Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

Quick Update:

Nope. Nothing works. OverDrive doesn't work for me no matter what I do. Will instantly freeze the system. I can boot at 220HT from BIOS stock, I'm running it right now but changing 1MHz on HT through OD from 200 to 201 will freeze it.

OverDrive detects wrong CPU voltage and wrong CPU speeds when you change NB speed through WPCEDIT. Like Neoseeker report, at least 300MHz higher a 8x NB multi.

"Phenom 3GHz" guy said he changed HT multi to 3x -> I can't change HT multi anywhere. 5x is stock.

I've tried max core volts (1.525V), max NB volts and max HT volts. It makes zero difference but to power/temps.

I've tried WPCREDIT (tictac's methods- thank you fella). Changed NB speed, only multi 8, 9, 10 work for me but that doesn't allow a single MHz more in anything.

Stock NB speed:


New NB speed:



Still have RAM problems.

Also some advice: The only thing I've done since my last reinstall on two new HDs is changed HT between 210-220, failed POST a few times, system reset values itself (no power off) and already my system is again finding plus 50 HD errors (checkdisk runs again and again). For me it's not too bad since I'm only testing it, I paid nothing for it and it's not going to be my day to day runner. For anyone with crucial info or HDs attached, this is a major problem.

So should I wait for a second revision of Phenom to build my AMD rig or are these going to OC fine once the mobo BIOS situation is squared away?

I don't want to get a chip that can't even make 2.6.

What you already have is better, I would stick to that or a Q9450 if I were you. This platform is far from ready yet, it's not advisable to buy it to run day to day at all. To test anything is fine but to run daily, you can totally forget about Phenom for now. Far far too many headaches.

2.4GHz stock volts with very low speed/high latency RAM is about max consistent stable for daily use. Until they sort out these major issues.

[Blacklash]

What you already have is better, I would stick to that or a Q9450 if I were you. This platform is far from ready yet, it's not advisable to buy it to run day to day at all. To test anything is fine but to run daily, you can totally forget about Phenom for now. Far far too many headaches.

2.4GHz stock volts with very low speed/high latency RAM is about max consistent stable for daily use. Until they sort out these major issues.

Thank you for the prompt reply. I was looking to upgrade my secondary rig to an AMD chip not my main. I would have to be high to replace my main rig in my sig with a 2.2-2.6GHz Phenom.

Right now my secondary rig has an E6320 in it @ 3.4GHz. From the look of things currently I'd be better off getting a Penryn for my prime, and moving this Q6600 G-0 to my secondary.

I really want to see AMD succeed though and I have very fond overclocking memories with AMD chips. Competition is win|win for us consumers and I am just looking for AMD to give me an excuse to buy from them again. We will see.

[LIKMARK]

Awsome work KTE! You just smacked up all those tests we've been asking for! Now, when (if) you get bored playing with your 790FX board, please try out your Phenom on a AM2 board. (if you got one at your hands)
Thanks!

[JumpingJack]

Cheers Jack.

It happened when I set VCore to 1.485V, NB to 1.350V, HT to 1.25V and RAM 1:2 at 4-5-5-5-12 2.3V at 300HTx11...

It boots up but no picture (POST). But then when I restarted at default values it started up but all my files including Windows files, directories and saved pictures did not open up. Windows read "the system cannot find the specified file". I lost some pics and CPU-Z valids like that, but luckily some survived.

Ahhhh, yeah to it looks like to froze at the start of loading the system files and the partition table wacked out. The boot strap loader started but the system files were corrupted, seen that error many times.

This is good news, it was more of a 'improper' windows start up than some wierd BIOS write back to the drive that screwed things up....

[The-PHENOMATOR]

Nice work KTE

lots of effort from you and lots of appreciation from us.

take a look here, you may find it useful

http://www.xtremesystems.org/forums/...20#post2592120

[Lightman]

Nice work KTE

lots of effort from you and lots of appreciation from us.

take a look here, you may find it useful

http://www.xtremesystems.org/forums/...20#post2592120

Great work KTE!!

Some of your Everest tests seems a bit low, I just tested my 9500 for 5 minutes this morning and it was getting better memory and Zlib numbers .

