ASUS M5A97 R2 not accepting OC values

[Teemsan]

This is my first time overclocking, and on a new build. I have an FX 6300, cooled with a Hyper 212 EVO, on an ASUS M5A97 R2.0 mobo
http://www.newegg.ca/Product/Product.aspx?Item=N82E16813131873
with 8Gb (2x4) GSkill Ripjaws DDR3 1600 cl8
http://www.newegg.ca/Product/Product.aspx?Item=N82E16820231445
running Win 7 on a Kingston Hyper X 120Gb SSD (a 1GB WD HDD storage no RAID)

THe mobo BIOS version was 1903 and I updated to the latest - 2103

The first thing I need to mention is that initially my RAM was only being recognized at 1498 Mhz both in BIOS and CPUz. With default settings in the BIOS there was no 1600 option. Next up from 1498 was 1700. So in the manual for the mobo it said to go to the Ai Tweaker>Ai Overclock Tuner and set it from Auto to DOCP, which I did. Then a 1600 Mhz selection was available in the Memory Frequency drop down box. I set it there, rebooted and the RAM showed up as 1600 but only cl11. Back to BIOS and confirmed it was cl11, in the DRAM Timing Control. I went back to the Memory Frequency drop down and realized at the bottom there was also my exact RAM profile available at the bottom of the selections. Set it, rebooted and all was good - CPUz confirmed the correct speed, latency and cycle.

If I'd stayed in DOCP to get the correct RAM settings I couldn't OC the CPU - It requires setting the Ai Overclock Tuner back to Manual - not DOCP - and of course my RAM settings are then lost. I can go back to manually selecting 1600 (not my previous exact profile), but it will then only give me cl11 - confirmed both in the BIOS and CPUz. I should also mention that when it is at 1600 - either cl8 or 11, it still gives me a WEI of 7.9, and was way better than the 1498 on stock BIOS settings. So I decided to stay with the cl11, and Ai Overclock Tuner at Manual to just get on with the OC.

As a base I just ran everyhting at stock on Prime95 for half an hour just to see if all cores were OK, and no problem, temps about 38C. I wanted to just OC by bumping up the multiplier, turning off Cool and Quiet, Turbo, Core C6 State, HPC Mode and Apm Master Mode. I bumped up the multiplier by increments of 1 and ran Prime95 for half an hour. No problems and at x20 I was at 4Ghz. Temps after a hlaf hour in Prime 95 were stable 54C, still OKish, all cores fine. So I went back to try and set the Timings on the RAM manually in Ai Tweaker>DRAM Timing Control. It would only let me select one value at a time and then if I moved onto the next it would reset the previous to auto - which of course was cl11. So I set CAS latency to 8, saved and rebooted and CPUz showed Timings of 8-11-11. Went back to BIOS and was then able to only set one again - CAS Delay - to 8. Reboot, CPUz showed the timings as 8-8-11. Back to BIOS and tried to set RAS pre to 8 and it accepted it but reboot still showed 8-8-11. It would go no further. I decided to just revert to all at 11, and find a different route or just leave it and ask others about the problem. However I was unable to manually set the values back. Tried several times and it would accept the input, but on reboot it would be stuck at 8-8-11, BIOS showing the same. So I just went to BIOS, loaded defaults, and I was back to square one having also lost the CPU OC settings, and memory back to 1600 - all cl11.

Now when I go to Ai Tweaker>Ai Overclock Tuner and set it to manual, I can get the 1600 back at 11, reboot and it's fine. However whenever I then try to again bump up the multiplier, no matter what I input it will always revert to the stock 17.5, 3.5Ghz I made sure Turbo was disabled, but there I'm stuck. It will accept the input of a multiplier number but on save and reboot it is back to stock numbers, confirmed in CPUz and rechecking BIOS. Not sure why. My intent is only to go as far as I can with the CPU on stock voltage, hoping for 4.1 or 4.2. If I could get some input on this it would be appreciated.

Thanks

 

[ingtar33]

ok...

you're using the automatic overclocking, not manual. stop using the ai overclock tuner... then you need to turn the cpu frequency from auto to manual, set it to 200, the cpu multiplier from auto to manual and set it to 17.5 (i think), which should put you to 3.5ghz, then turn off turbo mode, turn off quiet and cool, turn off c1 & c6 states, turn off voltage offset mode, and change the cpu and nb voltage from auto to manual. cpu spread spectrum, pciexpress spread spectrum, epu, all set to disabled. set the cpu LLC to ultra high, set the nb LLC to high, disable the VRM spread spectrum. Disable SVM and AMP Master, Enable HPC.

go to the nb frequency, set it to 2400, set the HT frequency to 2600

now you're ready to overclock.

