Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

madfaze: late this month first samples AFAIK and mainly to OEMs in Q2 in B3 step.

ragnarok95: nice. So any testing details yet? Max HT, Max NB, Max MHz, Max stable MHz, Max benchable MHz?

BTW another myth debunked which I told you about before very clearly.

I asked AMD why they had asked MB manufacturers to definitely put the TLB Fix on by default and they said no we haven't done any such thing. What the MB manufacturer chooses is their choice entirely.

So there you have it, just as I had told you around 4 weeks back now.

[Kayin]

My K9A2 refuses to boot twice in a row. I've got a 3600X2 on it, 2GB Dominator 6400C4, ATM 1 3850, and it's powered by a PCP&C Silencer 750 Quad. Is it BIOS related (IE, fixable) or would it sound more like a bad board?

[KTE]

Kayin which BIOS are you using?

[Kayin]

Was 1.2, now 1.3. I'm still not booting every time, but at least now I can get it into windows if I grit my teeth and try hard.

[jeremydde]

My K9A2 is doing the same thing. I have a 5000+ BE @ 3150 w/ crucial ballistix and a 3870x2. The thing rarely boots the first time. Usually I will have to hit the reset button then it boots fine. I have prime95'd the system for over 10 hours continuously, plus there are no errors in windows, games, or any other app. The only problem is cold boot. Is it possible that the board doesn't up the voltages soon enough at cold boot? I am also using Bios 1.3

[KTE]

Try BIOS P0H (BETA, but better than 1.3): http://www.box.net/shared/gsn4mgt6o4

Load defaults and then flash with parameters /p /b /n /c

Let me know if you get the same problem. It looks like memory to me.
Also check your PSU voltages on all rails at boot.

If you still get it, can you describe exactly what happens and at what stages?

[Kayin]

What memory would you recommend then? 8 GB worth, if possible-this machine does rendering duty on occasion...

[jeremydde]

I flashed the bios to P0H and my problem has disappeared. I applied the same overclock settings in this bios and it boots everytime. Also, it boots faster, and XP loads noticeably quicker as well. Also shutting down in Xp is quicker as well. Before it seemed like windows locked up while shutting down as it took a long time. Now it shutsdown in about 8 to 10 secs. Clearly there is something wrong with the 130 bios. Also, does anyone happen to know what the Auto Xpress feature is, and what it does? I currently have it disabled.

[Kayin]

My board just flat died (open circuit, no beeps) and I've gotten an RMA number for it. I'm sending it off tomorrow, will post back when the new one gets here.

Mine was not just a warm boot. I had trouble coaxing it to do anything at all, including post. BIOS settings could not be changed, etc.

[dr_drache]

Try BIOS P0H (BETA, but better than 1.3): http://www.box.net/shared/gsn4mgt6o4

Load defaults and then flash with parameters /p /b /n /c

Let me know if you get the same problem. It looks like memory to me.
Also check your PSU voltages on all rails at boot.

If you still get it, can you describe exactly what happens and at what stages?

M9K2 Plat.. 9600BE
I tryed this bios, and it causes a BSOD (nvidia driver) on boot :
I think it's nvd3dumx.dll
OS is x64 vista
i have flashed with 1.33/1.30, and no issues, but i cannot get over 208 HHT...
no biggie really, not trying to do what KTE does.

but I noticed that BIOS 1.13 was reccomended, but i DO have
OCZ PC2-8000 ReaperX HPC 4GB Dual Channel for my ram...
could I place them in slots 1&3 (i know 1&2 are needed for dual channel, i got it like that now) and place the memory as unganged and get the same performance? only ask this because i could add some XMS2 ram then... and not feel weird with differnt in slots (don't ask lawl)

eather way, i'm thinking i should get a bit higher than :
(with AOD on red, and RWeverthing patch applyed)


I don't want to start messing around with bios, till i know i have one worth messing with. bear with me, havn't overclocked sence pencil mod on the athlons!

[KTE]

If you're buying 2GBx2 or 2GBx4 kits then please be very careful and ask before doing so. Many kits are not yet supported especially with Phenom and if you have Vista 64-bit!

I will repeat as I've mentioned many times before; Vista 64bit = hundreds of problems with Phenom!

Most well known problem = BSoD 'Clock interrupt not received on second processor within the allocated time permitted'

Source of problem = BIOS/Vista, nothing else!

