Positive experience with ACC on Phenom II

[ecat]

ACC does appear to gain me a little extra stability but in tests so far it has not proved worthwhile. eg.

Up the speed on my 3rd core, from observation this is the weakest, until Prime95 fails. The fail is usually a re-boot at this point. Enable ACC -2. Prime95 still fails but on a rounding error, no reboot. ACC = -4 gives the same result, ACC = -6 and it's back to reboot. Hum, it is doing something, intentional or otherwise.

[mattkosem]

ACC does appear to gain me a little extra stability but in tests so far it has not proved worthwhile. eg.

Up the speed on my 3rd core, from observation this is the weakest, until Prime95 fails. The fail is usually a re-boot at this point. Enable ACC -2. Prime95 still fails but on a rounding error, no reboot. ACC = -4 gives the same result, ACC = -6 and it's back to reboot. Hum, it is doing something, intentional or otherwise.

I'm using +2% on my weak core. -2% seems to make my results worse.

With my Phenom II 940BE and Dfi 790FXB i first need to press reset after enabled acc and then it works to boot in to windows but i lose some stabilty with (acc enabled) becuse i geting blue screen next time i restart windows.

Maybe its the board or the cpu.

Since you didn't mention having one bad core, and didn't mention enabling ACC on it and it alone, it sounds like you're going about trying ACC with the wrong approach. Have you tried OCing core by core, and run into one core that clocks worse than the others?

--Matt

[Lastviking]

I'm using +2% on my weak core. -2% seems to make my results worse.



Since you didn't mention having one bad core, and didn't mention enabling ACC on it and it alone, it sounds like you're going about trying ACC with the wrong approach. Have you tried OCing core by core, and run into one core that clocks worse than the others?

--Matt

I can do super pi1m with 1.47v(True Black air temp 25c)

Core0=4040mhz
Core1=4060mhz
Core2=4040mhz
Core3=~3950mhz

OCCT no problem @ 3.8ghz on the good ones

Have problem with Core2/Core3 changing place sometimes when i´m restarting windows(dunno know why)

ACC setting i have tryied(have done more tests before also)

ACC Enabled(Per core)

Core0=-+0
Core1=-+0
Core2=+2(have also tryied 0 here)
Core3=+2

OCCT still crash @ 3.8ghz

And super pi1m cant run @4.0ghz @ core3

[mattkosem]

I can do super pi1m with 1.47v(True Black air temp 25c)

Core0=4040mhz
Core1=4060mhz
Core2=4040mhz
Core3=~3950mhz

OCCT no problem @ 3.8ghz on the good ones

Have problem with Core2/Core3 changing place sometimes when i´m restarting windows(dunno know why)

ACC setting i have tryied(have done more tests before also)

ACC Enabled(Per core)

Core0=-+0
Core1=-+0
Core2=+2(have also tryied 0 here)
Core3=+2

OCCT still crash @ 3.8ghz

And super pi1m cant run @4.0ghz @ core3

Bummer. I haven't had any luck with values over 2, but you could give 4 a shot. *shrug*

--Matt

[chris.y2k.r1]

folding is a different story. That probably is the hardest legit task since it goes on non-stop doing as many calculations as the cpu can handle (not even gaming will be close to the stress). However, for a person that does not fold, anything that lasts an hour in a stability test is more than enough

I could not disagree more - For as long as you run OCCT it can fail. I've seen it fail @ 1 minute, 1 second, 1 hour, 1 day, 5 hours ....on and on... When your system is right it will run anything forever - when it's not it won't period. If you think OCCT 1 hour is good enough that you are going to crash at some point whether it's gaming or whatever. Unless you got really lucky with your settings, 24/7 stability cannot be assumed at all after 1 hour of OCCT. Maybe after 30 hours you might be alright. I use tests of all sorts. Sometimes OCCT will runs for hours and SPI 32M will crap out at iteration 4.

[chris.y2k.r1]

I have a question - If ACC does nothing for Phenom IIs then why is it installed in AM3 boards? AMD is being tight lipped about this for a reason. I suspect it's something like the TLB issue.

