• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

SOLVED Strange LLC issue, ASRock 970 Ext3

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

iueras

New Member
Joined
Apr 9, 2008
So here is a strange thing that I can't find much info on and have no idea why it happens: When I enable LLC (setting enabled as opposed to auto) on my Asrock 970 Extreme3 (BIOS version 1.40) with a 955 Phenom II in it, the comp boots and initial CPU voltage appears normal (in fact, it is closer to what I set in BIOS than with it off). but, when I put a load on (such as a prime95 test), the voltage starts to DROP.

There seems to be some kind of rate limit on how fast it can change in either direction, but the voltage goes down... and down... and down. I'm not talking about just a little, I mean on the order of a 0.25V drop. Before I BSOD with LLC on, the lowest voltage I have seen it hit was 1.12V from a BIOS set voltage of 1.3375.

The exact opposite happens when I set LLC to disabled in BIOS. My initial voltages are higher (setting 1.3375 yields a BIOS HW monitor reading of 1.352V), and when I load the CPU with prime95, the voltages go UP. BIOS setting of 1.3375V yields a max voltage under max load of 1.375V as reported by both AIDA64, CPU-Z, and Asrock's tuner utility. All the software has exactly matched what the BIOS HW monitor shows for CPU voltage.

This behavior seems to be exactly the opposite of what I would expect. Has anyone else seen behavior like this?

PS: With LLC off and the load voltages going up, I am prime95 stable after 6 hours (longest I have run it, I don't like testing hardware while I sleep just in case). With LLC on and the voltages dropping under load, I error out on core 4 after 5 seconds to 2 minutes max.
 
We've seen that every board acts differently when it comes to LLC. You did the right thing by experimenting with it a bit. Even different boards by the same manufacturer act different.
You're right though that just seem ludicrous to me!
 
We've seen that every board acts differently when it comes to LLC. You did the right thing by experimenting with it a bit. Even different boards by the same manufacturer act different.
You're right though that just seem ludicrous to me!

You wrote what is about as accurate as can be written. "Storm-Chaser" took issue when I linked an FX cpu discussion of what CPU_LLC would likely cause when used on these cheaper boards with only 3 choices for CPU_LLC. AUTO > DISABLE > ENABLE. That particular user I gave the link to, had a Deneb processor and not FX. However to my mind any discussion of CPU_LLC is relevant to the situations we face today and how odd CPU_LLC can work even on the same brand and model of motherboard but for a different user with a different mobo.

"S-C" did say that I likely could do this stuff 'blind-folded' and that was sort of true. I can do it but I need to see the keyboard to enter the stuff and also to see the captures the users have of HWMonitor. To try and get to the bottom of LLC on AMD motherboards we must go back in history a bit.

Before the release of the FX processor there was NO Vdroop as a specification set forth by AMD. Nada. Zip. A very prominent Vdroop specification came with the release of the FX series of processors. That is why in early bioses used for Deneb and Thuban processors on 9xx chipset boards the CPU_LLC choices for Enabled said for FX processor only in the right hand pane of most early motherboard bioses on 9xx chipset boards using non-FX processor. Heck we had 9xx chipset boards in the market nearly a year before AMD release the first pig of an FX processor.

So the bios writers hid for most who did not look in the right pane that LLC should be off/disabled unless the processor in the socket was an FX processor. This is why "iueras" has better luck with his 955BE in the socket and CPU_LLC actually disabled and the bios voltage set to a voltage that is passing P95 Blend. Heck CPU_LLC Disabled is actually running his pre-FX cpu in the manner it was designed and all the Pre-9xx boards did operate.

Now comes some more of the story. When I tested an Asrock Fatal1ty 990FX motherboard, I found that the Asrock CPU_LLC worked just backwards to what I was accustomed to seeing on my Asus CHV motherboard. I want to state right here that once I understood that the PRIMO Fatal1ty mobo CPU_LLC was working backwards and responded accordingly...well it worked just FINE. I just had to understand that Asrock CPU_LLC called into use at a lesser amount of LLC equaled what calling "more" LLC into use did on the CHV. That is about as 180 degrees from being the same as you can get. However the knowing of how the bios writers interpretted CPU_LLC to function was the key to using either the Asus board or the Asrock board. Sadly the cheaper Asus and Asrock and Gigabyte motherboards do not have a CPU_LLC that can be 'fine-tuned' for use.

So where does all of that really leave us?
1. CPU_LLC is a spec and design parameter for FX processors and not Pre-FX processors.

2. Some cheaper boards with CPU_LLC on them will often work okay with a Pre-FX processor. Some will not and DISABLING CPU_LLC on those boards is to actually run that Pre-FX processor as it was originally designed to be operated. Set manually a Vcore that will allow the non-FX processor to pass P95 Blend and the "loaded" voltage will tend to rise over the voltage set in the bios. That was the way most Denebs ran on earlier chipset motherboards.

3. On the really good boards such as the CHV and Sabertooth and the Asrock Fatal1ty 990FX boards and likely just a few others of the upper echelon AMD 9xx boards, there is a much finer CPU_LLC circuit on the motherboard. This extra adjustability of the more robust motherboards, can likely allow for the 'better/easier' use of CPU_LLC on the Pre-FX processor.