Examples (multi-CPU):
http://forums.amd.com/forum/messagev...threadid=92110
http://forums.anandtech.com/messagev...readid=2122913
http://www.driverheaven.net/windows-...not-oc-ed.html
http://answers.yahoo.com/question/in...1225702AAUYnW2

AMD gave out their sample chips with quite some time for MFGs to troubleshoot but most MB MFGs did not have a clue of intricate understanding about Phenom functioning up until around January after user feedback and many users telling them word for word. First ones were in July 0723/0724/0725 B0 step AFAIK. Since then the BIOS/compatibility has got much better. Many many closed BETA testers of BIOSes had absolutely no idea what or how Phenom works and why and they were formulating wrong conclusions and opinions. Main reason being, it's a complex arch. and unlike K8 in it's functioning, which they now know. Just look at how long it took us to work it out and Me/Achim here are still finding important things starting non-stop from 15th Nov (latest in DFI RD790 thread).

For MSI K9A2 Plat., BIOS P0H has many major fixes and updates; apart from the lack of OC options for Phenom and the semi-enabled TLB fix, it is near flawless for me. Very stable as one should expect. I have tested them all, even some you haven't, and P0H is the most stable/hardware supporting out.

I flashed the bios to P0H and my problem has disappeared. I applied the same overclock settings in this bios and it boots everytime. Also, it boots faster, and XP loads noticeably quicker as well. Also shutting down in Xp is quicker as well. Before it seemed like windows locked up while shutting down as it took a long time. Now it shutsdown in about 8 to 10 secs. Clearly there is something wrong with the 130 bios. Also, does anyone happen to know what the Auto Xpress feature is, and what it does? I currently have it disabled.

Good to know. I told them not to use the old 1.32 code but use P0H but they didn't listen.

AutoXpress briefly: http://www.xtremesystems.org/forums/...postcount=1050

Boosts subsystem component performances. It's useful for systems with powerful PCIe 2.0 cards.

My board just flat died (open circuit, no beeps) and I've gotten an RMA number for it. I'm sending it off tomorrow, will post back when the new one gets here.

Mine was not just a warm boot. I had trouble coaxing it to do anything at all, including post. BIOS settings could not be changed, etc.

Bad board then.

M9K2 Plat.. 9600BE
I tryed this bios, and it causes a BSOD (nvidia driver) on boot :
I think it's nvd3dumx.dll
OS is x64 vista

I'll get back to you about the RAM but I've emboldened the problem near enough every user has had with Phenom so far.

The BSoD at that stage makes me think you installed OS on a different board and didn't format/install it new here?

(with AOD on red, and RWeverthing patch applyed)
http://www.imagehosting.com/out.php/i1609727_bench.jpg

That's very low, the patch is still applied to 3 cores there.

What speeds were you at?

[dr_drache]

thanks for check on the ram,

The BSoD at that stage makes me think you installed OS on a different board and didn't format/install it new here?

that right after OOBE, then install drivers, and reboot.
first install gave me no AREO, and acted like old xp/98 as in, REQUIRED
a reboot to detect new hardware.
2nd time, worked fine, but was getting that BSOD, then i thought,
what the heck, and flashed 1.30, and running pretty good right now
(it's a bit slower than my crosshair mobo though)
which also, weird enough had no issues with OC other than the lack of voltage control (bios 0905beta if you care)

That's very low, the patch is still applied to 3 cores there.

how would i make sure all 3 are applyed with RW? (version .29) i'll do it manually again see if i can "Fix it"

What speeds were you at?

that's about as stock as it gets, ~2301 everything in bios on default.


EDIT : changed the 0xc0010015 from 0x0000018 to 0x0000010 in all 4 cores. same results. tryed with RED AOD and all 4 cores, same result. intrestingly enough, when i brought up AOD after the first change, it was set at "yellow". and then when i shut it off (black), and rechecked CrystalCPU, only core 1 was back at 18.

[Kayin]

What RAM is safe to get in 2X2GB or 4X2GB?

[Suosaaski]

Did some winrar benchmark tests with different BIOS versions. Phenom 9600 at stock, no optimizations, Vista 32-bit. I must say I was a bit surprised.

The scores are all "about" scores, I did not record the numbers, just noticed they were about the same all the time regardless of BIOS-version (apart from P0H).

BIOS 1.1 : around 400KB/s (varies between 380 and 430, mostly around 400-410) at stock, around 580KB/s (varies between 570 and 620, mostly at around 58x) with AOD at "red"
BIOS 1.3 : about the same
BIOS 1.0B1 TLB-fix enabled : about the same
BIOS 1.0B1 TLB-fix disabled : about the same
BIOS P0H TLB-fix disabled : around 400KB/s at stock, around 490KB/s with AOD at "red"

First: Are the numbers fine?
Second: Why does P0H have such poor performance compared to the other versions?

