Phenom 9500 w/ MSI K9A2 Platinum

[jesteronyer6]

What is the best BIOS for the K9A2 and the Phenom?

[justapost]

OK. Is 1.25V (core) what you set in BIOS or what you get real through DMM measurements or what your Super I/O chip reads in Windows idle (i.e. AOD/EVEREST/HWMonitor) ?

KTE may I ask you how I can locate that superio chip. I have an osciloscope here for a week. EDIT: nevermind found it.

BTW: You remeber the intel rig I have to build for an employee. His neighbor works at intel and he came up with an qx6850 for free. I wonder what it's tdp at 2,2GHz will be.

[Swanie]

My last 9600 BE: http://www.xtremesystems.org/forums/...&postcount=569

OPN: HD960ZWCGDBOX
CAAWB 0747MPMW

My new one:



Hahaha

OPN: HD960ZWCGDBOX
CAAWB 0747MPMW

Man, they only had this batch/week/OPN there when I picked it up, every one of them spare was this. It's the exact same coded one I had before, I have yet to boot it up and check what the stock voltages/te,mps/power/vid are and if it reacts the same as my other one. Would be a very good checkup for "this stepping/week/OPN is [insert comment]" don't you think?

Also, both of these had a 2mm^2 dull metallic rough spot looking like a "pencil smudge" in the middle of the IHS. Did anyone else get something like this with their Phenom?

Hmm.. I wonder why though, the SB600 has given more problems than the Promise.

Don't plan to really.. you don't want to know, do you?

I need Admin to restrict me from posting unless I can verify a 3GHz Phenom. Maybe then I can actually start trying with this.... although that time may never come on earth unless I swap my C2Q in there somehow


Thanks Swanie. You can move the HT slider up though. Tried it yet?

(don't try above 2.6 HT, it'll lockup)

Have you tried lower RAM divider than 1.2.66 ? Do they work?
What about any other options like NB multi, NB VID, CPU VID, are they available in the BIOS like the older ones, or totally removed?

Same as the P0F and P0G then.

No options for most of P-State variables as we had before in the new BIOSes. Very few oc options now.

None of the 'Performance' BIOSes have been any decent for me since P0C. I wonder why this is.

Are you sure about this? Try P95 latest one which is 25.6 IIRC. I highly doubt anything will come near its power pull, load and hence what creates the high temps apart from Linpack. Orthos is IME 12W lower full load than P95 small FFT on the Phenoms with lower temps and easier to pass too, although it is supposed to be based on the same code, there are many bug and performance improvements with P95.

No idea about 1.2. 1.31 BETA was excellent for X2 when I tried it, better than 1.0 and 1.1.

Then we now know why it's BETA.

Try a stable release if you can instead but the temps/volts info should be provided with HWMonitor/CoreTemps perfectly anyway.

Well, it could be program incompatibility or just an incompatibility bug. Maybe they are both accessing and sharing the same resources, this can cause such freak behavior.


Hey Suosaaski, wlecome.

First the bad news...
Damn that is the exact chip I've had twice now and yes, those VID/Volatges will be pulling mega power. Almost the Quad FX again here.

Can you do something for me on Vista... at 2.4GHz run Super Pi Mod 1.5 XS 1M and post it back please.

(run it about 3 times before capturing ss of the last one)

All VId/DID options are gone in the new beta. I have tried that divider but my system don't boot. Running Mushkin as you know. The only time it boots with that divider is at default speed @ 1066mhz. Will try to play with the slider. got my machine to boot at 2760 and was running Prime the past 24 hours and all was ok but then tried 3mark06 and pc froze, then no boot

[jonspd]

Yea one thing I noticed the other day while testing is even if it is prime, 3dmark stable doesn't mean you want have a cold boot issue as I think thats what it would be.



Justapost if you want a good OCing chip like KTE you will have to try as many as he has which I think is around 4-6 now

[Suosaaski]

Hey Suosaaski, wlecome.

First the bad news...
Damn that is the exact chip I've had twice now and yes, those VID/Volatges will be pulling mega power. Almost the Quad FX again here.

