Overclocking an old AMD CPU (Brisbane). Any advice?

[pmv]

Was tempted to add this to the thread by azemsiB whose question is almost exactly the same as mine, but I'm assuming hijacking someone else's thread is frowned upon.

I'm doing almost exactly the same thing - trying to overclock for the first time, a 5600+ athlon64 x2. Only difference is this is Brisbane core (2,9Ghz) not Windsor.

Asus m2n motherboard.

Am using freezer 7 pro cooler, which I know is the cheapest non-stock cooler you can buy so presumably isn't great, but it was very cheap and _does_ give me temps nearly 20C lower under load compared to the stock (@ stock speeds - temps were hitting 60C now are hardly above 40C) so it surely has to be good for some degree of overclocking?

Also, have good case cooling with 4 12cm+1 8cm case fans.

Have installed prime95, want to be sure any o/c is stable. Firstly, which option for 'torture testing' in that program do I use? How long should I run it for? One source on the web says to only start testing with prime95 once you reach a 10% overclock, but 10% is pretty much my target o/c, because I want to get the RAM to the point where its running at the correct speed.

Because of the way the multiplier works for Brisbane cores by default the memory is significantly underclocked at normal bus speeds, by my calculations I can increase the bus to 220 (and CPU to 3.2Ghz) before the memory even reaches its 800mhz rated speed. So that's my target. Given which I'm assuming I don't really need to worry about testing the RAM?

So far I've only tried a very timid 203Mhz. My hope is to at least be able to overclock enough that the memory is running at its rated speed - which would mean 220mhz bus.

Is it really essential not to let HT speed go over 1000? At the moment with the bus of 203 its very slightly overclocked @1015 - does prime95 test this factor?

 

[RyanGreener]

You should have thread jacked. The more information the better

Sorry for the useless post.

 

[jvroig]

Have installed prime95, want to be sure any o/c is stable. Firstly, which option for 'torture testing' in that program do I use? How long should I run it for? One source on the web says to only start testing with prime95 once you reach a 10% overclock, but 10% is pretty much my target o/c, because I want to get the RAM to the point where its running at the correct speed.

Just go to 220 and test it from there.
Prime 95 Small FFT - tests CPU overclock
Prime 95 Large FFT - tests the CPU-NB (integrated memory controller) overclock and some RAM
Prime 95 Blend - mostly a RAM test, but also sensitive to CPU-NB overclocks. You are better off using Memtest HCI instead of Blend.

How long depends on you. You can run it for 1 hour, 3 hours, 6 hours, 12 hours, or 24 hours. Whatever makes you comfortable. You certainly won't find a single unanimous answer here. The longer, generally the better, but unless you've got time to waste, balance it with your regular use case.

It also won't hurt to also test it with other stress tests such as OCCT and LinX.

Given which I'm assuming I don't really need to worry about testing the RAM?

Download Memtest HCI and test anyway. It doesn't have to take you more than an hour. Just better to be sure, since you are overclocking your machine anyway, it won't hurt to be sure your RAM will work at its new settings.

Is it really essential not to let HT speed go over 1000?

Overclocking the HT Link produces no benefit, but may contribute to instability, which is why it is recommended to lower it. Just lower the multiplier by 1 and forget about it.

 

[pmv]

Thanks. OK, I guess I'll try going straight to 220 and running prime95 overnight. I guess if it fails its a case of tweak the voltage slightly and do it again?

At first look couldn't find a BIOS entry for HT multiplier, instead it just gave a choice of absolute figures of 1000, 800, 600 etc.

 

[jvroig]

At first look couldn't find a BIOS entry for HT multiplier, instead it just gave a choice of absolute figures of 1000, 800, 600 etc.

That's the multiplier. Some BIOSes are irritating like that. My Gigabyte board shows a proper multiplier, but my MSI board doesn't. Just treat the figures as the corresponding multipliers relative to a stock base clock (200). So the 1000 is actually x5, 800 is x4, 600 is x3, etc.

Just set it at 800 and you're good.

I guess if it fails its a case of tweak the voltage slightly and do it again?

Yes.

