Phenom II: Unlocking Cores, Cache, And A Free Lunch

[Shadow703793]

[citation][nom]cangelini[/nom]An hour, four-threads, FFTs before moving on. That's no guarantee of absolute stability, but it was "good enough" for getting the unlocks to fail.[/citation]
Only an hour? Most consider at LEAST 6 hrs to call it stable.

 

[megamanx00]

Honestly, AMD should just leave the SB750 microcode alone, at least as far as the ACC goes. I suppose I can see them being worried about some system builders buying the 720BE and then trying to sell a machine on it telling the customer it's a quad core only to have it crash and what not. Even so it's a good idea to keep overclockers and enthusiasts happy. I suspect that hacked BIOS builds will be popping up on the net soon anyway for boards with the SB750.

 

[ailgatrat]

[citation][nom]bustapr[/nom]I don't know much about cpu things and unlocking cores, but it looks as if a x3 with a disfunctional fourth core works better than a x4 with four perfectly good cores. Please correct me if I'm wrong. But good article.[/citation]
[citation][nom]neiroatopelcc[/nom]That is not the case. 810 is clocked at 2.6ghz per core, and 720 is clocked at 2.8ghz default. Thus if you use the same number of cores on both cpus the 720 will come out on top.[/citation]
[citation][nom]bustapr[/nom]Yes I would agree with on that, but you have to remember that its not just the same amount of cores, but one authentic quad core and one triple core with a disfunctional fourth core. The logic would be the that the part of it being disfunctional should have been able to change it all.[/citation]

Don't forget that the x3 720BE (Black Edition) also comes with an unlocked multiplier where as the x4 810 does not. So you can overclock this cpu to a certain degree without voiding the warranty. Or am I wrong?

 

[ailgatrat]

Anybody else notice the time is way off for US time zones? Looks like UK time zones. I had to look at the source of the article to make sure I wasn't imagining things...lol! Just mention it cause I just noticed that my post time is 7 hours off local time.

 

[neiroatopelcc]

[citation][nom]joeman42[/nom]I believe i7 has the capacity to deactivate idle and dynamically OC its cores in response to scenarios like this, yes? Anyhow is it possible to programatically disable the "free" core? This might make an interesting utility to code.[/citation]
Haven't got the finances at the moment to play with it myself, but far as I've read the 'turbo' feature of an i7 system overclocks the core in use if running singlethreaded stuff. But obviously that's not a favorable event if you've already overclocked your system to its limits - so it's irrelevant to most thg users. I don't know if it can disable cores completely, but i7 can disable certain parts of the circuitry at least. I'm a bit under the weather atm though, so perhaps I know more in a few days when I'm back not running in failsafe anymore.
Anyway, you can dictate how many cores to use on windows boot - so in a sense you can disable free cores. But hardly anyone'd want that ? The only use I can see for disabling a core is if you're using a wds server and need to load a uniprocessor system onto a dualcore system - and then that beforementioned system won't help.

 

[Stingingfork]

I've got a Biostar TA790GX Motherboard and a Phenom 2 X3 720 BE CPU. I had no asperations of trying to get a fourth core from the three core processor but after reading the article on Tom's Hardware concerning their attempts at unlocking the fourth core with limited success I decided to enable the ACC in my bios and instantly after rebooting I came up with what appears to be a stable fourth core. I ran Cinebench for establishing a bench mark between the two cores (three and four) and here are the results:

Three Cores:
Rendering (Single CPU): 3226 CB-CPU
Rendering (Multiple CPU): 8582 CB-CPU

Four Cores:
Rendering (Single CPU): 3177 CB-CPU
Rendering (Multiple CPU): 11533 CB-CPU

The change appears to be completely stable at this point but could someone suggest some other benchmark tests that would push the limits of the current four cores to check on it's future stability? At this point it recognizes the CPU as AMD Phenom II X4 20 Processor. I've never been one in the past to run benchmark tests on my new builds. Based on the article I'm not sure if the change from the three to four core processor provides me any added benefits but I'm tempted to leave as four if there isn't any stability problems. What do you think?

 

[cangelini]

Stinging,

Glad you were able to get your 720 working. Try running four threads of Prime95 for several hours to see if the four cores run stably together.

As you can see from our benchmarks, there are a number of different applications that'll demonstrate a benefit to that last core. While it certainly depends on what you run, even an app like AVG anti-virus will yield a speed-up if the configuration is indeed stable.