I am kinda amazed at how large a difference there is in this particular test between AOD at yellow/red and no AOD running...

[dr_drache]

UPDATE :

well, for S.A.G's sake (S**ts and giggles) I went into bios and reset to ganged mode (aka, 1x128bit, dual channel as we remeber it) (hopefully i am getting the terms right)
came into windows, opened AOD, set RED, open crystalCPU
went through and just checked the state of 0xc0010015, only core 4 was changed, (to 0x01000010), so it seems AOD for me is pretty much useless,
anyway, set the bits, and benchmarked :


much better!!

now am i supposed to be doing something with this MSR? :
0xc0011023, it's in the RW file, but i don't know if it applys to me?

on a side note, my install may be buggered, the progress bar pauses 3 times while booting...

[KTE]

i have flashed with 1.33/1.30, and no issues, but i cannot get over 208 HHT...
no biggie really, not trying to do what KTE does.

Should get much more than that. You'll need to work at it and spend time on it if you wanted it.

but I noticed that BIOS 1.13 was reccomended, but i DO have
OCZ PC2-8000 ReaperX HPC 4GB Dual Channel for my ram...
could I place them in slots 1&3 (i know 1&2 are needed for dual channel, i got it like that now) and place the memory as unganged and get the same performance?

RAM should be wok fine IMO. Nope, you need it in 1&2 or no dual channel. Single channel is poor performance. Like this: http://www.xtremesystems.org/forums/...2&postcount=23

I don't want to start messing around with bios, till i know i have one worth messing with. bear with me, havn't overclocked sence pencil mod on the athlons!

It's easy but just confusing.

that right after OOBE, then install drivers, and reboot.

For Phenom you may have to install OS fresh. Or it'll give too many problems unless you've deleted all the waste files and registries of old components manually. OOBE, out of body experience?

2nd time, worked fine, but was getting that BSOD, then i thought,
what the heck, and flashed 1.30, and running pretty good right now
(it's a bit slower than my crosshair mobo though)
which also, weird enough had no issues with OC other than the lack of voltage control (bios 0905beta if you care)

Hmm.. you should flash 1.13B if you want to oc through BIOS. Slower than your Crosshair, as in what way? Any benchmark comparisons?

how would i make sure all 3 are applyed with RW? (version .29) i'll do it manually again see if i can "Fix it"

Like this: http://www.xtremesystems.org/forums/...0&postcount=39
CPU1, 2, 3 etc parameters write MSR for each core separately. RWE has never worked for me yet though so I don't use it, no time to troubleshoot.

EDIT : changed the 0xc0010015 from 0x0000018 to 0x0000010 in all 4 cores. same results. tryed with RED AOD and all 4 cores, same result. intrestingly enough, when i brought up AOD after the first change, it was set at "yellow". and then when i shut it off (black), and rechecked CrystalCPU, only core 1 was back at 18.

It's because you need to write to all cores and two different MSR (AOD yellow will do fine). Like shown here: http://www.xtremesystems.org/forums/...postcount=1215

Try it.

What RAM is safe to get in 2X2GB or 4X2GB?

2x2GB, many incl. Crucial Ballistix/Tracer, Corsair Dominator (XMS2), Kingston HyperX (newer BIOSes), many OCZ such as ReaperX/Reaper, G.Skill such as F2-8500CL5D-4GBPK/F2-8000CL5D-4GBPQ and GeIL modules like the Black Dragon which can all boot 1.8V if need be. Higher, either 4x2GB or 4x1GB, it gets tricky and depends on the RAM latencies/volts itself. I have not seen or tested any where all four are populated with this board but two which worked with BIOS P0F and P0H: Corsair QUAD2X4096-8500C5DF (Twin2x 4x1GB) and GeIL Black Dragon GB24GB8500C5QC (4x1GB). YMMV as always, but the above are known to be working for others on this board. P0H is best to try them with especially since it can run all RAM dividers, all before P0E do not offer many 4x2GB, 4x1GB or 2x2GB RAM compatibility. The biggest problems are experienced by Vista users and especially Vista X64. Many will have no boot.

Did some winrar benchmark tests with different BIOS versions. Phenom 9600 at stock, no optimizations, Vista 32-bit. I must say I was a bit surprised.

The scores are all "about" scores, I did not record the numbers, just noticed they were about the same all the time regardless of BIOS-version (apart from P0H).