 

[VirtualLarry]

If your existing CPU is already at 2.9Ghz, there may not be much more left in it. Maybe 3.0, probably not more. I wouldn't bother, I would upgrade the CPU instead, especially if you have an AM2+ mobo, that has had it's BIOS updated in the last year.

 

[pmv]

If your existing CPU is already at 2.9Ghz, there may not be much more left in it. Maybe 3.0, probably not more. I wouldn't bother, I would upgrade the CPU instead, especially if you have an AM2+ mobo, that has had it's BIOS updated in the last year.

That's the problem - this motherboard won't support anything beyond AM2, unlike some other M2N motherboards. I'm not sure if those other motherboards were available when I got this one, if they were I'm irritated with myself for not getting one of them as then I could have easily upgraded all the way to an athlon ii quad.

A new cpu means a new mobo (and possibly consequently new DDR3 RAM) and a rebuild and new windows and possibly a new psu. At which point I might as well build a new PC.

Is there something significant about 2.9ghz? Is 3Ghz known as a point its hard to get beyond?

 

[pmv]

If your existing CPU is already at 2.9Ghz, there may not be much more left in it. Maybe 3.0, probably not more.

Seems you might be right. Can get 3.05Ghz (stable 8 hours of prime95) with stock voltage but 3.1G or higher and prime95 falls over. BIOS only offers 3.125V and 3.15V as higher voltages but the former makes no difference and when I tried the latter I got a BSOD soon after boot.

Rats.

 

[maniac5999]

Well, if it makes you feel better, a 3ghz K8 x2 still isn't anywhere near obsolete. I'm running a Kuma 7750 (K10 x2) which is probably only 10-15% faster per clock in games at 3ghz, and for what I play, (SC2 @ 1080p, almost everything maxed) I'm happy with it. (And YES, I play as Protoss, and YES my favorite strategy is carriers)

Mine doesn't seem to have much left in it either. I suspect that AMD's 65nm process just isn't good for much over 3ghz period. Athlons, Phenoms, whatever, nobody seems to report much over 3ghz. the best I've heard has been about 3.25ghz with some of the Kumas, but they have the advantage of being x4s with the two worst cores shut off.
K8 was a champ, when it came out. and is still respectable, but was never known for it's overclocking headroom either. Heck, back at 130nm Pentium 4 was easily blowing past 3ghz, but the Athlons (K8) at 2ghz were wiping the floor with it.

 

[DrMrLordX]

My x2-3600+ would do, at most, 3.2 ghz with 1.6v vcore. Running the voltage that high was a hair-raising experience: entire temp sensors would disappear in Speedfan, for example (only to return once the vcore was lowered). 3.1 wasn't too hard, but 3.2 was just asking too much from the chip.

I believe the 6000+ could hit 3.5ghz for some dedicated overclockers, but it wasn't an everyday sort of thing. 3 ghz is pretty respectable as is.

 

[maniac5999]

My x2-3600+ would do, at most, 3.2 ghz with 1.6v vcore. Running the voltage that high was a hair-raising experience: entire temp sensors would disappear in Speedfan, for example (only to return once the vcore was lowered). 3.1 wasn't too hard, but 3.2 was just asking too much from the chip.

I believe the 6000+ could hit 3.5ghz for some dedicated overclockers, but it wasn't an everyday sort of thing. 3 ghz is pretty respectable as is.

Haha, my Kuma's about the same, 3ghz @1.2875v. 3.1ghz @1.3125v, 3.12ghz requires 1.3375v, 3.2ghz? not stable at 1.5v

 

[DrMrLordX]

65nm process, RIP. We don't miss you! Much.

 

[cusideabelincoln]

The memory to bus ratio will change depending on the CPU multiplier. If you set your CPU multiplier to 14 and your bus is at 200, then your memory should run at DDR2-800 speeds. but of course you'll only be at 2.8 GHz. I doubt this helps you but I just thought I'd let you know if you didn't.

Anyways, don't expect much over 3.1 or 3.2 GHz with these chips. That is about as far as they'll overclock for everyday use.

And I have my HT running at 1100. I haven't seen any negatives to this so far, but I am using a motherboard capable of higher HT speeds which yours wasn't intended to do.