Can you do something for me on Vista... at 2.4GHz run Super Pi Mod 1.5 XS 1M and post it back please.

(run it about 3 times before capturing ss of the last one)

I could... if the main page wouldn't time out when trying to open it. So will try it later.

Note that my system or os is not optimized in anyway, and I have F-Secure Antivirus which is not the lightest by any means, so when I do test it, it might not perform at the level expected

[Suosaaski]

nevermind

[Brother Esau]

How are the Bios turning out from M.S.I guys? Also on the more recent beta releases has anyone using Dual Core noticed that the new Bios is not quite close to as good for Dual core as the first releases were?

Don't know if this is a DFI thing or if its common for all boards with recent Bios coding?

[jonspd]

NOT to swell IMO according to what I can tell there isn't anything better then the 1.13 betas. Which is the 1st or one of the 1st betas...

I'm gonna give 1.3 a shot if they will ever release it

[Daveburt714]

I agree with you Jonspd... I flashed v1.2 Release version tonight, and although it does seem a little more stable, the performance hit seems pretty bad... All I did was a few Sandra test but it was a pretty big hit, even with AOD on Yellow&Red, Mem Bandwidth dropped below 9000 (I've hit 11k stable!) ....

I've only played with it for a couple hours, but I still can't boot at 260x10, looks like ~255 HT might be do-able but theres no way it's gonna make up for the TLB patch hit...

Oh well, I'll give it a few days...

[KTE]

Hehe, yeah, so.

I was gaming and then started running P95 yesterday... nothing errored. I've never found instability not caught by bench>game>memtest>12hr P95 in anything so far. Folding loads are no good for testing 100% stability IME because it never picks up the errors there and then to give you a visual output and stop the load with critical/small core errors (i.e. prime number calc catches the tiniest of computational error in a rounding up), instead they just carry on like many benchmarks. I tested this with the Phenom 9500 intentionally unstable at 2.560G earlier, no errors appeared, although it is a constant static load, good for testing ideal load temperatures, power and 75% stability when at 100% each core, first priority. I want to try Linpack preferably if I could get it too run on this machine (?). Anyway, I then had to leave P95 at full load, with front priority, finishing off HL2: LC.

It ran 28C ambient, stock boxed fan and the load temps were ~54C 2.7G @ 1.25VID, 1.305V BIOS/1.336V OS Idle/1.328V OS load @ 230W AC.
BIOS is bad, after 1.21V I only get a 1.305V option which is selectable. But I've noticed with Phenoms, VCore offering is chip dependent. At times 1.30V gives me 1.33V or 1.32V and at other times it'll give me 1.344V.

The minute I was set to go away, a USB port shorted (this is a daily problem with this board, anyone else experience it?) and I had monitoring tools set up so found the +5V and +12V rails sagged to +11.75V and +4.76V from +11.98V and +4.98V for a slight second. At that time, you guessed it, CPU VCore dropped to 1.264V and Core 0 failed prime instantly in that second at 5hr25min.
Since then, I let the rest of the cores which never failed all run on without any more voltage drops. All the rest of the 3 cores which withstood the 1.264V sag were fully stable at the +11hr mark.



YES I tried to give my 2x 102CFM delta's a 1500RPM whirl there at 21C ambient

BUT, then I was busy and away. The explanation of that failure is pretty clear : large +12V rail fluctuation. Will it fail without this bad fluctuation?
Now was night time, aka my time, and I've since then stopped and restarted P95 straight away. If there is no lengthy spike now (meaning not transient fluctuations which are common with electrical systems, but one holding for more than 1s, which is lengthy -- I've left that USB hub free), it shouldn't fail even up to 1.28V load across the cores, like I said. I'm waited for it to.

Even if it did fail because the CPU was unstable, let us say for instance, I have CPU VID, CPU volts, NB Multi, HT multi, NB VID, NB/HT volts still to touch apart from the first and there is ample room ahead.