I know this to be true for a C2 stepping 955BE which is not even the better C3 that everyone wants. I used CPU_LLC to get the cpu Vcore correct and run that old C2 955BE at 3.9Ghz P95 Blend stable without undue temps or Vcore. The same was also doable on the Asrock Fatal1ty motherboard I used for a while. Just had to remember the CPU_LLC was operating backwards from the CHV by Asus.

4. So I expect the truth of the matter is that most users with Pre-FX processors might do well to run with CPU_LLC disabled on 9xx chipset motherboards, simply because the Vcroop spec which CPU_LLC is trying to counter-act was not enacted until the FX processor came on the scene. However if the cheaper brand board, no matter the brand can be sorted to run CPU_LLC, then sure use it. But the settting of CPU_LLC is going to be a 'per-user' setting that ONLY trial and error at the keyboard of each system can determine the actual outcome.

Most of this only comes after hours and hours of digging and searching. Seeing early bios screenshots of the early bioses on AMD 9xx chipset boards before later and later bioses were released, to have seen that disclaimer about LLC for non-FX processors. I mean it has been months since this question of CPU_LLC oddity has come up on a Deneb processor installed in a cheaper 9xx chipset motherboard.

H*ll my memory is not what it used to be. I had written most of this down in my "AMD Cpu Notes" or I might not even remember the time-line as it has unfolded. I just had never taken the time to put this in a post since CPU_LLC equal OFF is the setting to use if the cheaper board acts a fool with CPU_LLC on and a non-FX processor is in the cpu socket.

This was not intended to speak to the use of CPU_LLC on better boards with an FX processor in use, but only to give a little insight as to a reason the Pre-FX processor may well be easier to manipulate for stability if CPU_LLC was in fact DISabled. AS in anything computer related, it seems the standard YMMV applies.
RGone...
 
Thanks for the history lesson there RGone. It seems that in this particular BIOS, they have labeled the settings backwards. In BIOS it says to enable LLC for AM3 processors and disable for AM3+.

Also, i don't know if it makes any difference, but the 955 I have is a C3 revision, but it is NOT BE. Somehow I managed to order a non-BE, 95W Deneb 955. But it clocks up to 3.6GHz on air with a $15 tower heatsink at slightly under stock volts without any trouble -- as long as I leave LLC off! :D This coming week I am getting a full sized tower case (GROne) and a Swiftech upgradable AIO water system as my first foray into water cooling. Gonna be a lot of fun!
 
Last edited:
How they say, "you got it Toyota", backwards I can believe out of these bios writers.

Almost as crazy as backwards is how in h*ll you got a non-BE cpu is surely backwards. I saw a few of them but they were few and far between. Hard to get one is what I am trying to say.

Well it is all life in the fast lane anyway. You should have fun with some water for sure. Luck to you man.

I thought of this just as I was fixing to click on submit the post. If I were you I would steer clear of the FX processors as that EXT3 board just lacks heavy enough VRM circuit. Users flooded us with that board and had to upgrade so many it was not funny. That cheap Asrock series is just that > cheap and too cheap to really push an FX. Have fun with what you have and when "steamroller' comes then upgrade the whole thing.
RGone...

Thanks for the history lesson there RGone. It seems that in this particular BIOS, they have labeled the settings backwards. In BIOS it says to enable LLC for AM3 processors and disable for AM3+.

Also, i don't know if it makes any difference, but the 955 I have is a C3 revision, but it is NOT BE. Somehow I managed to order a non-BE, 95W Deneb 955. But it clocks up to 3.6GHz on air with a $15 tower heatsink at slightly under stock volts without any trouble -- as long as I leave LLC off! :D This coming week I am getting a full sized tower case (GROne) and a Swiftech upgradable AIO water system as my first foray into water cooling. Gonna be a lot of fun!
 
That's the plan. Although I think my next upgrade will be to an i7 rather than another AMD, unless there is some compelling reason to go with the new "steamroller". And that's disheartening to hear that this ASRock board is super cheap; I mean, I paid more for the board than the CPU! Right at $100 for the board... that was a big step up from the usual $50-$60 boards I have used for the last 10 years. :(
 
That's the plan. Although I think my next upgrade will be to an i7 rather than another AMD, unless there is some compelling reason to go with the new "steamroller". And that's disheartening to hear that this ASRock board is super cheap; I mean, I paid more for the board than the CPU! Right at $100 for the board... that was a big step up from the usual $50-$60 boards I have used for the last 10 years. :(

Times have changed and 10 years of use and thinking is just not current with AMD and the power they want to use to run super fast with FX processors.

Nothing wrong with an i7 for sure. You should cool it very well at 4.5Ghz with a CM 212 EVO and be quite comfortable. Not so an FX if pushed hard to reach 4.6 to 4.8Ghz and greater. Like I said, enjoy your rig. Next time just search the forums that have some truthful not ticklish users in them and then you will get the parts to begin with. Buying twice is always a pain.
RGone...
 
Back