[chew*]

I have a question - If ACC does nothing for Phenom IIs then why is it installed in AM3 boards? They can't run Phenom I's anyway - So that's just straight BS. AMD is being tight lipped about this for a reason. I suspect it's something like the TLB issue.

Future proofing? Sort of like sideport on ati cards...its there but it doesn't work. No clue really.

[mattkosem]

Future proofing? Sort of like sideport on ati cards...its there but it doesn't work. No clue really.

As I've been saying, it definitely works. It may not be as drastic of an improvement as it was on PHI, but it does make an improvement. It also seems to need additional increases as the clock and/or voltage is increased. I had to bump it up to 2%/2%/4%/4% at 1.6125v for 3.82Ghz on my M4A79 Deluxe. I require 4%/4%/4%/4% at 1.6375 to bench at 3.92Ghz. Asus actually fixed it in the 1102 bios for this board. It was very apparent that it wasn't working in the 90x and 100x bios builds.

--Matt

[chew*]

As I've been saying, it definitely works. It also seems to need additional increases as the clock and/or voltage is increased. I had to bump it up to 2%/2%/4%/4% at 1.6125v for 3.82Ghz on my M4A79 Deluxe. I require 4%/4%/4%/4% at 1.6375 to bench at 3.92Ghz.

--Matt

Yes i know it works in your case for AM2+ matt, its also included in am3 board bios's and the for the record chat is " it does nothing for am3 " but it's still being coded into the bios. Its apparently doing something though as it is enabling dead cores on AM3.

Your chip seems to be an anomally, like one rare case of it making a diff.....and actually getting measurable performance increases when enabled.....its possible your chip made it through binning that it shouldn't have somehow with a weak core.

[Smartidiot89]

Yes i know it works in your case for AM2+ matt, its also included in am3 board bios's and the for the record chat is " it does nothing for am3 " but it's still being coded into the bios. Its apparently doing something though as it is enabling dead cores on AM3.

Your chip seems to be an anomally, like one rare case of it making a diff.....and actually getting measurable performance increases when enabled.....its possible your chip made it through binning that it shouldn't have somehow with a weak core.

Have AMD actually said that ACC doesn't work with Phenom II?

It's not impossible it does actually help, I have yet to see extensive testing proving the opposite, or a database of people saying it didn't work.

And why wouldn't hes chip make it through the binning? It is stable at stock, I don't think AMD cares if the chip can overclock to 3,5ghz or 3,7ghz as long as stock values (and lil bit above) is working fine.

[chew*]

Have AMD actually said that ACC doesn't work with Phenom II?

It's not impossible it does actually help, I have yet to see extensive testing proving the opposite, or a database of people saying it didn't work.

And why wouldn't hes chip make it through the binning? It is stable at stock, I don't think AMD cares if the chip can overclock to 3,5ghz or 3,7ghz as long as stock values (and lil bit above) is working fine.

Yes they have thats why I have said what i said....I say made it through binning becasue without acc on his performance is somewhat hindered.

[keithlm]

Have AMD actually said that ACC doesn't work with Phenom II?

As far as I know: The ONLY source on this was a third party reviewer at a website that claimed that ACC was built into the PHII chips and it was not needed or used. He claimed that his source at AMD confirmed that.

BUT when you have a few bios updates that specifically mention fixing ACC for PHII chips that kind of contradicts the reviewer. Plus not having any concrete information from AMD doesn't help.

Perhaps the reviewer's "source" at AMD meant that it wouldn't work with the bios that was then available and he interpreted that in his own way. (And of course the way the internet works... pretty soon other sources were quoting him.)

[mattkosem]

This guy seems to be having pretty good luck with ACC on his PHII as well: http://forums.ncix.com/forums/?mode=...t=15&subpage=1

--Matt

[chew*]

This guy seems to be having pretty good luck with ACC on his PHII as well: http://forums.ncix.com/forums/?mode=...t=15&subpage=1

--Matt

Matt your missing the big picture. 2 people so far with good accounts.......out of how many? I still say anomally.

My sources are not stuff you guys read on the net

[mattkosem]

Matt your missing the big picture. 2 people so far with good accounts.......out of how many? I still say anomally.