BIOS 1.1 : around 400KB/s (varies between 380 and 430, mostly around 400-410) at stock, around 580KB/s (varies between 570 and 620, mostly at around 58x) with AOD at "red"
BIOS 1.3 : about the same
BIOS 1.0B1 TLB-fix enabled : about the same
BIOS 1.0B1 TLB-fix disabled : about the same
BIOS P0H TLB-fix disabled : around 400KB/s at stock, around 490KB/s with AOD at "red"

First: Are the numbers fine?
Second: Why does P0H have such poor performance compared to the other versions?

I am kinda amazed at how large a difference there is in this particular test between AOD at yellow/red and no AOD running...

There is definitely a performance difference across BIOSes, 1.13B is best I've tested in pure performance and P0H is second. But you need to disable the TLB errata patch first

Try this link method and then switch off all basic extra's, AOD yellow and rerun the WinRAR bench. Should be much higher: http://www.xtremesystems.org/forums/...postcount=1215

[dr_drache]

For Phenom you may have to install OS fresh. Or it'll give too many problems unless you've deleted all the waste files and registries of old components manually. OOBE, out of body experience?

OOBE = Out of Box Experence, the wizard before you can first log-in

Hmm.. you should flash 1.13B if you want to oc through BIOS. Slower than your Crosshair, as in what way? Any benchmark comparisons?

no benchmarks that i can post, but i KNOW that i was getting near 1900 in the winrar test, it was stock running, no matter what i tryed. switched because I saw that a 790FX chipset may be better for me for tweaking (kinda irked about lack of SLI, but i'll live) love love this crosshair sitting on my desk :P

It's because you need to write to all cores and two different MSR (AOD yellow will do fine). Like shown here: http://www.xtremesystems.org/forums/...postcount=1215

Try it.

ahhh, i've been doing only the 1/2 of it, i have not done the 0xc0011023 registers.
i'll give it a shot.

There is definitely a performance difference across BIOSes, 1.13B is best I've tested in pure performance and P0H is second. But you need to disable the TLB errata patch first

Try this link method and then switch off all basic extra's, AOD yellow and rerun the WinRAR bench. Should be much higher: http://www.xtremesystems.org/forums/...postcount=1215

i'll look into it, not really looking into formatting again, will if it needs it. will attempt to build a new iso with vlite, see if i can get a promise raid driver intagrated right this time. as soon as i decide on the format, i'll switch back to trying P0H.

UPDATE :
i think i may got an issue, my 0xc0011023 doesn't match






UPDATE #2 : flashed 113, did BOTH register changes and got :

[jeremydde]

Did some winrar benchmark tests with different BIOS versions. Phenom 9600 at stock, no optimizations, Vista 32-bit. I must say I was a bit surprised.

The scores are all "about" scores, I did not record the numbers, just noticed they were about the same all the time regardless of BIOS-version (apart from P0H).

BIOS 1.1 : around 400KB/s (varies between 380 and 430, mostly around 400-410) at stock, around 580KB/s (varies between 570 and 620, mostly at around 58x) with AOD at "red"
BIOS 1.3 : about the same
BIOS 1.0B1 TLB-fix enabled : about the same
BIOS 1.0B1 TLB-fix disabled : about the same
BIOS P0H TLB-fix disabled : around 400KB/s at stock, around 490KB/s with AOD at "red"

First: Are the numbers fine?
Second: Why does P0H have such poor performance compared to the other versions?

I am kinda amazed at how large a difference there is in this particular test between AOD at yellow/red and no AOD running...

Funny, I did the winrar benchmark after I flashed to P0H and increased from 580 to 642 KB/s.

[jeremydde]

Sorry, didn't have the multithreaded box checked. I got 1306 kb/s , dunno what was before the flash.

[Mathos]

Hmmm, so if I'm understanding this right, does P0H have the ability to disable the TLB fix? Or does it only disable it for one core?

[Suosaaski]

There is definitely a performance difference across BIOSes, 1.13B is best I've tested in pure performance and P0H is second. But you need to disable the TLB errata patch first

Try this link method and then switch off all basic extra's, AOD yellow and rerun the WinRAR bench. Should be much higher: http://www.xtremesystems.org/forums/...postcount=1215

I might test that when I get back home.

By the way, do you happen to have any idea why AOD shows my memory is in single channel mode?

I have 4 * 1GB PC2-6400 and BIOS is set to unganged.

And I was to understand BIOS 1.1 would not have the TLB-fix, howcome it gets the same performance as the newer ones?

