OCing A10-5800k noob

[Fortkentdad]

Let's try this post again ~ in mid post the screen jumped to "Thank You for Logging in", strange, happens when I tried to enter the degree sign° (which now is working??)

Anyway, I was expressing my doubts about the validity of HW Monitor. I had tried to run HW and Prime95 all afternoon, but when I returned to the PC it had gone to sleep and when I went back to it, both HW and Prime85 had "not responding" messages.

I checked the log and was concerned as it showed a max temp of 127° C.

But when I reset HW Monitor with no CPU load, it gave a very high reading even though ASUS AiSuite monitor reported normal temps.

I started HW Monitor and then ran a Win8 CPU test App (Speed Test Pro). It is a short test and reports on the Hash MB's. During the test I watched HW Monitor, temps rose a little as expected then there was a blimp, and the max posted the same 127° C for a split second and then back to the normal range. So I have my doubts about this 127° C max. I also am not sure what is being measured by the TMPIN3 or AUXTIN, but they always report very high temps.

Will try again but is there another temperature monitoring program?

 

[Fortkentdad]

Ok. so I watched it this time, about 90 minutes of Prime95 testing. Appart from those blimps where HW jumped to 127° the CPU temp raised gradually to about 44°C where it remained constant. I tracked it in ASUS's Sensor Recorder and the temperatures were virtually identical to HW except there were no 127 blimps. You can see the flat line in the ASUS sensor. I used GPU-Z to monitor the GPU temp, which remained pretty constant at 41° C. Curiously GPU-Z also can monitor the iGPU temp and reports it at about about 54° making it the hot spot on the system. But how can there be a 10° C difference between the iGPU and the CPU - that's the same chip (not sure what GPU-Z uses to measure the iGPU's temp. The second GPU is running just a couple of degrees cooler than the main GPU at 39°. I'm not at all sure what "Package" HW is measuring but if that temp is accurate it is the hotspot at over 81° C, but what is this temperature package??

 

[Johan45]

Others have said that HWmonitor isn't very reliable with the FM2 chips. You could try Aida64 it has a 30 day trial period

 

[Johan45]

Do you have all the auto OC features turned off?

 

[RGone]

Real good build journey at another forum. Pics and stuff galore.

Undervolting + Underclocking the A10 5800k Is the thread title here in the AMD Cpu Section. MacClipper posted a link to a long thread he is doing in another forum. Quite the read actually. Darn informative.

 

[Johan45]

There aren't alot of people with the APU'. I for one don't have one I'm just trying to give general advice for OC'ing so bear with us.

 

[Fortkentdad]

Prime95

Do you have all the auto OC features turned off?

For that test above I believe I did. I used reset to defaults in AI Suite and in the Bios I also selected the defaults settings. The tests at those lower settings ran a little cooler than the the following test.

I reset the system to run a a 45x multiplier and turned on the auto OC features to their extreme settings and re-run the torture test with Ai Suite's sensor recorder running. It seems that the CPU hi temp mark is about 54° C with it running at about 50° C for most of the test. Prime95 torture test ran for about seven hours before the system shut down. Not sure what to make of that. The Prime95 log shows all test it ran were positive. HW Monitor was also run during this time, again it shows a 127° max which can't be right.

So, now I know that using the auto OC features in the Bios and/or Ai Suite the system runs a hotter (about 10° hotter: 54° vs 44°) is this a concern, how hot is too hot?

I also know that my system has some instability after several hours of torture testing. Is this a concern?

 

[Fortkentdad]

Do you have all the auto OC features turned off?

Real good build journey at another forum. Pics and stuff galore.

Undervolting + Underclocking the A10 5800k Is the thread title here in the AMD Cpu Section. MacClipper posted a link to a long thread he is doing in another forum. Quite the read actually. Darn informative.

Thanks for the link, lots to digest there.

 

[Johan45]

I also know that my system has some instability after several hours of torture testing. Is this a concern?

That depends on what you are using it for. And is kind of objective, heres a link to a thread started by RGone about stability if you're interested some good opinions by some old pros.
http://www.overclockers.com/forums/showthread.php?t=686655
It could have been a hiccup in windows or your ram.

 

[Johan45]

So, now I know that using the auto OC features in the Bios and/or Ai Suite the system runs a hotter (about 10° hotter: 54° vs 44°) is this a concern, how hot is too hot? QUOTE]

This is quite normal since the board is adjusting your multi and voltage by presets this is to try to guarantee stability. You could probably achieve better results manually but your system would always be at those settings so it won't slow down when not being pushed.

