Overclock.net banner

1 - 20 of 25 Posts

·
Registered
Joined
·
11 Posts
Discussion Starter #1
Hi All
So I've finally given in trying to solve this and have submitted to asking for help! I've had my gaming PC for the last 18 months and I'm convinced there is 'something' not quite right with the level of performance I seem to be getting. The performance is significantly lower than what I should getting. For example, I will be playing a game and get framerate drops and general stuttering and low framerates. It just doesn't feel stable. When I run monitoring software it doesn't seem like anything is maxed out; one of the worst performing games (Wolfenstein: New Order) My GPU load = 50% and my CPU load = 20%. So, even when performance is bad, the utilization seems quite poor. On the other hand, some games run butter-smooth @ 60FPS; The Witcher 3 for example, but this has very high CPU & GPU utilization.

So, a little more information:

+ CPU - FX-8350 run at stock 4.0GHz; all default settings in Frequency, voltage etc
+ Motherboard - Gigabyte GA-970A-DS3P using the latest BIOS "FD" (release 2016)
+ GPU - Radeon RX 480 8GB with Dual Fan
+ RAM - 16GB DDR3 @ 1866MHz (1.5V)
+ Windows 10 64bit DX12

It is worth noting at this point I have not OC'd anything (maybe later once it is stable). I have run some temperature monitoring under heavy loads and nothing is getting higher than 60°C, and actually 50°C is more typical.

Finally, I have run Prime 95 on Blend mode and I do get some errors; primarily 'Rounding was 0.5 expecting 0.4'. It's usually never the same CPU core that fails; so I'm wondering is I have a memory problem as opposed to a CPU problem.

Oh yeah, finally finally; I used to get an issue where if the GPU got too hot the system would crash (White Screen of Death!). I found this is because the Radeon tool, when left to it's own devices, didn't really get the fans going that much. So I set the fans to kick in much earlier and this doesn't happen anymore (keeping the GPU under 80°C does the trick).

If you could offer any advice or insights I'd be hugely grateful!!!
 

·
Senior OCN Member
Joined
·
39,815 Posts
What PSU is in that system? and yes that matters
 

·
Registered
Joined
·
5,374 Posts
Also curious on the PSU.
I'd try turning the ram down to 1333 and see if that helps.

Is the CPU downclocking when this happens? I doubt it is since it's running stock but it could be a problem in some cases.
 

·
Registered
Joined
·
1,415 Posts
Rounding errors in Prime95 usually indicate bad memory. (I see this most often OC:ing memory)
 

·
Registered
Joined
·
799 Posts
Do you have good air flow in your case? Also the CPU cooler is properly installed?

Check your VRM temperatures, if you can install a FAN blowing air into the VRM area, usually helps.

Clear memory and GPU contacts, check for memory problems as some friend sayd.

Good Luck !
 

·
Registered
Joined
·
11 Posts
Discussion Starter #6
It's an EVGA 600W 80 Plus.

I did a bit of a Power calc to check what headroom I had; The processor gets to about 60W under load, 100W for the GPU. I then added some rule of thumb "worst case" ratings (correct me if I'm miles off on any of these); RAM sticks @ 10W each, Mobo @ 50W, 2 disk drives @ 10W each, 3 fans @ 10W each.

That comes to just under 300W with everything flying....or am I miles off?
 

·
Registered
Joined
·
11 Posts
Discussion Starter #7
Also curious on the PSU.
I'd try turning the ram down to 1333 and see if that helps.

Is the CPU downclocking when this happens? I doubt it is since it's running stock but it could be a problem in some cases.
That's a serious underclock you're talking there, what are you thinking? I'll give it a try as a look-see.

The CPU doesn't seem to be downclocking at all. If anything I regularly see 4.1GHz on all 8 cores, which I guess it does because it's underloaded. I also see drops to 1.4GHz, but only at rest, not during gaming.
 

·
Registered
Joined
·
11 Posts
Discussion Starter #8
Rounding errors in Prime95 usually indicate bad memory. (I see this most often OC:ing memory)
Can you be more specific? what would be bad exactly? If its something i can resolve manually through reseating or adjusting settings then great; id rather not have to spend more money on new RAM if at all possible?
 

·
RGB Rainbow Vomit!
Joined
·
990 Posts
My 8320 did not like the RAM running anything above 1766, let alone 1866 or higher.
1333 is the base clock however you may find 1600Mhz as a suitable speed depending on the memory itself, BIOS update may help if available.

GPU should not be getting that hot if its an AIB card with a non reference cooler. May be a sign of poor airflow in general which would result in motherboard VRM being too hot.
What CPU cooler do you have in there? What case model and fan layout?