 

[Teemsan]

ingtar33 are you familiar with this exact mobo? In order to manually set values for OC'ing, the Ai OC Tuner HAS to be set to manual. If you read the post you'll see I didn't have it set to Auto - and then try to over-ride those values. It was set first to DOCP to get the correct RAM setting, but that of course doesn't allow you to manually set the CPU values for OC'ing.

DOCP: Select this item and the CPU Bus frequency, CPU ratio, and memory settings are auto optimized.

This was the only setting that had my correct RAM speed and timings.

Then I set the Ai Overclock Tuner to manual to start the CPU OC, which lost the RAM settings.

Manual: Select this item and the CPU Bus frequency, CPU ratio, and memory settings can be customized.

Set to manual, after initial OC, the BIOS wouldn't accept the values I was inputting after I readjusted the RAM - and lost the CPU OC settings. It would allow me to change the number, but then saving and rebooting it would revert to stock settings. I know I had it right before because it had OC'd up to 4.0 stable. And you're right - 17.5 for 3.5Ghz is default. On manual, CPU Bus Frequency is already at 200, and most of the other settings you mentioned I had set - I could try the others. However I can't get there unless I figure out why the BIOS now won't accept my multipliers numbers and hold them, while it did before.

Here's a thread with pics of the BIOS settings form another OC. scroll down

http://www.tomshardware.com/forum/283351-29-asus-m5a97-overcloking

I just realized that the above linked post is in the Motherboards section where I probably should have asked this. How do I move it over?

 

[ingtar33]

yeah... i know that bios. I've got an almost identical bios on this computer (asus m5a99x evo). I'm having issues following exactly what you're doing that's causing the issue though.

1) turn AI overclock turner to manual
2) disable AMD turbo core and all that other stuff i told you to do; make sure the cpu frequency and multipliers are set to stock 200 x17.5
3) page down to dram timing control, enter your ram values by hand
4) save this as an oc profile no.1, and then save and exit.

see if that works.

 

[Teemsan]

OK ... the problem was the BIOS. I understood what you were saying and asked me to do. I just couldn't do it because the BIOS was locking up after entering some values. Only way to get it unstuck was to reset to defaults. As was the case with it not allowing me to manually set DRAM timings or CPU multiplier - or one setting, then locking up. Suspect maybe it had something to do with the BIOS update ??? So I flashed the BIOS, problem solved. Now it will allow me to make changes. So I started at x17.5 3.5Ghz then messed around with various combinations of what you suggested. But if I set both LLC's to high (enable), nb freq to 2400, sb to 2400 (doesn't go to 2600), then C6, Apm Master, Turbo, Cool n Quiet off, it gets hot quick in Prime95 when I bump up the multiplier to anything past about x18.5 (not increasing voltage manually). Like 66C socket temp hot. So I've taken a more modest approach to start. With Cool n Quiet enabled, C6, Apm and Turbo off, both LLC's on Auto, nothing else changed, now at x20 (4.0 Ghz), I'm just one hour into Prime95 and socket temps are staying at 55 \ 56C, core temp 40C. I have 0 errors 0 warnings, all cores passed. Not an OC to write home about, but I'm definitely seeing I have cooling issues. I only used Arctic MX 2 with the 212 EVO, and have 3 case fans, 1 1200 rpm intake, and 2 900 rpm exhaust. Tomorrow I'll try to get some MX 5, some stronger fans, and go at this again. So my initial post was more a BIOS \ mobo issue and I should have posted in Motherboards. Thanks for the suggests and I'll have a go at this again tomorrow.

 

[ingtar33]

yeah... you're having a "standard" problem with the evo. You see airtowers, while really good at their jobs don't handle temp spikes from the cpu well. no air tower does really... this is the advantage a sealed loop cooler has over a tower that doesn't show up in benches.

you see, piledriver is not a core i cpu... which is what most coolers were tested on. Piledriver suffered from rapid temp spikes and changes, which pretty much all air coolers can't really stop. This makes it almost required for serious piledriver overclocking to use a closed loop or DIY water loop. i'd be very surprised if you were able to get that evo cooled fx6300 above 4.4ghz under any circumstance.