 

[nyker96]

man it's been so long since I had my brisbane, I believe I had one forgot the model and speed. but I did remember it did oc well and ran cool by yesterday's standards. I think I sold it either that or still in some box in closet. gosh, been too long.

 

[pmv]

Is there any particular reason why this stupid motherboard (ASUS M2N, plain vanilla M2N) won't support anything better than AM2? I notice the M2N-SLI-Deluxe does as does the M2N-E and several other M2N types with BIOS updates, but not this one. Also I was sure I saw motherboards by other manufacturers with the same (chipset nforce430) that say they support AM2+/AM3. Went from 3800 single core to this, and...dead end.

Is it a hardware limitation or does ASUS just not want to bother giving it a new BIOS?

(I'm baffled now why I didn't go core2 duo in the first place. Must have been out of my price range at the time.)

 

[LoneNinja]

I still have a Brisbane X2 5000BE, 3.1Ghz was easy, 3.2 not even with 1.5V. It's been in my closet for nearly a year now, I had replaced it with a Kuma X2 7750BE, which I than replaced with a Phenom X4 9850 that I still use.

 

[heyheybooboo]

Even having the *8* divider I would try to OC with the memory ratio dropped from 400 to 333 until you find the max speed of the mobo. Though your logic would seem reasonable on your RAM speed, I think the 'technical' issue is the IMC and how it handles DDR2.

Even though your RAMs are running 2900MHz/8 x 2, when you up the clock to 220MHz the IMC has been OC'ed 10%. For whatever reason, dropping the memory ratio from 400 to 333 reduces the stress on the IMC and allows you to bump the clock.

Drop the CPU multi and see if your mobo can handle the higher clock speed --- try 12x240MHz.

If that doesn't work give 12x235MHz a shot.

Once you find the max *bus* you may then experiment with bumping the CPU multi. Once you reach the max you are comfortable with the last thing you want to tighten up is your ram speed.

If you are lucky maybe you can hit something like 13.5x235MHz with a little volt bump (which I think may be the over-riding factor in all of this).

There are so many Asus m2n mobos, many of which have chipsets which are archaic by today's standards. I suspect your version is in the lot of those that only run 65w CPUs and are really volt-limited.




--

 

[pmv]

Even having the *8* divider I would try to OC with the memory ratio dropped from 400 to 333 until you find the max speed of the mobo. Though your logic would seem reasonable on your RAM speed, I think the 'technical' issue is the IMC and how it handles DDR2.

Even though your RAMs are running 2900MHz/8 x 2, when you up the clock to 220MHz the IMC has been OC'ed 10%. For whatever reason, dropping the memory ratio from 400 to 333 reduces the stress on the IMC and allows you to bump the clock.

Drop the CPU multi and see if your mobo can handle the higher clock speed --- try 12x240MHz.

If that doesn't work give 12x235MHz a shot.

Once you find the max *bus* you may then experiment with bumping the CPU multi. Once you reach the max you are comfortable with the last thing you want to tighten up is your ram speed.

If you are lucky maybe you can hit something like 13.5x235MHz with a little volt bump (which I think may be the over-riding factor in all of this).

There are so many Asus m2n mobos, many of which have chipsets which are archaic by today's standards. I suspect your version is in the lot of those that only run 65w CPUs and are really volt-limited.




--

Thanks.

I get what you are saying...reducing the multiplier means checking the mobo independent of the cpu. Plus seeing if memory controller speed makes any difference. Will give it a go when I can face leaving the thing running prime95 for ages to check it again, but its looking like it might not be worth the effort for what will probably be a small overclock!

Odd thing is, this mobo _will_ accept 125W Windsor core AM2 CPUs. Its not the wattage it just has no BIOS available to support AM2+ or AM3. The majority of M2N variants listed on Asus's site now have such a BIOS, but not this one. Just wondering if its a hardware limitation (and hence no such BIOS will ever be available, even a hacked one).

 

[WT]

I still have a Brisbane X2 5000BE, 3.1Ghz was easy, 3.2 not even with 1.5V.

Same. I actually did get 3.2 out of mine, but this was on an Asrock SATAII board, so I left it at 3.1 and its running great as a workbench rig. Typing this on it now.