 

[Fortkentdad]

Win7 results

I decided to go back to defaults and test in Windows 7 rather than Windows 8.

I was able to have the system run Prime95's torture test for about six hours without any problems. I tracked the temps using Ai Suite's Sensor Tracking and the temps rose slightly to about 45° C and remained constant.

I did that test with the two HD 6670's Xfired.

I tried to run the Cinebench 11.5 benchmark, the graphics one returned an error but the CPU came in at 3.19 (this was tested with the Xfire disabled but I don't think that affects the CPU test?).

I decided to use Furmark to benchmark the graphics in part because it is quick, and it seems to work well in Win7 or Win8 and I can get some comparisons as I make adjustments.

Curiously when the Xfire is enabled the graphics FPS is less????

Onward with the testing, I'll OC and see what I get. ....

 

[Fortkentdad]

Spent some time using PCMark7, I tried PCMark11 as I do have Directx 11 on this system but Mark11 kept failing. Maybe it does not support this APU or MB? Odd that the older PCMark7 worked flawlessly?? Curious and curiouser.

Anyway it did demonstrate a clear advantage to using the auto OC features, with a little OCing and the use of a discrete GPU there is an advantage. Now having said that it only goes so far. It is not clear if there is much of an advantage to pushing to OC past the 43x factor, or pushing all the settings to "Extreme" (one step above Turbo or High depending on the item being adjusted). Taking the GPU to the max seems to be fine. And I used the published settings for the Kingston HyperX RAM which had to be manually set as the mBoard used the 1333 settings and not the 1600 settings. I also pushed the OC on the memory a bit and was getting 1800.

I did run into some trouble with 46x and greater in BIOS - it refused to boot, and I had to let it cool down and get in, quickly set to defaults, reboot, and then go in and tinker, if I was not quick enough the system just shut off. Now I had set all the OC factors to the max when this happened. Note I have not gone to manual settings in the BIOS, I'm just using the pick from a menu of options for the various settings.

These are the PCMark7 scores for various settings.

 

[Fortkentdad]

NovaBench

I also used NovaBench ~ why? Because it works, its easy and its free. I'm not sure how accurate it is, at least not in Win8. I notice they are promising a new version, the current version was released in 2011 so it probably is not designed for Win8 (although it works well in Win8). It also would not have been designed with AMD's A-Series APU's and their dual graphic capacities.

Anyway, in Windows 7 I was able to consistently get results in the 1000-1080 range. In Win8 my last test was 647, and most scores are in the 700 range, that means Win7 is 50% faster than Win8 according to NovaBench. I think their Win8 scores are off.



https://novabench.com/profile/46614


I have 40 or so scores on NovaBench for this set up if you are interested just go to their score site and look for Fortkentdad 's scores.

 

[RGone]

One of this forum sections regulars posted some NovaBench stuff a week or so ago. He said it gave odd results. I downloaded it and tried it and got a score. Said okay now let us get 'more' score. Bumped cpu speed up; raised ram speed and CPU_NB; defragged hard drivve and tested again and NovaBench politely gave me "less" score. Less score everytime I ran it. That is plainly stupid in my mind. and un-installed the crap.

 

[Johan45]

I have had limited experience with the xfire/sli setups in the past and I'm fairly certain that all three of the GPU's should be running the same frequency. Sometimes that isn't important but I think maybe the auto OC feature on the iGPU may be throwing it off.
I'm certain you can shut that off in the bios.

 

[Fortkentdad]

A-Series APU and multiple dGPU's

I have had limited experience with the xfire/sli setups in the past and I'm fairly certain that all three of the GPU's should be running the same frequency. Sometimes that isn't important but I think maybe the auto OC feature on the iGPU may be throwing it off.
I'm certain you can shut that off in the bios.

I had hopes of a three way Xfire on my rig when I found out I needed to get a different dGPU, one with a displayport if I was going to run three monitors the way I wanted them (Eyefinity single display across 3 monitors).

Anyway after trying too many times I was able to find out on the AMD website that I can run two of the three. Either on HD 6670 in what they call "Dual Graphic" mode (still shows up as xfire in AMD's own Vision Control Centre/Catalyst program) or Xfire the two HD6670's I have, but I cannot "triple fire" all three at once. One of the three will just be sitting there dormant. At least that's my understanding.

Hate to say this but my hopes of gaining much benefit of Xfire-ing the two dGPU over one dGPU and the iGPU have been dashed. There is only the smallest margin of benefit and some tests show the benefit goes to the dual graphics mode.

