Looking for a Suggestion (FM2)

[imaputer2]

Bearing in mind I was setting up some data for an attempt at 5GHz stability, rather than also exploring Ram and Gpu OCs as well, the only OC record I have available - inserted below - comes from my top stable OC just before I undertook these tests. (I hope it shows up in the post - I don't see it in the preview)

If it doesn't show there's a copy on display at OCN.

I should also say that this level of driving under constant load has had a degrading effect on the APU, so that the test is now producing voltages 1 or even 2 increments above the original results.

 

[TsunamiJuan]

Asus seems to have the best implementations of A85X so far. I have an ASUS F2A85-V PRO and a Gigabyte F2A85X-UP4 and have severe overheating problems when trying to overclock an A10-5800K on the UP4 past 4600, with temps in Prime95 going over 80C, meanwhile the V-PRO easily does 4900 in Prime95 at about 54C.

Asus gives you access to Digi+ VRM in windows through its AI Suite II (recently updated again), while Gigabyte's UEFI only gives you the two basic LLC settings. The following test I did on the ASUS wouldn't be possible on the UP4 as it is implemented at the moment.

Using the standard APU frequency of 100MHz and incrementing the APU Multiplier from 38X to 50X I've attempted overclocks with the default Digi+VRM settings and then using the Smart Digi+ key in AI Suite II from 3800MHz to 5000MHz. LLC settings in my test were made using the optimized VRM setting facility in AI Suite II on Asus boards. I first set and apply the Multiplier/Frequency in Turbo EVO then go into DIGI+ Power Control and click the colored Smart DIGI+ Key which generates an appropriate setting for the applied OC.

The CPU voltage required in each OC to pass 30 minutes in Prime95 is given with the number of 0.00625V steps above or below the default setting of 1.3125V.

+------+---------------+---------------+
|CpuMHz|.Default VRM...|.SmartDigi+Key.|
|ApuFrq+-----+---------+-----+---------+
| =100 | -/+*|..Vcore..|.-/+.| .Vcore..|
+------+-----+---------+-----+---------+
| 3800 | -15 | 1.21875 | -19 | 1.19375 |
| 3900 | -10 | 1.25000 | -16 | 1.21250 |
| 4000 | -05 | 1.28125 | -12 | 1.23750 |
| 4100 | +02 | 1.32500 | -07 | 1.26875 |
| 4200 | +07 | 1.35625 | -02 | 1.30000 |
| 4300 | +13 | 1.39375 | +05 | 1.34375 |
| 4400 | +18 | 1.42500 | +10 | 1.37500 |
| 4500 | +26 | 1.47500 | +17 | 1.41875 |
| 4600 | +38 | 1.55000 | +24 | 1.46250 |
| 4700 | +50 | 1.62500 | +34 | 1.52500 |
| 4800 | +62 | 1.70000 | +46 | 1.60000 |
| 4900 | 80+ | 1.8125+ | +60 | 1.68750 |
| 5000 | 80+#| 1.8125+ | 80+ | 1.8125+ |
+------+-----+---------+-----+---------+
| * 1.3125V to 1.9V = 94 x 0.00625V....|
| # 80+ = Prime95 fails below 1.8125V..|
+--------------------------------------+

Generally you shouldn't need to play with LLC at all when pushing to 4.5 ghz unless your using an extreme overclock on your IGP. The powerstate of the chip generally will dictate and scale cleanly aslong as temp throttling and OCP controls are not coming into play. If your pushing the IGP alot then it comes into play more but in most cases the OCP starts pushing you back more than the default core voltage does.

