should've been seeing 100% GPU usage if vsync off, not 90%, but still. New windows installation, he should run his games again and make sure vsync is off.

I hate it when games default to vsync on. Yeah, you get rid of tearing, but unless you're aware of it, things like this happen.

 

[GammaLazer]

I will try skyrim now. BTW my table clock shows temperatures of 29.2 current.

 

[.vodka]

Nevermind that. Your card is working just fine. Maybe you're going through summer to get 92°C load temperatures on that card, but it's fine. Nothing suggests it needs RMAing, so far.


Make sure vsync is off.