Your CPU voltage is rather high for only 4Ghz on an FX 8 series, my 8320 on a Sabertooth board would give a stable 5.3Ghz on little more than 1.4v.
The motherboard in question has woeful power delivery to the CPU: 4+1 phase delivery and no heatsink, only a thin heatsink over the MOSFETS. There is a common complaint on various forums about these boards officially supporting 125w CPU's but burning out with FX 8 CPU's even at stock speeds. Reminds me of the old MSI K9A2 which would fry the MOSFETS if you put one of the decent Phenom's in there.
 

·
Registered
Joined
·
11 Posts
Discussion Starter #10
Do you have good air flow in your case? Also the CPU cooler is properly installed?

Check your VRM temperatures, if you can install a FAN blowing air into the VRM area, usually helps.

Clear memory and GPU contacts, check for memory problems as some friend sayd.

Good Luck !
I'd say its OK. i have 3 internal fans creating good airflow, bringing cool air in at one end and pushing warm out the other. Cooler looks square and well seated; it gets warm so is pulling heat away from the chip. Like i said, i monitored temps and nothing got above 60, which i wouldn't consider 'hot'.

Which sensor is the vrm? i have 3 motherboard temperatures but they're just called 1, 2 & 3! will try blowing direct onto the vrm and see if that helps, cheers!
 

·
Senior OCN Member
Joined
·
39,815 Posts

·
Registered
Joined
·
11 Posts
Discussion Starter #12
My 8320 did not like the RAM running anything above 1766, let alone 1866 or higher.
1333 is the base clock however you may find 1600Mhz as a suitable speed depending on the memory itself, BIOS update may help if available.

GPU should not be getting that hot if its an AIB card with a non reference cooler. May be a sign of poor airflow in general which would result in motherboard VRM being too hot.
What CPU cooler do you have in there? What case model and fan layout?

Your CPU voltage is rather high for only 4Ghz on an FX 8 series, my 8320 on a Sabertooth board would give a stable 5.3Ghz on little more than 1.4v.
The motherboard in question has woeful power delivery to the CPU: 4+1 phase delivery and no heatsink, only a thin heatsink over the MOSFETS. There is a common complaint on various forums about these boards officially supporting 125w CPU's but burning out with FX 8 CPU's even at stock speeds. Reminds me of the old MSI K9A2 which would fry the MOSFETS if you put one of the decent Phenom's in there.
BIOS is the latest, I checked. So in theory I could underclock my RAM and see an improvement in performance? Have you any idea what the theory is behind this, or is it purely a thermal issue?

I'll have another look at airflow; I've got 3 fans in there creating a flow from front to back. The card has got quite a high profile though and I could see how it might block cool air getting to the VRM; it's in the way!! It's funny actually because on GPU heavy games like Witcher 3, the GPU fans fire up properly and the whole system runs better when this happens!

I have a Cooler Master Hyper 212x on my CPU which, again, seems to be doing a job. I'm not seeing temperatures that look worrying at first glance, at least. And I'm not overclocking anything yet.

The CPU voltage is at around 1.2/1.3V. The other voltage I quoted was the DRAM voltage which is at 1.5V (although in reality when you measure it, it sticks at 1.488V).

I always suspected my Mobo as the weak link in the chain...also the hardest part to replace >_< Anyway, I'm looking at reducing voltage in order to not overload the VRM's right?
 

·
Registered
Joined
·
11 Posts
Discussion Starter #13
You really need a better PSU because what you have is not very good
https://www.overclock.net/forum/31-...ou-should-not-buy-evga-400-600-500b-600b.html

Might be why your system runs so badly or at least its part of the problem
A PSU like that does not belong in a system like yours
Very Interesting! thank you.

So, the overall inefficiency of the Voltage regulation (crappy components) is leading to unstable rails? So if the 12V/5V/3.3V rails are jumping around and generally a bit rubbish then the system struggles with stability? That's what I'm reading into your article...

How does the ripple affect things? It just looks like noise, and I'd be interested to know how this affects the system.

Incidentally, do you have any recommendations? Ideally I'd like to spend under £50 but will take your advice on what's best.
 

·
Senior OCN Member
Joined
·
39,815 Posts
Very Interesting! thank you.

So, the overall inefficiency of the Voltage regulation (crappy components) is leading to unstable rails? So if the 12V/5V/3.3V rails are jumping around and generally a bit rubbish then the system struggles with stability? That's what I'm reading into your article...

How does the ripple affect things? It just looks like noise, and I'd be interested to know how this affects the system.