I had ZERO problems with my RIG, just plug'n'play going form X2 6000+ to Phenom. It even booted first time already OCed because I left HTT BIOS setting from X2 at 215MHz . I will do more research this evening and then we can compare results and opinions !

[KiD0M4N]

Great work KTE!!

Some of your Everest tests seems a bit low, I just tested my 9500 for 5 minutes this morning and it was getting better memory and Zlib numbers .

I had ZERO problems with my RIG, just plug'n'play going form X2 6000+ to Phenom. It even booted first time already OCed because I left HTT BIOS setting from X2 at 215MHz . I will do more research this evening and then we can compare results and opinions !

His results are all single channel. Maybe thats making the difference?

[Andi64]

KTE, are you runing default 1.0 BIOS?

There are 3 BETA BIOS to test, v112, P0D (performance BIOS, doesn't work with my K8) and v113. All of them have HTT multiplier from 1x up to 13x and a few new options.

I've flashed P0D and v113 with AFUWIN. They both work, but P0D slows my system down.

The version 113 works perfectly, I can change my HTT multiplier now, but there is no way to make this thing work at 300Mhz BUS. I have an Athlon64 X2 5200 rev. F3, 3250Mhz is easy with only 1.44V, but I can't make it stable at 3300Mhz.

You can find the BIOS at MSI.de forums, or PM me and I will send it to you.

Edit: also AMD Overdrive is working now

Edit2: I'm at GMT -3:00. It's allways late when I post and nobody is around! lol

[KTE]

Even I have an X2 that works fine on the board. It's just the Phenom.

A word of advice: if you have OverDrive working then you'll get higher MHz. I should get 3GHz. My chip needed no volts or changes to anything but HT speed for 2.6GHz with no WPCREDIT used. NB is what throws everything away though.
To check if you have OD working, all you need to do is go to the 'Preferences' tab and click "Apply". If it works on your system, it will apply nothing and be OK. If not, then like me your system will freeze.

Also I "think" that NB frequency should be 200MHz higher at stock. Not sure.

Soldner: Like I said, OverDrive doesn't work for me though this chip can bootup at 235HT x11 compared to 215HT x11 you were doing stock all volts. Also you have 1.560VCore and my BIOS only gives max 1.525V option. Your NB volts are very high though, it will be heating up pretty bad.

216MHz HT boots perfect for me now consistent but in between that 217, 218, and 219 do not boot. However, higher boots up. I tried to replicate stock Q6600 2.4GHz, closest I got to it was 220x11 2420 which is what I'm running now. Passed all stability tests stock volts.

Going to try BETA BIOSs later.

Memory bandwidth results are very NB frequency linked. The higher, the better. 550 5-5-5-5 2T is not much different to 450 4-4-4-4 2T unless the NB frequency is higher.

No windows tweaks apart from LSC, nor any hardware tweaks apart from what is mentioned.

Ran some more tests early today. I can re-run the above 2486MHz runs at 2420MHz for anyone if they want too but bear in mind there won't be much difference. So here's some other ones today at 2420MHz to compare to Q6600. Enjoy.

Setup











That's all all for now=>

Some of the benches are real bad. I mean POV-Ray built in bench took around 30 minutes and only ever used 114W AC. That's just above idle and I really don't think if coding was making use of cores efficiently it would be 114W. That shows there's a bottleneck somewhere which the coding is limited to.

Thank you for the prompt reply. I was looking to upgrade my secondary rig to an AMD chip not my main. I would have to be high to replace my main rig in my sig with a 2.2-2.6GHz Phenom.

Right now my secondary rig has an E6320 in it @ 3.4GHz. From the look of things currently I'd be better off getting a Penryn for my prime, and moving this Q6600 G-0 to my secondary.

I really want to see AMD succeed though and I have very fond overclocking memories with AMD chips. Competition is win|win for us consumers and I am just looking for AMD to give me an excuse to buy from them again. We will see.