+8hr Testing Results

So I tested:
1) Power cycling, hot and cold.
2) Cool 'n' Quiet.
3) Start-stop full load fluctuations.
4) Gaming.
5) Idling.
a) 25C ambient with stock box fan at full speed.
b) Adding a 120mm 1000RPM fan (v.low noise) at 25C ambient to the system (exhaust).
c) Then test by moving it up to 3000RPM 152CFM at 25C ambient.
d) Then keep fans at max and drop the ambient to 18C.

One thing you'll notice: Power draw falls with lower core temps. VCore will increase with lower core temps.

1) Passed without any issues each time.
2) Don't enable it even if BIOS oc'ing, no point and messes up the stability in my case.
3) No problems experienced.
4) No problems at all.
5) Nothing odd experienced.

a)

225 x 12 = 2.7G @ 1.25VID 1.305V BIOS/1.336V idle/1.320V load
2.025G NB @ 1.25VID
2.025G HT @ 1.2V
DDR2-1200 5-5-5-15-28-75 2T @ 2.2V
Max temps: 51C, 51C, 51C, 51C
Power: 111W AC idle, 239W AC load

b)

225 x 12 = 2.7G @ 1.25VID 1.305V BIOS/1.336V idle/1.328V load
2.025G NB @ 1.25VID
2.025G HT @ 1.2V
DDR2-1200 5-5-5-15-28-75 2T @ 2.2V
Max temps: 45C, 45C, 45C, 45C
Power: 111W AC idle, 236W AC load

c)

225 x 12 = 2.7G @ 1.25VID 1.305V BIOS/1.336V idle/1.328V load
2.025G NB @ 1.25VID
2.025G HT @ 1.2V
DDR2-1200 5-5-5-15-28-75 2T @ 2.2V
Max temps: 38C, 38C, 38C, 38C
Power: 111W AC idle, 230W AC load

d)

225 x 12 = 2.7G @ 1.25VID 1.305V BIOS/1.336V idle/1.328V load
2.025G NB @ 1.25VID
2.025G HT @ 1.2V
DDR2-1200 5-5-5-15-28-75 2T @ 2.2V
Max temps: 31C, 31C, 31C, 31C
Power: 111W AC idle, 225W AC load

Stopped @ 8hr 38mins stable:



Didn't I say it was easily stable?
One thing I will mention though -> whatever you get, don't run above 2G NB with stock volts, it won't last long (as my 9500 witnessed) and it'll mess your core oc'ing up. I won't be running more than 2G NB until I flash the older BIOS which'll let me put 1.265VID on it to run that speed.

Another note: Jack mentioned something to me earlier and he was right. The stock AMD 9600 BE boxed fan cannot handle overclocking and the heatload produced by it when you raise volts slightly. It will overheat itself and the cores very quickly, touch around it at 50C and the heat is very easily felt, pipes will be all hot, because the measly fan cannot remove the heat quick enough... core temps will thus FLY high.

New TDP to be dissipated: ~130W

BUT, here's how too see the bottleneck:
Add a 500RPM-3000RPM 120mm fan near by and watch the core temps drop 8-20C across the cores.
3rd party coolers will show much better load handling and temperatures, but at least we know that the temperature probes are working. I managed to get a 74C load if I set the stock fanspeed to low at 1.4V load (which makes the system shut down for safety).

AND if you're going plus 64C, stability testing may error at any time just because of the heat. Cores don't like it, simply put, like Wolfdales hating plus 55C.

And in case you didn't already know, AMD Power Monitor only detects bootup Multiplier and VID changes for its shown frequencies.

SScchittenn my USSB porrt (keybooardd) hass fflocked up agaain. Ccoreess hhaveen't faailed though but VCCoree ddid ddrrop to 1.296V. Noww this iss hoow my keeyboard is typinn eveery leetterr
Taakiing mee 2200minss too type oneee linee heree

What is the best BIOS for the K9A2 and the Phenom?

Depends what you're looking for with it. For OC I would say 1.13 BETA (113) is best so far without any patch and excellent system performance (compared to the rest).

KTE may I ask you how I can locate that superio chip. I have an osciloscope here for a week.

You can find it listed in the manual; I've checked and informal in your M3A thread has given you the details by now. You'll have to search for the Super I/O IC electrical datasheet and how the chip works to find out the relevant pin mappings and their corresponding readings (many add calibrated values mentioned).