As for the auto-OCing of the iGPU. That setting seems to be applied to all GPU's integrated and discrete, but for an experiment I should see if that can be turned off and just use AMD's Catalyst Vision Control Centre's GPU overdrive to pump up the volume on the GPU, worth a try.

Thinking of pulling that second HD6670 and seeing if I can sell it. Anyone with a A10 or A8 chip looking for the dGPU that works with your chip???

I also learned that the Lucid MVP that came with the mboard as an extra only works when you have a iGPU and dGPU in that dual graphic mode.

If I'm wrong about any of this, someone please correct me.

 

[Fortkentdad]

One of this forum sections regulars posted some NovaBench stuff a week or so ago. He said it gave odd results. I downloaded it and tried it and got a score. Said okay now let us get 'more' score. Bumped cpu speed up; raised ram speed and CPU_NB; defragged hard drivve and tested again and NovaBench politely gave me "less" score. Less score everytime I ran it. That is plainly stupid in my mind. and un-installed the crap.

What free benchmark utility do you recommend?

 

[Johan45]

I couldn't find your mobo listed in the posts but from the pics I guessed at this one ASUS F2A85-M PRO. Does it have the lucid logix chip on it ? I checked the mobo manual at http://dlcdnet.asus.com/pub/ASUS/mb/SocketFM2/F2A85-M_PRO/E7502_F2A85-M_PRO.pdf and it is very specific as to how you should set this up. Earlier when you said you bought another card because of the ports it didn't dawn on me. But started thinking when you reported odd video functionality. The M3N board I had included an onboard gpu and supposedly had the ability to run a hybrid solution which would switch from card to chip depending on the load but it never worked very well and Asus stopped trying to support it. But the monitor was connected by the port on the board not the card.
Any way what I'm saying is you may need to re-examine your whole Xfire set up to get it functiong properly and verify you followed the steps in order.

 

[Johan45]

Sorry I was composing while you posted. Tha's kinda lame about the 3 way. You Certain?

 

[Fortkentdad]

I couldn't find your mobo listed in the posts but from the pics I guessed at this one ASUS F2A85-M PRO. Does it have the lucid logix chip on it ? I checked the mobo manual at http://dlcdnet.asus.com/pub/ASUS/mb/SocketFM2/F2A85-M_PRO/E7502_F2A85-M_PRO.pdf and it is very specific as to how you should set this up. Earlier when you said you bought another card because of the ports it didn't dawn on me. But started thinking when you reported odd video functionality. The M3N board I had included an onboard gpu and supposedly had the ability to run a hybrid solution which would switch from card to chip depending on the load but it never worked very well and Asus stopped trying to support it. But the monitor was connected by the port on the board not the card.
Any way what I'm saying is you may need to re-examine your whole Xfire set up to get it functiong properly and verify you followed the steps in order.

First, I was sure I posted a reply but I don't see it here??? odd.

Anyway, thanks for the reminder to go back to the basics and make sure I covered my bases. I had used the manual when I built this rig and followed it pretty religiously.

But I still learned something in review. I learned that I probably didn't need to buy that second HD 6670 to get a three screen display, looks like I could have used Lucid and added one HD 6670 but made the iGPU the primary which means I could have used the displayport on my my mboard and saved myself $75.00. Lesson learned.

Now do I want to go back and take out that second HD 6670 and see what happens when I use the APU-iGPU and take advantage of that feature?

I have tried enabling Dual Graphics through AMD's Vision Control Centre (which I learned is just AMD's name for their Catalyst program when it is installed on an AMD system, "catalyst' is used on intel systems). Anyway, there is not much difference in performance between the Dual Graphics and the Xfire of the two HD 6670's. Which would be expected as the iGPU, the HD 7660D is supposed to be matched up with the HD 6670 which suggests they both have about the same capacity. But it might be worth the effort to open the case up and play with my cards.

From AMD's website, from what I can read, if I had not configured the cards properly, or not used the recommended drivers, then AMD's VCC would let me know. I jumped to this conclusion reading through their FAQ's and seeing the number of times that "if this blah blah blah problem, then your cards not installed properly" reply. AMD's VCC reports both HD 6670's and reports them in Xfire formation, with the HD 7760D (the iGPU) disabled. I've also used AMD's utility to check to make sure I'm using the latest drivers for all my cards and the VCC ~ it reports no new updates available.

But a good review, and I wish I had re-read it carefully before rushing out and buying that second HD 6670 just to get a displayport that I already had. Anyone want to buy a HD 6670? Makes a perfect companion to your A10 or A8 chip.