that said... on stock vcore you probably can get that fx6300 up close to 4.4ghz... (somewhere between 4.2 and 4.4, depending on the silicon)...
~some thoughts on maximizing your overclock with the evo
-get a 2nd fan for push/pull... this extra horsepower will help the evo overcome those temp spikes
-test it in different orientations... both in the standard "facing back" and "facing up" positions (only if you have vents on the top of your case)
-put 2 80mm or 92mm or one 120mm fan over the northbridge/vrm heatsinks on your motherboard. Cooling those down will give you better overclocking results.
-find a way to put a 120mm fan behind the motherboard pointing at the back of the cpu socket. you can gain 10C just by adding airflow back there (this will likely require a case modification of some sort)

 

[Teemsan]

I have the NZXT source 210 case
http://www.newegg.ca/Product/Product.aspx?Item=N82E16811146075
The problem is that the M5A97 mobo only supports one CPU fan, and three case fans. Like I say one 1200 rpm fan is front intake, one 900 rpm is rear exhaust and one 900 rpm is top left. Of the two exhaust ones, the top one is definitely removing the most heat. I can feel both pushing air out, but the top one exhausts more hot air. The two 900 rpm ones came with molex adapters which of course I didn't use. So I'm thinking I could buy two more stronger fans, and just plug them into a molex connection off of the PSU - daisychain them I can do that right? I guess they would just run at full rated speed and not off of the fan controller. Then I could experiment with more \ better placement. I gotta say the cable management inside the case is really good. I spent a lot of time ensuring every possible thread of wire was tucked away. It's clean in there.

My apt is pretty toasty though. It's winter here in Canada and this apt has s**t insulation. Baseboards low it's freezing in here, baseboards up, it's 26 \ 28C in here. My tower is positioned equidistant between them so it's not in direct heat though. On my old laptop I actually had an 8 inch circular fan blowing air across it and my usb backup drives and it worked pretty well. In summer it's an oven in here, so I'm going to have to figure something out. Probably back to external fans in the area of the tower.

And I'm not at all looking for a crazy OC. My plan was just to bring it to 4.1 or 4.2 if I could get that on multiplier only. I'd even settle for a stable 4.0 I just wanted to give it what the Turbo would, except in OC it'll have all 6 cores always active, which is what I want. The turbo feature is kinda lame imo. So based on results so far, I should be able to achieve that. On idle my Socket temp is about 36 \ 37C already and that seems a bit high to me though. I'd like it down about 5C.

So what do you think of the idea of fans off of molex connectors? This would give me the EVO in push \ pull, and four case fan options. And of course the second EVO fan would be controlled off of one of the mobo connectors.

 

[ingtar33]

you could always get a fan controller. something like the bitfenix recon fan controller, would probably do the job for you.

when i was running my PhII under a 24/7 4.0ghz overclock in the hot phoenix summer without AC, i had my tower air cooler oriented in the "up" position, facing the top of the case and the huge 200mm fan on the top of this antec 902. I also used a push-pull and used a few extra fans inside the case oriented just right to help the airflow along. It was 40C ambient, and the CPU never broke 55C under load (which was important because the CPU heatcrashed at 58C). That settup moved the hot air through the case so well that the metal mesh at the top vent got too hot to touch (you could burn yourself on it).

I've got AC now, and a h100 cpu cooler... but the same case. And i can assure you I'll never underestimate how important good case airflow is again after seeing what this case gives me right now with this piledriver. it's got me hoping i'll be able to keep this 5ghz clock speed in the summer.

i guess what i'm saying is airflow is something many people overlook and unappreciated. Good airflow can make an average cpu cooler or motherboard and make them outstanding. Its also pretty cheap to do right. So i would say, work at it. I know 95% of piledriver cpus out there can hit 4.2ghz on stock voltage. it's just a matter of keeping your temps down when you take a shot at it.

 

[Teemsan]

I added a couple case fans, and that seems to have made a big difference. Idle temps dropped from 34 to 29C. One of them is a side fan which blows air on the northbridge and I think that's making the biggest difference. It's only on the molex connector and so runs full tilt. It's a bit loud so I may look into a fan controller.

I decided to try to the Asus AI Suite OC option. It OC's raising the FSB only. It slowly raised, then stress tested until my system crashed at 4384 Mhz. Then when it rebooted it had selected 4147 Mhz as the stable speed. (FSB 237 x 17.5 Multiplier). CPU voltage is @ 1.237. It is noticeably faster. That is about where I wanted to go to. What I'm thinking is that now I will go back and manually get there through a combo of FSB and multiplier. Don't know if I want to stress this board that much just leaving it as FSB alone, plus it seems like kind of a waste of an OC'able CPU. Now I know that FSB max is 250 - that's CPU @ 4384 (where it crashed) / 17.5 = 250. And CPU multiplier max was x20, where it only got 6 minutes into Prime95 before it got too hot - but that was also b4 the extra case fans. Given those two variables I have a fair bit of room to work with.