The unlocked chips should be good. IIMO the version to look out for is when the NB is step 3. It's the NB frequency with HT link which messes everything up. You should easily get 235HT even on the x13/x14 chips, so plus 3GHz. But when you leave the NB default, you'll have problems overclocking. Penryn is a optimized Core 2 so its better, especially due to 45nm.

Awsome work KTE! You just smacked up all those tests we've been asking for! Now, when (if) you get bored playing with your 790FX board, please try out your Phenom on a AM2 board. (if you got one at your hands)
Thanks!

I have a Sapphire RD580 AM2 board but DIMM slots are faulty on it and they've not returned me a board yet.

Ahhhh, yeah to it looks like to froze at the start of loading the system files and the partition table wacked out. The boot strap loader started but the system files were corrupted, seen that error many times.

This is good news, it was more of a 'improper' windows start up than some wierd BIOS write back to the drive that screwed things up....

Yep, its not bad. If you mess up an overclock, leave it running, and it'll reboot after a while by itself with stock settings and let you know that the oc failed and has been restored to default.

KTE, are you runing default 1.0 BIOS?

YHPM.

Thank you to all and you're welcome.

[PhilDoc]

Wow, great work KTE! You've been working your behind off. I know myself and everyone here appreciates everything you've done. Good luck with the beta bios.

[Andi64]

Thank you KTE!!! Very good work ;D

[JS1234]

Very nice KTE! Great posts!

Sorry for the stupid question, but how do you think Phenom would be for everyday use with a small OC? I was thinking of the 5000 BE or this, and im really thinking about the 2 extra cores

[kl0012]

Some of the benches are real bad. I mean POV-Ray built in bench took around 30 minutes and only ever used 114W AC. That's just above idle and I really don't think if coding was making use of cores efficiently it would be 114W. That shows there's a bottleneck somewhere which the coding is limited to.

It seems you're using version 3.6 which doesn't support multithreading. You need to update it to version 3.7 (only patch is available, no stand-alone version exists).
http://www.povray.org/download/

[KTE]

Quick update 2:

You see my work boss is pretty good. Let's me bring in my system to work if I can get my job finished early.

Today, only once OverDrive worked and never crashed. BIOS voltages for all except RAM worked once. HT multi 3x worked once. WPCREDIT worked once to set NB multi 5x. Sadly RAM did not allow more than 1.9V and auto which is multi 1:2... but still, I have some quick new results before BIOS change. Read on further below...

The BIOS is very bad. Everything is to do with the IMC(NB). It controls yur RAM ratio. You can set 1:1 and 1:2 and that's it. Any higher and no go. You need to give the IMC more volts for it (mine only allowed it once). I mean, how many of you can run DDR2 at 1.9V tWTR 6? Yep, the BIOS only allows this value and this is very low -> will crap out before most 1.9V RAM can reach 820MHz.

Sorry for the stupid question, but how do you think Phenom would be for everyday use with a small OC? I was thinking of the 5000 BE or this, and im really thinking about the 2 extra cores

Inquiry isn't stupid bro. You have nothing to be sorry about.

You can run 2300/2400/2450 quite easily stock volts stable. Problem is RAM. Can your RAM bootup fine at 800-900 5-5-5-18 tWTR 6 @ 1.9V? FYI RAM is normally rated tWTR 10-14 at DDR2-800 2.1-2.2V. If yours can run tWTR a6 at 1.9V, then you can later drop timings from 3-3-3-3 to whatever and you should be OK. If not, then you'll have too many problems, probably not even a bootup.

More than 2.4-2.5GHz would start to get pretty unstable because of the IMC. They need to sort out IMC issues.

It seems you're using version 3.6 which doesn't support multithreading. You need to update it to version 3.7 (only patch is available, no stand-alone version exists).
http://www.povray.org/download/

Thanks mate. Yep, v3.6 is single threaded and I used it because 3.7 Beta was stuck at 4kB/s download the last I tried. I've just d/l 3.7 again for you now at same setup. There were two versions in there: one was povengine and one was povengine_sse2. I ran both:

POV-Ray 3.7


POV-Ray 3.7-SSE2


Power draw: 173-177W AC

IIRC stock QX6700 B3 gets ~2200 in that... ?