My sources are not stuff you guys read on the net

I don't think the big picture is worth that much in this case. Your sources must not be folks at AMD or at any of the mobo manufacturers that have added improvements for ACC in recent bios builds. I'm sure they wouldn't bother with making these improvements if they were all for naught. If it works for two it can work for others. The "ACC Doesn't Do Anything on Phenom II" attitude is preventing people from giving it a shot in the first place.

--Matt

Edit: here's my latest achievement, thanks to a bit more voltage and some additional ACC tweaking.

[soundood]

I don't think the big picture is worth that much in this case. Your sources must not be folks at AMD or at any of the mobo manufacturers that have added improvements for ACC in recent bios builds. I'm sure they wouldn't bother with making these improvements if they were all for naught. If it works for two it can work for others. The "ACC Doesn't Do Anything on Phenom II" attitude is preventing people from giving it a shot in the first place.

--Matt

Edit: here's my latest achievement, thanks to a bit more voltage and some additional ACC tweaking.

why does the AM3 boards have ACC in the SB750 SB?

[mattkosem]

why does the AM3 boards have ACC in the SB750 SB?

It doesn't seem to be known at the moment. There are reports that it isn't needed for AM3, but there are also similar reports about the AM2+ Phenom IIs. I'll certainly give it a try when the 955 comes out though!

--Matt

[chew*]

Matt my source is AMD, I don't know what else to tell you.....it seems to work on a couple chips.....buy 4 more 940's and do a comparison....best I can tell you.

[G0ldBr1ck]

I would guess that in some cases where ACC creates more stability its due to a shortcoming of the MB. When ACC is enabled from what I understand, the clock is generated off of a diffrent PLL and accesses the die by a different circuit. If the boards default clockgen and circuits are not up to 100% perfect in function and the ACC southbridge clockgen is, it could make a large impact on stability.

I would also speculate that its threw this other link from the southbridge that is accessing the fused off cores and cache that aparently were only fused off from the default link.

This of course is all just speculation.

[mattkosem]

Matt my source is AMD, I don't know what else to tell you.....it seems to work on a couple chips.....buy 4 more 940's and do a comparison....best I can tell you.

I surely won't buy any more 940s with 955 right around the corner. My 940 does 3.82Ghz 24/7 stable with the help of ACC. Aside from having you test this processor yourself, there's nothing else that I can really do to prove this to you (if you really want to see it, I'd entertain an X3 920BE to x4 940BE swap). Trust me though, it does actually help in some cases. It helped on two different boards with this chip. It does seem to have one or two cores that aren't as strong as the rest but ACC definitely makes a difference, even on the stronger cores (as can be seen in the 3.825ghz shot).

--Matt

[chew*]

I surely won't buy any more 940s with 955 right around the corner. My 940 does 3.82Ghz 24/7 stable with the help of ACC. Aside from having you test this processor yourself, there's nothing else that I can really do to prove this to you. Trust me though, it does actually help in some cases. It helped on two different boards with this chip. It does seem to have one or two cores that aren't as strong as the rest but ACC definitely makes a difference, even on the stronger cores (as can be seen in the 3.825ghz shot).

--Matt

Oh i believe you matt 100% your cinebench was more than enough to convince me that its definitely helping your chip, by quite a bit I might add especially performance wise, It's just not supposed to make a diff and I have no explanation why its working for you.....I have same board, it kills all my 940's performance. I tried duplicationg results

[chris.y2k.r1]

I would guess that in some cases where ACC creates more stability its due to a shortcoming of the MB. When ACC is enabled from what I understand, the clock is generated off of a diffrent PLL and accesses the die by a different circuit. If the boards default clockgen and circuits are not up to 100% perfect in function and the ACC southbridge clockgen is, it could make a large impact on stability.

I would also speculate that its threw this other link from the southbridge that is accessing the fused off cores and cache that aparently were only fused off from the default link.

This of course is all just speculation.

Sounds like it's reasonable to me. I was curious as to why there's no PLL control for AMD chips - and I like how this explanation fits the reasoning for the 4th core activation. I guess the question is how does AMD typically deactivate cores? Is it by not sending a loop?

