AMD FX Overclocking

humbug · 27 Mar 2015 at 14:22

craig_t said:
My package temp rises close to 60 and CPU is going to 70+. I only have a dual rad but my GPU is also cooled in this loop. CPU is before GPU in the flow sequence and directly after the RAD. I am upgrading to a tripple rad but, I would have expected to get better temps. Currently I am running:
240 FSB, 2648HT Link, 1.32V, and 19 multiplier, why are my temps so high? This is just doing a prime 95 test no graphics etc are working.

Prime95 goes way beyond what the CPU is intended for, and to be frank its not even that good at finding instability, all it does is hammer the CPU with prime numbers calc.

Use AMD's own tool, it bombards the CPU with multiple tasks simultaneously. Asus Real Bench is pretty good for pushing the CPU hard with tasks that you do on a daily basis.

Prime95 is old hat, CPU's are far more complex than that, all it does is boil it.

LtMatt · 27 Mar 2015 at 15:04

Nice to see this thread going strong.

craig_t · 27 Mar 2015 at 15:27

humbug said:
Prime95 goes way beyond what the CPU is intended for, and to be frank its not even that good at finding instability, all it does is hammer the CPU with prime numbers calc.

Use AMD's own tool, it bombards the CPU with multiple tasks simultaneously. Asus Real Bench is pretty good for pushing the CPU hard with tasks that you do on a daily basis.

Prime95 is old hat, CPU's are far more complex than that, all it does is boil it.

Thanks I will try that out

Currently seems stable at 4.575GHz @ 1.308V with 240FSB

thebennyboy · 27 Mar 2015 at 15:40

craig_t said:
Thanks I will try that out

Currently seems stable at 4.575GHz @ 1.308V with 240FSB

That's an incredible clock speed for that voltage. I have to use 1.404v to get that clock.

craig_t · 27 Mar 2015 at 15:54

thebennyboy said:
That's an incredible clock speed for that voltage. I have to use 1.404v to get that clock.

I need to do more testing, but it does seem to run ok. I will feed back when I have done more, 4.5 is the target I wanted to run at so I can keep the voltage lower but still get good gains over the 3.3 standard. I might push up to 4.6XX, depending how it goes on 4.57

gpuerrilla · 27 Mar 2015 at 16:09

thebennyboy said:
That's an incredible clock speed for that voltage. I have to use 1.404v to get that clock.

Yes mine would be similar benny, over 1.4 for that clockspeed.

Scurburg · 27 Mar 2015 at 16:43

craig_t said:
My package temp rises close to 60 and CPU is going to 70+

If thats in prime95, then that is fine. Nothing will get close to those temps in normal use. As I said before and others have said too, use Asus realbench for stability, you have got all the details you need from prime now (highest possible temp and llc settings)

Other things you can try is keeping NB and HT around the same speed (I think I ran mine both at 2600-2700, but NB didnt like much higher)

craig_t · 27 Mar 2015 at 16:54

Scurburg said:
If thats in prime95, then that is fine. Nothing will get close to those temps in normal use. As I said before and others have said too, use Asus realbench for stability, you have got all the details you need from prime now (highest possible temp and llc settings)

Other things you can try is keeping NB and HT around the same speed (I think I ran mine both at 2600-2700, but NB didnt like much higher)

Yep absolutely I will be using realbench. I am at work and have no internet access on my own PC to download it, until I get back home...

I will have a play with NB

craig_t · 27 Mar 2015 at 19:03

Ok I am home now and have my PC plugged in...

Here is my figures for realbench after 10 min stress test:

Not sure why voltage showed 1.6 highest, it just popped up there at the start doesn't normally do that. Both NB and HT Link are 2648, I will see if I can step them up, but the jumps are big and I need to go to a 300 bus to get my memory back to 1599mhz, otherwise I need to underclock it or seriously overclock it.

A question though, if I was running prime earlier at a lower voltage it never crashed but it stopped a test on 1 core so I put the voltage up. Was it just coincidence or was I right that there was not enough voltage?

Scurburg · 27 Mar 2015 at 20:27

Generally it is just a lack of voltage that will cause that, but it could indicate a weaker core if any increase in voltage does not help (Like my core 7&8 above 4.9 even at 1.55v).

Not sure about the 1.6v though, Are you using offset or a set manual voltage?

craig_t · 27 Mar 2015 at 20:34

Offset, like I say though it was a 1 off it sits at the steady 1.308 I think it just loaded that when opening HWM.

Bench tests all work fine, gaming works smooth. I changed my RAM timings from 11-11-11-28 to 10-10-10-28, I am tempted to try getting to 9-9-9-27 but I will leave that for another day I think.

Bonjour · 30 Mar 2015 at 17:08

Couldn't get my 8320 stable at 4.6, no matter how much I added to the offset.

It's solid at 4.5 with +0.125 offset and LLC 'medium'. Going to 4.6, though, AMD stability test will crash the computer within 10 minutes even with offset at +0.15.

Should upping LLC to 'extreme' help?

Dog · 30 Mar 2015 at 17:14

Try it at extreme then reduce the offset to find the minimum for stability.

Are your temps ok with that offset?

Rich

OhEsEcks · 30 Mar 2015 at 17:16

I think I had my LLC set to extreme on my 990XA-UD3 with an 8320e and could do 4.6GHz stable. But it was pulling 1.4v at that. If I tried going any higher, it started to reach thermal limit for the chip. But I was using Intel Burn Test for the stress test. If I'd used Realbench and had a better cooler, it might have been capable of 4.7 or more as it wouldn't have been hammered as hard.

Edit: My post from earlier in the thread: http://forums.overclockers.co.uk/showpost.php?p=27591740&postcount=137

Bonjour · 30 Mar 2015 at 17:20

Dog said:
Try it at extreme then reduce the offset to find the minimum for stability.

Are your temps ok with that offset?

Rich

Yeah, temperatures stay below 50 currently when doing the AMD stability test.

How accurate are the readings? I read somewhere they can be out by as much as 20C due to the algorithm method utilised.

Bonjour · 30 Mar 2015 at 17:22

OhEsEcks said:
I think I had my LLC set to extreme on my 990XA-UD3 with an 8320e and could do 4.6GHz stable. But it was pulling 1.4v at that. If I tried going any higher, it started to reach thermal limit for the chip. But I was using Intel Burn Test for the stress test. If I'd used Realbench and had a better cooler, it might have been capable of 4.7 or more as it wouldn't have been hammered as hard.

Edit: My post from earlier in the thread: http://forums.overclockers.co.uk/showpost.php?p=27591740&postcount=137

Cheers. Will take a look. I'm on a similar motherboard - 990FXA-UD3.

gpuerrilla · 30 Mar 2015 at 20:21

Depends on what you are using to read the sensor. Go for the package temp, your cooling should keep it below 60c.

craig_t · 31 Mar 2015 at 11:29

My LLC is only on high too, just to add that in..

Bonjour · 31 Mar 2015 at 12:01

Ah. I'll try 'high' rather than 'extreme'. Cheers.

Kidge · 31 Mar 2015 at 20:46

On a bit of a side note to the current discussion but still related to this thread. My OC throttles like hell under load and the only way I was able to resolve it was to install AMD Overdrive - Enable Turbo Boost then disable it and click OK and then it stopped throttling.

Anyone else experienced this?