Here's some quick OC results:

PCIe = 110


PCIe = 172


PCIe = 195


Never gave any errors. Should be good for 3D benching.

2.6GHz to 2.86GHz



I was at 2.9GHz when I knew the RAM is going to bust any minute. Opened EVEREST fine. Opened CPU-Z fine. Opened Memset to change RAM timing to higher and on Apply-> it froze.

I hadn't even captured it yet. But I took a pic of it with phone. Here ya go:



If you want to see the volts, don't look at anything but these:

CPU VDDQ is VCore.
Mem VDDQ is VDIMM.
NB Core is what it says.
CPU HTT is HT voltage.

All of those are real BIOS values.
CPU VID isn't related to real VCore.

I set 1.480V VCore BIOS, 1.375V NB and 1.3V HT in BIOS. RAM remained auto at 1.9V. It reached 526 at 5-5-5-5/12 tWTR 6 at 1.9V. Not bad at all.

Some CPU-Z 1.42 validations:

2596MHz 472 5-5-5-5: http://valid.x86-secret.com/show_oc.php?id=275584

2706MHz 492 5-5-5-5: http://valid.x86-secret.com/show_oc.php?id=275586

2805MHz 510 5-5-5-5: http://valid.x86-secret.com/show_oc.php?id=275588

Also, a request. Less of the PMs please. Make your queries or asks here, it's much better. Cheers.

[JS1234]

Inquiry isn't stupid bro. You have nothing to be sorry about.

You can run 2300/2400/2450 quite easily stock volts stable. Problem is RAM. Can your RAM bootup fine at 800-900 5-5-5-18 tWTR 6 @ 1.9V? FYI RAM is normally rated tWTR 10-14 at DDR2-800 2.1-2.2V. If yours can run tWTR a6 at 1.9V, then you can later drop timings from 3-3-3-3 to whatever and you should be OK. If not, then you'll have too many problems, probably not even a bootup.

More than 2.4-2.5GHz would start to get pretty unstable because of the IMC. They need to sort out IMC issues.

Great!

About the RAM, i havent exactly bought it yet. Im hoping to buy this computer when all the parts come to a certain computer store nearby...but i was thinking about going out one of these days to get some Ballistix 8500. So im hoping RAM wont be a limiting factor! For 90$ Canadian..i dont know if it can be beat!

[KTE]

The Gigabyte 790FX board has far far less issues than this MSI board. If you really want a Phenom quick then I would get that. Their IMC can run faster, they can control all voltages including RAM, modes and dual channel, their ratios and multi's are easily changed, OD works easy on that and HT can go higher than 245 on it with ease. I can get it to 2.86GHz right now but RAM is very iffy without more volts.

The 9700 is being shipped at places, I know corporations have got hold of them but none are being used. I picked that at first to test but I was called to return it later and given this instead. I wouldn't waste my money, it won't be better than this 9500 at 2420 anyway and you can get 2.85GHz out of this on a decent board pretty easy I would say.

[informal]

I must applaud the ~2.9GHz achievement on your board KTE
Also,those PCIE frequencies of above 150Mhz and stable,are amazing and jaw dropping!

Oh i just saw this over the inq(who knows if it's true,Theo had it's fair share of mistakes in the past):
http://www.theinquirer.net/gb/inquir...re-infamous-l3

So,it seems something goes wrong with above 2.4Ghz speeds which would be addressed in the microcode update soon.All chips will benefit it seems(5-10% gain would be nice if this is true)

[KTE]

You're welcome informal.

The chip was spare idling so I picked it up to test. I also want to sell it off before Xmas -> rather have the extra money. Just wayy too many issues on this board. People will RMA them instantly, that's how bad it is. I haven't messed with it yet again, but I did do some tests earlier for ya'll.

Remember TDP?

Well, I left NB multi default, HT multi default and CPU to 2420MHz.

Threw RAM to 2.3V, CPU HT to 1.40V, VCore to 1.560V, NB to 1.458V

Orthos large in-place FFT max: 248W AC

Prime 95 in-place large FFT max: 264-266W AC



Max power draw observed: @ bootup => 271W AC ~about 225W DC.

That's system power draw.