 

[Stingingfork]

Cangelini,

Thanks for the advice. I've located Prime95 (Win 64 bit) and I've set it to run for two hours as a blended test. I'll follow up later with the results but just as a side note I installed and ran Sisoft Sandra 2009 and it referenced it as a AMD Phenom II X4 940.

 

[apache_lives]

[citation][nom]Shadow703793[/nom]Only an hour? Most consider at LEAST 6 hrs to call it stable.[/citation]

i get the feeling most people dont know how to test hardware properly - most important for new builds even if not overclocking/modding!

 

[neiroatopelcc]

I run prime for a maximum of 2 hours unless I suspect a stability issue. Usually after 30 minutes, if it's still running it'll keep running (when the enviroment temp has risen due to the heat level).
So I'd start pime95 and watch a movie or something, and that's it. While actually trying to find oc limits I'm only running a couple minutes before rebooting and trying higher settings.

 

[Stingingfork]

After reading apache and shadows comments concerning the run time to verify stability I've left the unlocked CPU running Prime95 for just at nine hours with the self tests passing for both worker #1 and #2 at each stage without any blue screens or failures for the complete time. Since Sisoft Sandra 2009 recognized it as an AMD Phenom II X4 940 when I ran it last night and stablility doesn't seem to be an issue at this point would you leave it unlocked and if I do will this void the warranty if down the road it fails within the warranty period? Also if I ever disable the ACC in my bios will it retain the original three cores it started with?

 

[fudgeboy]

wheres the free lunch!?

 

[neiroatopelcc]

[citation][nom]Stingingfork[/nom]After reading apache and shadows comments concerning the run time to verify stability I've left the unlocked CPU running Prime95 for just at nine hours with the self tests passing for both worker #1 and #2 at each stage without any blue screens or failures for the complete time. Since Sisoft Sandra 2009 recognized it as an AMD Phenom II X4 940 when I ran it last night and stablility doesn't seem to be an issue at this point would you leave it unlocked and if I do will this void the warranty if down the road it fails within the warranty period? Also if I ever disable the ACC in my bios will it retain the original three cores it started with?[/citation]
Technically the warranty is void simply because you tried to unlock it, but the thing is, amd neither can nor try to check if that was the reason it broke if it did. That said, I haven't seen a cpu fail since socket a.

 

[Guest]

I use a 720Be & ASRock M3A790GXH/128M with BIOS v1.10. This version of BIOS was updated at April 22, 2009. But I can use 720BE with 4-core by converting ACC to Auto. (I could use 720BE with 4-core at BIOS v1.0) of course, I tested my computer by Prime95 during 4 hour. My computer passed it.

 

[Guest]

I have a 720BE on Gigabyte's GA-MA790FXT-UD5P (bios ver: F3L, F3K, F5) and the 4th core unlocked. Running stable (12+ hours Prime95) @ 3.3GHz so no problem with the cores.

Two things I want to comment related to the article.

1) Board manufacturers (and/or AMD) that are more likely to "block the unlock":

With the first BIOS versions for my board, you had to enable the ACC to unlock the cores but since F3K version Gigabyte introduced a special option called "hybrid mode". This mode will unlock the 4th core regardless of the ACC setting. Seems to me that this board manufacturer is not trying to block the 720BE -> x4 transformation. On the contrary the F5 BIOS states: "Update ACC code and CPU ID (Support AGESA 3.3.2.4 code for Phenom X3 CPU)"

2) About the failed attempts to unlock the 4th core:

I noticed that Vista x64 was used by Chris in his tests. From MY experience with MY 720BE & board I can say that unlocking & overclocking this CPU on 64-bit OS is more difficult. For my board, the first BIOS that supported the ACC unlock was the F3B but that one refused to boot under Vista x64. I read about successful unlocks on Win XP or Vista but on 32-bit with the same BIOS version so I thought I had a "defective" 4th core on my CPU. But I just had to wait until the F3L BIOS version to successfully unlock my CPU under Vista x64. Also I read on various forums about guys that could overclock better the 720BE on 32-bit OSes than on 64-bit ones.

 

[isamuelson]

I must be lucky. I have a Gigabyte GA-MA78LM-S2H motherboard and I purchased a boxed Phenom II X3 720 BE and unlocked the 4th core without a problem. Running Prime95 showed no problems with the CPU and it definitely made a difference with applications such as 3dMark06/Vantage and GTA IV.

Running 3 cores, not overclocked, 3dMark06 gave the CPU a 3232 score.

Running 3 cores, overclocked by increasing multiplier to x16, the CPU score was 3342.

Running 4 cores, not overclocked, the CPU score was 3970.

Definitely makes a better difference than overclocking at this point for me.