BTW: You remeber the intel rig I have to build for an employee. His neighbor works at intel and he came up with an qx6850 for free. I wonder what it's tdp at 2,2GHz will be.

Yep.. hmm. 2.2G TDP depends entirely on the volts and VID required to get it 100% stable really. Should be fairly low but that depends on the starting VCore and obviously, the chips capability itself.

All VId/DID options are gone in the new beta. I have tried that divider but my system don't boot. Running Mushkin as you know. The only time it boots with that divider is at default speed @ 1066mhz. Will try to play with the slider. got my machine to boot at 2760 and was running Prime the past 24 hours and all was ok but then tried 3mark06 and pc froze, then no boot

Are you sure?
Let me tell you why I'm asking: 2k6 is absolutely no where near as extreme loading as P95 is. P95 load you won't find anywhere else but with the EVEREST FPU Julia bench and Linpack on the desktop that I've tested so far. There are science apps which load the system a few watts more but not many. This is why you have uber fast 2k6 records, because 4G stable and 4.5G suicide validated can still usually bench 4.35G 2k6 and 4.25G 2k5.

What that mean is, something oddly wrong is occurring with your stability/chip at those frequencies/voltages. Make sure you're not topping 1.48V, Phenom doesn't seem to like it at all.

Yea one thing I noticed the other day while testing is even if it is prime, 3dmark stable doesn't mean you want have a cold boot issue as I think thats what it would be.

That occurs only with my systems if it isn't 100% stable but stable only at lucky times. Like I mentioned, my 9500 was like this at 2.8G boot, once +8hr stable but many times it won't even boot at those speeds and runs badly in even little taxing games.

Justapost if you want a good OCing chip like KTE you will have to try as many as he has which I think is around 4-6 now

Or, you can buy a 9500, much better chance of getting a decent chip. This one is still not better than my 9500, not even close. That did 2.8G benched at stock volts.

I could... if the main page wouldn't time out when trying to open it. So will try it later.

Yep, for me too. Here ya go: super_pi_mod.zip

Try to stop/pause other applications in the background or set their affinity to other cores for best results.

How are the Bios turning out from M.S.I guys? Also on the more recent beta releases has anyone using Dual Core noticed that the new Bios is not quite close to as good for Dual core as the first releases were?

Don't know if this is a DFI thing or if its common for all boards with recent Bios coding?

Not too sure with K8, but this board has had about 15 BIOSes now that I'm aware of. Lately things have moved on and I hope, if what I hear is true, that they'll bring back the AM2+ options but just in a friendlier format.

[KTE]

BTW, I run AMD stock cooler SMART fan controlled (BIOS lowest speed = 0&#37 set to keep sub 45C, so basically, it's at 25-35C at 0-1000RPM nearly all the time with general loads and near whisper quiet whilst keeping good temps and raising speeds on demand. At 2.7G I'm talking about. You guys with this feature should try it. I have because I test pretty much all the boards features I can.

[Swanie]

Yes i am sure it was 3dmark06 but found the issue. It was 3dmark itself as when i tried to run it again it gave me errors- Vista/3dmark issues perhaps. Well only have this weekend to play with my 9500 as it is sold....waiting for my BE.

[Swanie]

New bios released: 1.33
http://www.msi-forum.de/thread.php?t...user=0&page=13

[justapost]

Justapost if you want a good OCing chip like KTE you will have to try as many as he has which I think is around 4-6 now

LOL, atm i wait for a bios with a vcore setting. You can be sure I think about ways to introduce phenoms in our office to get a few more for testing but it's oversized atm.

One thing you'll notice: Power draw falls with lower core temps. VCore will increase with lower core temps.

You mean the vcore cpuz reports i guess. With lower temps phenom seems to be stable with lower cpu/nb vid and vcore settings here.

congratulations to your new results 2-7-2,8GHz stable was what i expected from the be.

AND if you're going plus 64C, stability testing may error at any time just because of the heat. Cores don't like it, simply put, like Wolfdales hating plus 55C.