No case fans, ambient is around 19C, just one CPU fan and max 15 minute P95 all core load temp: 45C. It was steady at that since the 30 second mark.

Very impressive I reckon. But MOSFETs need good cooling.

[informal]

Yep very impressive indeed .How long will you be able to play around with this system?If you are lucky,soon the microcode update will be released and MSI will bring out the new bios.But knowing MSI,it will take 'em way too long to do this..

[JS1234]

Yep very impressive indeed .How long will you be able to play around with this system?If you are lucky,soon the microcode update will be released and MSI will bring out the new bios.But knowing MSI,it will take 'em way too long to do this..

Hmm...is this what you mean?

http://www.theinquirer.net/gb/inquir...re-infamous-l3

We managed to learn that AMD was able to fix that problem through a micro-code update that is being prepared as we speak, and this micro-code update should be implemented in every Phenom from 9600 Black Edition and beyond. If that does not happen, micro-code should come no later than Phenom 9900 (2.6 GHz) in late January.

If you already own a Phenom, this micro-code update will be packaged with one of next versions of BIOS for your motherboard. This fix will enable normal performance at clocks starting from 2.4 GHz and beyond.

This mean ALL Phenom's will be 'fixed' with this microcode update?

[informal]

Yes,that's it.But knowing Theo Valich and his sources,we just can't be sure if he used some "aids" (crack pipe? ) in process of getting these informations .

We have to just wait and see.

[JS1234]

Yes,that's it.But knowing Theo Valich and his sources,we just can't be sure if he used some "aids" (crack pipe? ) in process of getting these informations .

We have to just wait and see.

So you do you think that will address the NB issue? (i think that was a problem, it was running slow...?)

[informal]

Yes,the L3 which had a design flaw(errata) is in NB block(domain) so the fix will address the NB in broader sense.It will allow 2.6Ghz stock speed within the 125W envelope and will probably allow AMD to cherry pick really good samples and label them as BlackBox parts or bin them as 3Ghz parts somewhere in Q1.

[KTE]

The system is mine, he told me to keep it yesterday on phone. I only got it to separate FUD from lies from truth around and first hand know what Phenom is like and to share what I find. TBH I don't need it at, I don't need any system more than a P4 and a C2 lappy I have, let alone this PSU is not something I want to use with it, so I'm going to sell it all ASAP. I also have an E6750, two P4 Northwoods, a 5000 BE and an E6850 lying here unused with some mainboards which is all useless, and I need to still test most of them and want to get rid of them before Xmas.

I might still have Phenom by January, who knows. I don't want to, but I'm not sure yet. I'll definitely be picking up the Penryn when it comes though.

As I mentioned elsewhere, if he had a spare PSU bigger than this or if he would've let me keep the 550W S12II I had before this, I would've picked up 4x Powercolor 3870 800/2240 to test with it. They were there unused for the taking but I didn't have a PSU for them. Although I can still get them or two 8800GT's but no PSU for them and apart from running one or two tests, I won't need them so it becomes pointless travelling 1000 miles for it.

I checked out other guys hitting high NB speeds and high HT speeds which means its definitely the board BIOS. I tried it again twice but it only gets to 2.9GHz and freezes at that. It's confusing because the board is very inconsistent. Something is limited and it's not RAM/HT.

BIOS/system only allows tRFC at 26 max. DDR2-1066 won't work despite volts (NB thing), max is ratio 1:2.

Trying to find OC bottleneck, I've tried very low NB speed, very low RAM speed/high timings, very high volts on all things, mid-high PCIe, very low HT speed -> the result is the same so far.

I'll test BETA BIOSes and then I'll try to see if I can get it Prime95 stable at 2.6-2.7GHz and the volts needed. Will post temperatures and settings used so it'll be easier for others. I might leave it for a week and play with the 50000 now instead. Should be easy, far easier than this anyway.

[JumpingJack]

KTE, when you get a chance... on your next update (and subsequent updates) summarize your clocks, and stability prime95 so far... simply by stating.... "Best prime95 stable 2.8 GHz at 1.4 volts so far"

That would help as I watch this amazing piece of work play out...

Thanks,
Jack