[Smartidiot89]

As far as I know: The ONLY source on this was a third party reviewer at a website that claimed that ACC was built into the PHII chips and it was not needed or used. He claimed that his source at AMD confirmed that.

BUT when you have a few bios updates that specifically mention fixing ACC for PHII chips that kind of contradicts the reviewer. Plus not having any concrete information from AMD doesn't help.

Perhaps the reviewer's "source" at AMD meant that it wouldn't work with the bios that was then available and he interpreted that in his own way. (And of course the way the internet works... pretty soon other sources were quoting him.)

Yepp, so unless I see a statement and a source directly from AMD I won't believe otherwise.

ACC might not help as much as it did to Agena (K10 65nm), but it could be working to some extent.

The fact that many motherboard manufacturers just "fixed ACC" many have already forgotten about this function and have not bothered to try it with new BIOSes and to fully take advantage of ACC actually requires loads of patience, as in trying to find out which core is weak, before using this function.

I am not calling you a liar Chew* but unless you can show me a concrete source from AMD, I won't be the one saying that Phenom II isn't supported by ACC "unless a defect passed through binning" which isn't a viable theory at all to me.

[chew*]

Yepp, so unless I see a statement and a source directly from AMD I won't believe otherwise.

ACC might not help as much as it did to Agena (K10 65nm), but it could be working to some extent.

The fact that many motherboard manufacturers just "fixed ACC" many have already forgotten about this function and have not bothered to try it with new BIOSes and to fully take advantage of ACC actually requires loads of patience, as in trying to find out which core is weak, before using this function.

I am not calling you a liar Chew* but unless you can show me a concrete source from AMD, I won't be the one saying that Phenom II isn't supported by ACC "unless a defect passed through binning" which isn't a viable theory at all to me.

The defect is my analogy for it as there is no logical explanation for it.....call it a hypothesis if you will.........

As far as my source.....laughable?

You seriously can't just expect me to drop a name or drag someone in here........take or leave it for what its worth....but ACC is not designed for PH II and PH II is not designed to work with ACC........whatever the bios guys and motherboard manufacturers are doing I have no clue.....makes little sense to me.....I can see patches for existing support for am2+ boards making sense........the AM3 boards are making no sense......I believe most of the patches are killing the ability to use the extra core though so that should tell you something in itself.

Btw yes ACC enables a core on my asus with a specific bios, disable ACC and the chip runs smooth as butter and is very responsive......enable it and I have to fight with the board the entire time to do what I want to do.......ya ya its a bad core.....does more than my 940 so it's not that bad ......it at least boots over 1.6v on air.

[Smartidiot89]

The defect is my analogy for it as there is no logical explanation for it.....call it a hypothesis if you will.........

As far as my source.....laughable?

You seriously can't just expect me to drop a name or drag someone in here........take or leave it for what its worth....but ACC is not designed for PH II and PH II is not designed to work with ACC........whatever the bios guys and motherboard manufacturers are doing I have no clue.....makes little sense to me.....I can see patches for existing support for am2+ boards making sense........the AM3 boards are making no sense......I believe most of the patches are killing the ability to use the extra core though so that should tell you something in itself.

Btw yes ACC enables a core on my asus with a specific bios, disable ACC and the chip runs smooth as butter and is very responsive......enable it and I have to fight with the board the entire time to do what I want to do.......ya ya its a bad core.....does more than my 940 so it's not that bad ......it at least boots over 1.6v on air.

No I don't expect you to drop a name, but unless AMD gives out info in public clearly stating ACC isn't working A link from a trustworthy website with a "anonymous" source from AMD would be enough for me.

Also that the new BIOSes doesn't allow people to activate the 4th Core, is because AMD kindly requested all motherboard manufacturers to make this change to ACC, cause they didn't like the idea of people buying X3's and unlocking them, I thaught everyone knew about this? So no it doesn't tell me something in itself.

And you said it yourself you have no idea what the AM3 manufacturers are doing implementing ACC.

And as we can see ACC is working for this guy, he does gain stability from it, hes clock was unstable until he used ACC.