During watch Speedfan charts I found that the CPU temps has seldome peaks about 5° above the average (swinging in +-1° range). Peaks should be higher with stock cooling as I used watercooling. Thouse can be the source of hangups

SScchittenn my USSB porrt (keybooardd) hass fflocked up agaain. Ccoreess hhaveen't faailed though but VCCoree ddid ddrrop to 1.296V. Noww this iss hoow my keeyboard is typinn eveery leetterr
Taakiing mee 2200minss too type oneee linee heree

You have a few typos here.

You can find it listed in the manual; I've checked and informal in your M3A thread has given you the details by now. You'll have to search for the Super I/O IC electrical datasheet and how the chip works to find out the relevant pin mappings and their corresponding readings (many add calibrated values mentioned).

Yeah thanks I thought the superio chips is different from the chips I called clockchip and informal already pointed me in the right direction.

Yep.. hmm. 2.2G TDP depends entirely on the volts and VID required to get it 100% stable really. Should be fairly low but that depends on the starting VCore and obviously, the chips capability itself.

If the co-worker could decide wheter he wants an X38 or an P35 chipset and needs additional DDR3 sockets for better lifetime i'd already test that cpu here.
Little OT but do you know if the X38 is more power efficient?

[jonspd]

Thats not typos that's what his usb port is causing him to type when regularly using it.....

[KTE]

Yes i am sure it was 3dmark06 but found the issue. It was 3dmark itself as when i tried to run it again it gave me errors- Vista/3dmark issues perhaps. Well only have this weekend to play with my 9500 as it is sold....waiting for my BE.

OK. Let us know how the BE goes

New bios released: 1.33
http://www.msi-forum.de/thread.php?t...user=0&page=13

Ah, that makes it 16 BIOSes now but these last 10 or so really do have bad perf. I'll see what I can do about this.
I'll hopefully test 1.33/P0H tonight before moving back to 1.13.

LOL, atm i wait for a bios with a vcore setting. You can be sure I think about ways to introduce phenoms in our office to get a few more for testing but it's oversized atm.

Did you know, VCore is linked to VID? If you're at a high VID, you can't have a low VCore. Lowest VCore at 1.55VID is 1.48V BTW.
How many BIOS releases has ASUS provided you yet?

You mean the vcore cpuz reports i guess. With lower temps phenom seems to be stable with lower cpu/nb vid and vcore settings here.

AOD/EVEREST/HWM/DMM VCore.

CPUZ hardly picks up my VCore. Yep, lower temps help with everything in an IC, it is scientifically true.

congratulations to your new results 2-7-2,8GHz stable was what i expected from the be.

Thanks but it's the chip rather than me.
I could've got one that only did 2.3G but again, that would've been the chip itself. I just hope users paying would get at least this much out of Phenom BE, I reckon it would be far better than what we see at present. I mean, I have even ran my RAM >1300 with Phenom, so it is worth it especially when you have NB/HT at 2.5G too.

During watch Speedfan charts I found that the CPU temps has seldome peaks about 5° above the average (swinging in +-1° range). Peaks should be higher with stock cooling as I used watercooling. Thouse can be the source of hangups

All that shows is, the heat produced by the core is low, only some decent air cooling is needed to be able to oc it. WC usually gets higher oc than air simply by its higher thermal capacitive nature.

You have a few typos here.

Yep

Yeah thanks I thought the superio chips is different from the chips I called clockchip and informal already pointed me in the right direction.

Do you mean the MB PLL?

If the co-worker could decide wheter he wants an X38 or an P35 chipset and needs additional DDR3 sockets for better lifetime i'd already test that cpu here.
Little OT but do you know if the X38 is more power efficient?

X38 is a power hog compared to all the latest Intel/AMD chipsets right now. 36.5W TDP whilst P35 is a 16W TDP chip. P35 is a very good offering, one of the biggest boosters of Intel sales in Q3/Q4 '07.

[jonspd]

KTE did you get my p.m?

I have pretty much found my chips max. Gonna try to get it stable around 2.65 2.68 but for now I'm pretty sure 2.62 is my max.