With a more air unfriendly chips I have consistently pushed 4.7ghz on water (ghetto closed loop) the setup is roughly equivilant to my noctua nh-d14 in cooling capacity. With around 1150ish IGP speed, and a core voltage of only around 1.5. I am sorry but the 1.8v to push past 50x also sound extremely horrible. Sounds like way to much voltage for air/water. Even to much for a phase change setup. I think i was at around 1.6 - 1.65vcore to push 5.4ghz stable on phase change. I could of probably gone alot lower with spending added time to tweak but I was in a rush to put numbers up.

http://www.hwbot.org/submission/2324726_

Obviously this changes quite a bit depending on the quality of a chip. If you have a Low leakage chip your more likely to see better air performance with the voltage curve scaling quicker in the higher clocks. The Higher leakage chips tend to scale better under cold and not quite so far on air. I am currently using a High leakage chip with a rather low initial voltage requirement.

As for boards. Heres why i would not buy asus right now. I've spent the last about 16 years buying asus products in a buisness enviroment. In the last 6 months or so their build quality has gone down the toilet. A long with their quality control. When you order 20 motherboards and 3/4's of them have some sort of defect that shows up in 30 days that requires them to be sent back under warranty. Theirs something wrong there. Not to mention when the warranty returns have problems themselves and have to be sent back again, you end up having to just pretty much throw them all away. (5 of them where doa out of the box to begin with). This isn't the first time in this time period I have received Asus products with horrible build quality and quality control. Until they Ditch ECS as their Contractor for building their products. I am staying away from them.

 

[azanimefan]

I should also say that this level of driving under constant load has had a degrading effect on the APU

can you elaborate what you mean?

 

[imaputer2]

can you elaborate what you mean?

So where Prime95 in the first test ran for 30 minutes at 4800MHz with 1.6V (+60) it now needs 1.6125V (+62).

Generally you shouldn't need to play with LLC at all when pushing to 4.5 ghz unless your using an extreme overclock on your IGP. The powerstate of the chip generally will dictate and scale cleanly aslong as temp throttling and OCP controls are not coming into play. If your pushing the IGP alot then it comes into play more but in most cases the OCP starts pushing you back more than the default core voltage does.

With a more air unfriendly chips I have consistently pushed 4.7ghz on water (ghetto closed loop) the setup is roughly equivilant to my noctua nh-d14 in cooling capacity. With around 1150ish IGP speed, and a core voltage of only around 1.5. I am sorry but the 1.8v to push past 50x also sound extremely horrible. Sounds like way to much voltage for air/water. Even to much for a phase change setup. I think i was at around 1.6 - 1.65vcore to push 5.4ghz stable on phase change. I could of probably gone alot lower with spending added time to tweak but I was in a rush to put numbers up.

http://www.hwbot.org/submission/2324726_

Obviously this changes quite a bit depending on the quality of a chip. If you have a Low leakage chip your more likely to see better air performance with the voltage curve scaling quicker in the higher clocks. The Higher leakage chips tend to scale better under cold and not quite so far on air. I am currently using a High leakage chip with a rather low initial voltage requirement.

As for boards. Heres why i would not buy asus right now. I've spent the last about 16 years buying asus products in a buisness enviroment. In the last 6 months or so their build quality has gone down the toilet. A long with their quality control. When you order 20 motherboards and 3/4's of them have some sort of defect that shows up in 30 days that requires them to be sent back under warranty. Theirs something wrong there. Not to mention when the warranty returns have problems themselves and have to be sent back again, you end up having to just pretty much throw them all away. (5 of them where doa out of the box to begin with). This isn't the first time in this time period I have received Asus products with horrible build quality and quality control. Until they Ditch ECS as their Contractor for building their products. I am staying away from them.

Firstly, I was using a special (ie. loud) OC air cooling setup in an open case using a Mugen 3b and a set of 3000rpm GT fans modded for use in a PWM chain which allowed me to operate these voltages at around 54C, as I said earlier. I had found a wide variation in the voltages being registered for the A10 at CPUZ and wanted to see what the limitations actually were. The point of the test was to establish some performance curves rather than searching for sweet spots.
Your trade report on ASUS quality control is really worrying, but I have had FM1 and FM2 Asus boards in the last 12 months with no problems at all.