One thing is that in the AI Suite , it also auto selects memory speed. It had clocked my memory back to 1263 Mhz cl10. (It's 1600 cl8) Next selection up from 1263 was 1579, so I selected that and rebooted. It was there, but now only at cl11) I can change that to cl8 no prob, but I don't understand enough about the FSB😀RAM ratio. If I select cl8 is it just going to be sending garbage readings because the FSB \ DRAM are now out of sync?

At any rate, the extra case fans made a big difference, so I'm good to go, whichever route I choose.

 

[ingtar33]

mmmm... good questions... but i think you've got it a little tangled up.

CL - is also known as "CAS" it's sorta the backbone of your ram's timings. a lower CL will improve your ram read times. Your FSB doesn't really have much to do with CAS. It's your MEMORY chip who's stability is determaned by such things.

think of it this way... my ddr 1600 ram sports timings of 9-9-9-24-T2, it is completely stable at this speed, with 1.6V (it's stable with 1.55V if i use 2 chips, but by adding 2 more, i need 1.6V to make it stable... yay AMD's weird memory controller). If I wanted to overclock my ram up to 1866, unless the ram is very overclockable, i'd have to do something called "loosening the ram timings"; which would require me to increase the CAS and other timings across the board. Probably take hours of work to find timings it would be stable at. but it would probably be something like 10-10-10-27... but i don't know... it depends on the ram... it also likely would take an increase in dram voltage. Generally losing 1 or 2 points of CAS will completely invalidate a single step up in ram frequency increase, so it takes some experementation.

My ram is completely unable to be overclocked. i know this from 18 months of experimenting. It's very stable at 1600 with the 99924, but drop that CAS down a step or increase the speed more then 100mhz and the whole system will fail to boot.

I found the ASUS overclock tuner tends to pump too much vcore and loosen up the ram timings too much in an effort to improve system stability. So i don't like it. Most piledrivers overclock perfectly well with the multiplier. Furthermore, piledrivers aren't like phenomIIs... for the phenomII the FSB and ram speeds had to be closely played with for overclock stability. It was almost an artform to balance the two. Furthermore PhIIs saw better performance gains (for the most part) from overclocking the ram and northbridge then they did from clock speed improvements because of the crappy memory controller bottle-necking the chips.

Piledriver was a large step forward with the northbridge/memorycontroller... as a result piledriver really doesn't see any gains in benching results with northbridge clocks over 2600... regardless of the ram speed. furthermore the closer you can make your NB and HT match their speeds, the better results. So with piledriver, the ideal settup is 2600/2600 on the NB/HT; and the highest clocks you can get with the highest and tightest (tight meaning lowest CAS, Trcd, Trp, and Tras timings) ram speeds you can get.

When overclocking a piledriver the best way to do it is lower your ram speed down to 1333 or lower, and keep your NB/HT as close to stock (2400/2600) as you can. This takes your memory and the northbridge out of the equation. Once you stabilize the cpu at the clock speed you want, then you bring the ram back up to it's "stock" settings or as close to it as possible. If you overclocked with your FSB/Cpu Frequency, then you might need to add some voltage to the dram and northbridge, vcore, and possibly even the cpu/nb voltage. if you keep mostly to the multiplier your main voltage will be the vcore; and if you are trying to overclock the ram, the dram voltage as well.

 

[Teemsan]

Thx ingtar33, in the end I found the most stable was bumping the multi to x20.5 for 4100, actually undervolted to 1.225 - honestly it was the most stable. I ran 3hrs of Prime95 blend and 2.5 hrs of custom with 75% RAM. I honestly can't see anything I do taxing it that much and I'm satisfied. Temps were good - maxed at 55 socket and 39 core. I was able to get my RAM at it's rated 1600 cl8, by matching FSB Freq and HT Link at 2400, without changing voltages. For some reason that did the trick. All this was after much testing over 3 days. So I'll see how it goes. I may still run an overnight Prime95 tonight. Thanks for all your input.

 

[ingtar33]

looks great!

yeah... i'm not surprised you found better stability lowering the vcore... too much vcore can be as bad as too little for stability. glad to see you were able to get your ram timings nice and tight and your nb/ht at a good stable point. a lot of things deturman where you can go with those settings including the motherboard, your airflow, your psu and cpu.

which is why overclocking is more an artform then it is a science. so many things go into it, it's more about experimenting to find the best set-up then it is about doing x and getting y. Glad you took some time and got it right. I bet you learned a lot more about your bios, your hardware and how it all works together.

=D thanks for the update! love to know people get their stuff working right