[Brother Esau]

16 beta releases huh? Something tells me that from what I see D.F.I is sweeping this board under the rug All of the Bios that have been written for this board after 11/22/07 suck and are half assed attempts more than likely just for show

[jonspd]

thats not good for DFI as I thought there where more enthusiast related over any other mobo producer....

[Brother Esau]

They are but I call a spade a spade and tell it as I see it!

Thats another reason I would not even consider the Phenom right now, not even just to help Debug the damn thing is because of the lack of effort and quality displayed by DFI right now from the Bios team appointed to Bios coding for AMD atm .

[justapost]

CPUZ hardly picks up my VCore. Yep, lower temps help with everything in an IC, it is scientifically true.

Found this on the asus homepage.

Question

Both Asus and third-party utilities such as PcProbII/AISuite/CPUZ shows that my CPU Vcore reading drops as the CPU load goes up.
This also happens after I set CPU Vcore to a set value instead of the preset [AUTO]. Why does the motherboard act like this?
Will this effect the overall stability of my system?

Answer

As Intel specification clearly states Vcc (CPU Vcore) should drop in propotonal to the increase of Icc (CPU current consumption), in order to prevent causing permanent damage to your CPU. When CPU is under stress (in other words, under higher load), the current consumption of the CPU will go up, and the CPU Vcore will then drop to conpensate this change. This is a perfectly normal behaviour, and will not effect the overall stability of your system.

For more information, please refer to Intel CPU datashee, under "DC Voltage and Current Specification".

If phenom behaves similar that explains why the vcore drop is lower (vcore higher) with lower temps as the current consumption of the cpu will be lower.

Thanks but it's the chip rather than me.
I could've got one that only did 2.3G but again, that would've been the chip itself. I just hope users paying would get at least this much out of Phenom BE, I reckon it would be far better than what we see at present. I mean, I have even ran my RAM >1300 with Phenom, so it is worth it especially when you have NB/HT at 2.5G too.

Good chips need skilled users. AMD should preselect the good ones and sell em for a higher price.

All that shows is, the heat produced by the core is low, only some decent air cooling is needed to be able to oc it. WC usually gets higher oc than air simply by its higher thermal capacitive nature.

That is true. I have exactly one +5-8° peak here in 15 minutes prime95. Mounted aircooling now will see if this peak is still there.

Do you mean the MB PLL?

Yep. MB PLL must be controlled by the super-io chip. That's what i deduce from the fact that one has to select that chip in for example systool or clockgen to modify the ref HT/fsb.

X38 is a power hog compared to all the latest Intel/AMD chipsets right now. 36.5W TDP whilst P35 is a 16W TDP chip. P35 is a very good offering, one of the biggest boosters of Intel sales in Q3/Q4 '07.

Thanks that's what I thought also. Was bushed by the EPU feature on some asus boards resulting in the same consumption than an p35 mobo.

[jesteronyer6]

Anyone have a link to 1.13 or to all of the BETA BIOS's archived somewhere?

[KTE]

It's true. CPU manufacturers implement Vcc drop across the cores through the VRM circuitry due to high load Icc, and the colder it is, the higher the Icc values due to less leakage, quicker on off transitions, quicker switching transistor speeds with less resistance -> higher voltage, less heat, less power. But Vdroop/Vdrop doesn't necessarily exist in the amounts you want in all boards, which is why ASUS provides the option of "load line calibration" and "voltage damper" to avoid it, because when overclocking, too many times Vdroop is much more than it should be for perfect stability at those particular settings, surpassing the low limit core voltage for stability, i.e. 1.3V idle -> 1.25V load, when you need 1.265V minimum load for perfect stability. So to limit this droop to less is good for oc and finding oc stability. Things get complicated when you complain about Vdrop and Vdroop for no reason, i.e. why does Vcore idle so high (which is good actually, provides a large offset from the low limit which'll cause instability).

We've had 16 BIOSes that I know of, including 3 official final releases, and another one coming up soon.

Here's 1.13 BETA jesteronyer6: 1.13 BETA

[Brother Esau]

We've had 16 BIOSes that I know of, including 3 official final releases, and another one coming up soon.



You suck

[jesteronyer6]

Thanks KTE !