A10-5800k OCs 15/10/2012
at valid.canardpc.com
--------------------
Apu X  Vcore CPU  Ram  FsMm Cpu order
100 40 1.288 4000 1200 1:12 3 
101 47 1.428 4747 1077 3:32 7
122 32 1.452 3904 1301 3:32 1
099 40 1.464 3991 1064 3:32 2
100 46 1.520 4600 1066 3:32 5
099 46 1.524 4591 1064 3:32 4
099 48 1.548 4791 0931 3:28 9
140 34 1.560 4760 0933 3:20 8
105 45 1.592 4726 0840 1:08 6
100 51 1.616 5100 1066 3:32 12
108 47 1.620 5076 0864 1:08 11
131 39 1.620 5112 1398 3:32 13
100 50 1.648 5000 1066 3:32 10
120 61 1.856 7320 1120 3:28 17
114 58 1.896 6635 0762 3:20 14
113 62 1.920 7007 0904 1:08 15
118 62 1.956 7317 0629 3:16 16
119 62 1.968 7384 0800 3:20 19
119 62 1.984 7422 0798 3:20 20
119 62 2.052 7379 0634 3:16 18
120 62 2.076 7441 0640 3:16 21

 

[TsunamiJuan]

The really big thing about running those voltages on air, is that You might pull off 4core CPU results but your gonna enter a hole different world of tweaking and OCP/voltage problems once you start playing with a 4/c + IGP load. Which is really the most important thing about these setups for most people. They are generally forgiving of high voltages. Fm1 would take 2.4volts and keep asking for more. Though there where diminishing returns if you really didnt take the time to really tweak things right it still didn't do you much good.

The real voltage lines that matter on these chips are the NB voltage line. Since that gets split in the chip Into standard northbridge voltage and IGP voltage. This line by itself is the single most annoying voltage to work with. With a the Mix of OCP control and general power output and the general lack of from vendor support for playing with it currently.

As for asus quality right now, I have a few that have been working just fine aswell. But the amount of money that they have cost me in shipping alone lately I would of been better off buying boards from another Company. Which Is what I am doing till I can see an improvement in their build quality. Its really a shame cause I have built and sold countless systems with their hardware over the years and never had problems like this. Or the lip that they seemly have been giving about it aswell. White box replacement boards that have problems with them aswell doesnt improve my faith in them. Which is sad.

As far as records and top end performance are concerned i'll use whoever works the best and lets me achieve what i need. But Right now everyones deck is a little different. So I will probably end up obtaining just about every OC target A85x full ATX board on the market, at some point.

BTW power phase wise right now, DIGI vrm which is a remarked chil chip who got bought by IOR. Is also used on the ASRock Extreme 6, and Gigabyte Up4. They just adjusted different and matched with different Mosfets. But the controller chips are pretty much all the same. ASRock and Gigabyte have both had a few interesting flickers of added voltage control in some of their beta bioses. Hopefully they will show up at some point to the release bios. I'll be talking with ASRock again sometime this week and be asking them for these features.

 

[imaputer2]

I sure hope Gigabyte do something about power phase on the UP4 soon because this should be a good board.

As far as I can see the two respective sets of BIOS controls for the Asus F2A85-V PRO and the Gigabyte F2A85X-UP4 probably explain the overheating problem the UP4 is giving me.

ASUS F2A85-V PRO
CPU Load Line Calibration
CPU Current Capability
CPU/NB Load Line Calibration
CPU/NB Current Capability
CPU Power Phase Control
CPU Voltage Frequency
Fixed Frequency Mode
CPU Power Duty Control
CPU Power Thermal Control
DRAM Current capability
DRAM Fixed Frequency Mode
DRAM Power Phase Control

Gigabyte F2A85X-UP4
VCore Loadline Calibration
NBVID Loadline Calibration