[KTE]

no benchmarks that i can post, but i KNOW that i was getting near 1900 in the winrar test, it was stock running, no matter what i tryed. switched because I saw that a 790FX chipset may be better for me for tweaking (kinda irked about lack of SLI, but i'll live) love love this crosshair sitting on my desk :P

It should be a little better for RD790. Check here, this is what you should be getting at stock (end): http://www.xtremesystems.org/forums/...postcount=1215

On X64, it should be more because X64 has access to higher memory bandwidth. Recheck that CPUZ 1.44.1 is reading dual channel. Post your Memset latencies next time you run it too. What WinRAR shows is essentially your memory bandwidth.

i'll look into it, not really looking into formatting again, will if it needs it. will attempt to build a new iso with vlite, see if i can get a promise raid driver intagrated right this time. as soon as i decide on the format, i'll switch back to trying P0H.

UPDATE :
i think i may got an issue, my 0xc0011023 doesn't match

Copy all the register EDX and EAX but for the 22. Change that to 20 and writeMSR for each core. Check multi-threaded and make sure something else is not heavily taking up your memory.

UPDATE #2 : flashed 113, did BOTH register changes and got :
http://www.imagehosting.com/out.php/i1611527_bnc.jpg

1.13B doesn't have any patch. Post me a screenshot of AMD Power Monitor, CPU-Z CPU, Mainboard, Memory, SPD tabs and Memset altogether please when you run this bench (make sure either the pic is resized to less than 1024x768 or thumbnailed). Your RAM perf. there is like Athlon 64 X2 5000+ BE rather than how Phenom should be.

Hmmm, so if I'm understanding this right, does P0H have the ability to disable the TLB fix? Or does it only disable it for one core?

When you choose disabled, it's only disabled on the 1st core and not on the other 3 = bad perf.

I might test that when I get back home.

By the way, do you happen to have any idea why AOD shows my memory is in single channel mode?

Always does, an AOD bug. Check in latest CPU-Z, if it shows dual channel, it'll be dual channel.

And I was to understand BIOS 1.1 would not have the TLB-fix, howcome it gets the same performance as the newer ones?

Looks like you ran it single threaded and not started by ticking the multi-threaded box (?).

[Suosaaski]

Looks like you ran it single threaded and not started by ticking the multi-threaded box (?).

The multithread was selected by default.

PS. Ever since I first installed BIOS 1.2, I think the performance went down and never came back up even with 1.1. At least in 3dmark. Almost like the TLB-fix would stay there no matter what. Or something else is wrong.

[KTE]

^Post your config screenshots and details please.
Also post your temperatures and powersupply.

Make sure you've flashed using the parameters I stated because these are BETA ( /p /b /n /c). The only way you could get that perf. using Phenom is if something was wrong.

-To check TLB errata patch, just download CrystalCPUID, Function>open MSR Editor>type MSR number: 0xc0010015 and press RDMSR.
-Read the EAX register, it should end in 10. If not post back what it shows.
-Then change core to 2/4, 3/4, 4/4 using the drop down in CCPUID upper right corner and for each one press RDMSR again and read the value (note it).

-Next, type 0xc0011023 in the MSR number box and hit RDMSR again for each core.
-Check EAX now ends with 20 and if not, post back what it does.

[Suosaaski]

^Post your config screenshots and details please.
Also post your temperatures and powersupply.

Make sure you've flashed using the parameters I stated because these are BETA ( /p /b /n /c). The only way you could get that perf. using Phenom is if something was wrong.

-To check TLB errata patch, just download CrystalCPUID, Function>open MSR Editor>type MSR number: 0xc0010015 and press RDMSR.
-Read the EAX register, it should end in 10. If not post back what it shows.
-Then change core to 2/4, 3/4, 4/4 using the drop down in CCPUID upper right corner and for each one press RDMSR again and read the value (note it).

-Next, type 0xc0011023 in the MSR number box and hit RDMSR again for each core.
-Check EAX now ends with 20 and if not, post back what it does.

I'll try to flash BIOS 1.1 when I get back home and check those values then. Will also check them with current values for 1.0B1 (which has the option to disable TLB-fix) and P0H probably too.

My config is as follows:
MSI K9A2 Platinum
OCZ Platinum DDR2-800MHz, 4 * 1GB
AMD Phenom 9600 Black Edition, stock cooler: idle @30-32C, load with prime, under 50C (fan ramps up from 50% at 45 degrees, set from BIOS)
MSI HD3870-OC
Western Digital 250GB SATA
Antec NeoHE 550W PSU