Incidentally, do you have any recommendations? Ideally I'd like to spend under £50 but will take your advice on what's best.
Sory but you wont get anything priced at £50 that is good
A Corsair CX650M is £70 and even that is not great, a Seasonic Focus Plus Gold 650 watt is around £90 and that is about the best option under £100

https://www.amazon.co.uk/Corsair-CP...ie=UTF8&qid=1533211312&sr=8-1&keywords=CX650M
https://www.amazon.co.uk/Seasonic-S...sr=8-1&keywords=seasonic+focus+plus+gold+650w

Scan.UK or Overclockers might have those cheaper but i did not check

Edit: Scan has it for £85
https://www.scan.co.uk/products/650...d-sli-crossfire-single-rail-54a-120mm-fan-atx
 

·
RGB Rainbow Vomit!
Joined
·
990 Posts
BIOS is the latest, I checked. So in theory I could underclock my RAM and see an improvement in performance? Have you any idea what the theory is behind this, or is it purely a thermal issue?

I'll have another look at airflow; I've got 3 fans in there creating a flow from front to back. The card has got quite a high profile though and I could see how it might block cool air getting to the VRM; it's in the way!! It's funny actually because on GPU heavy games like Witcher 3, the GPU fans fire up properly and the whole system runs better when this happens!

I have a Cooler Master Hyper 212x on my CPU which, again, seems to be doing a job. I'm not seeing temperatures that look worrying at first glance, at least. And I'm not overclocking anything yet.

The CPU voltage is at around 1.2/1.3V. The other voltage I quoted was the DRAM voltage which is at 1.5V (although in reality when you measure it, it sticks at 1.488V).

I always suspected my Mobo as the weak link in the chain...also the hardest part to replace >_< Anyway, I'm looking at reducing voltage in order to not overload the VRM's right?

Not really about RAM thermals but more to do with rated speeds, supported speeds and stable timings.
Try running your RAM at 1333 and 1600 to see if either speed is more reliable. Is your RAM actually rated for 1866 out of the box?

RAM voltage may be dropping too much causing instability when its under load. Try manually setting it to 1.55v.
When my Corsair CX600M was dying, the motherboard was constantly giving me warnings that various voltages had fallen below spec before the system would hang.

Airflow from the 212 will not be blowing over the VRM; being an AMD board, the 212 will mount so air is blowing out of the top of the case, not across the VRM's which are between the socket and IO. This may or may not be the cause of your problem but it is definitely a weakness of that motherboard.
 

·
Senior OCN Member
Joined
·
39,815 Posts

·
Registered
Joined
·
1,415 Posts
Can you be more specific? what would be bad exactly? If its something i can resolve manually through reseating or adjusting settings then great; id rather not have to spend more money on new RAM if at all possible?
Usually bad timings that cause errors (as in doing a ramtest it will return errors at one time or another).

But the returned 0.5 when expecting less than 0.4... Usually I have noted is about tWRWRSCL/tRDRDSCL being too tight for the speed/voltage. A notch or two more voltage usually fixes it(or loosen the timing), though this is experience from Ryzen clocking, so unsure how it relates to the older architecture.
 

·
curmudgeon
Joined
·
5,930 Posts
Try setting the ram voltage to 1.6v. It will not hurt anything and should help with stability. I was told this by MSI in 2013.


That motherboard does not have vrm sensors or heat sinks so it really isn't suited to the 8350. Try removing the I/O cover where all the cables attach to get better air flow in that area. The I/O cover creates a dead air zone right where the vrms are.



What monitoring software are you using? HWINFO64 works best with FX and is free. If everything is on auto then it is probably running in the 1.38-1.4 volt range. Consider 1.4 volts to be the redline for that motherboard. You might consider undervolting it in the future as a properly tuned 8350 should run at 4 ghz @ 1.25-1.28 volts. I got my 8370 stable at 4 ghz @ 1.256 volts.
 

·
Registered
Joined
·
11 Posts
Discussion Starter #19
Update:
So I've tried a few suggestions and I've just run the Prime 95 Blend test each time. Not tried any games yet

+ Set RAM to 1333. Run Blend. Fail @ 4 mins. Worker 6 "rounding error"
+ Set RAM to 1.6V (still @ 1333). Run Blend. Fail @ 4 mins. Worker 6 "rounding error"
+ Set RAM to 1600 @ 1.6V. Run Blend. Fail @ 4 mins. Worker 6 "rounding error"

It's interesting that It's the same worker every time.

attached is typically what I see furing test. My core voltage is around 1.2V. The CPU gets to around 59°C (but it's a warm day here in the UK which won't help!)

I'm going to try bumping down the CPU clock speed a touch to see if that helps. I've not tried any games yet; might give Wolfenstein a try and see if performance is different.
 

Attachments

·
Registered
Joined
·
5,374 Posts
Since all this stuff is used it's pretty hard to figure out the exact cause here, the CPU could be bad as well as the board or ram.

Have you used any of this stuff before in a different rig? I'm trying to rule something out here, can you give me something to go on?
Testing the ram in a different PC would be great or past experience too.
 
1 - 20 of 25 Posts
Top