Phenom 9500 w/ MSI K9A2 Platinum

[Suosaaski]

If you have problems, then it's probably your flash technique, as is with most users. You have to use the / pbnc parameters I stated or old BIOS values may be retained.

EDIT: Just seen your above post Suosaaski. You seem to have flashed using the wrong method. Check mine, it works, all registers change.
Which OS you on?

OS is Vista Ultimate 32-bit with SP1

I flashed using the following options

afud412 a7376ams.110 /p /b /n /c

And still I get values 18 and 22 there... even with BIOS 1.1. I need to recheck the flash options... does it matter if it has spaces between the options etc...

Edit: Nah, doesn't even work if I remove the spaces in between.

Edit2: I'm starting to suspect it this has something to do with Vista. Especially SP1. I'll install XP and see if things change... Anyone think TLB-fix might be built into Vista SP1? And thus even with a BIOS without an option to disable it, would in fact have it enabled?

[KTE]

OS is Vista Ultimate 32-bit with SP1

I think I've commented on that note already.
It's known to be a very bad performer, esp. MEM/IMC perf.

Uninstall SP1 or doubt that's changeable. Meanwhile you can use the app. I linked at the start of the thread to disable it.

afud412 a7376ams.110 /p /b /n /c

That's right (for BIOS 1.10), yup. BIOS is fine then, it's Vista SP1.

1.1 doesn't have patch, if you get those values, then Vista SP1 has it instead.

[madfaze]

thanks Suosaaski got your messsage

[Suosaaski]

Installed XP 32-bit. Good news.

With BIOS 1.1 all values read 10 and 20 instead of 18 and 22.

Performance with the WinRAR bench went up to a bit under 1800KB/s (no optimizations, RAM at default 800MHz CL5, clean install, no drivers and so on).

So, even though I did not test with Vista without SP1 thoroughly, I did earlier notice that my performance went down for some reason (I got higher 3dmarks etc, with Vista RTM), just didn't figure out what caused it... seems it was SP1 and TLB-fix.

My conclusion is simply that SP1 for Vista enables the TLB-fix via software, so regardless of BIOS-version it is enabled... unless bypassed via software (like AOD, which disables it for first core only, though). I should also test with Vista RTM, but since SP1 is coming, performance eventually goes down.

Hope it is not included with XP SP3. Anyone tried the test versions of SP3 to check this out?

[madfaze]

hmmm updated my bios to p0h, had the 9500 phenom how come the multiplier seems unlocked up to x25 multi?!? can i use that for OC? thought 9500 is locked to x13 multi..

[KTE]

Hope it is not included with XP SP3. Anyone tried the test versions of SP3 to check this out?

Nope. SP3 is better and quicker than SP2. I've used both.
Vista was known to be messed up this way even though Microsoft did not state anything as such in their logs. We need initiatives where they are fined if they withhold such information we can clarify.

hmmm updated my bios to p0h, had the 9500 phenom how come the multiplier seems unlocked up to x25 multi?!? can i use that for OC? thought 9500 is locked to x13 multi..

Options or if chosen doesn't mean it works. It just doesn't change multi's above max multi the CPU can run which are: 11x CPU, 9x NB, 9x HT for yours.

[madfaze]

Options or if chosen doesn't mean it works. It just doesn't change multi's above max multi the CPU can run which are: 11x CPU, 9x NB, 9x HT for yours.

i manage to change 11.5, 200fsb and it OCed to 2300mhz accrding to cpuid...

one thing i notice can i change my NB setting to x13 or is this the HT speed?am running it to auto..

when i change it to x13 NB frequency reading is 2600mhz and HT speed is 1800mhz, is this normal?

[dr_drache]

any chance i can get P0J from someone to see if i can make it work? (on accound P0H still causes BSOD even with new install)

edit : i don't know if i even should.. hmmm

[Mathos]

Hmmm does the new P0J bios have the option to change nb/IMC multi again? And possibly voltage/vid if needed? I may consider updating to that one since the TLB fix can be fully disabled. Granted as far as NB volts go I shouldn't have much issue since I was doing 2.4 at stock VID before.

[KTE]

i manage to change 11.5, 200fsb and it OCed to 2300mhz accrding to cpuid...

one thing i notice can i change my NB setting to x13 or is this the HT speed?am running it to auto..

when i change it to x13 NB frequency reading is 2600mhz and HT speed is 1800mhz, is this normal?

Pretty normal, a known bug.

Check AMD OverDrive and see what real values it reads
If it changed, then it would read what they read for mine.
Also even better, benchmark it.

Why was your HT stuck at 9x? Because so is your NB.
Start AOD, you won't be able to move HT plus 9x because your NB is only 9x real. If you actually moved it up, the HT will move up along with it. We've been through this many times in this and many other threads.

any chance i can get P0J from someone to see if i can make it work? (on accound P0H still causes BSOD even with new install)

On MSI site for a while, but courtesy of Suosaaski: http://www.yourfilehost.com/media.ph...e=A7376AMS.P0J

Your own liability entirely.

Hmmm does the new P0J bios have the option to change nb/IMC multi again? And possibly voltage/vid if needed? I may consider updating to that one since the TLB fix can be fully disabled. Granted as far as NB volts go I shouldn't have much issue since I was doing 2.4 at stock VID before.

Nope.

On a side note, while underclocking using MSR values, have a look at this using decimal multiplier and DID. Very very very slow, took long to save a basic 200KB image, about 45 seconds.








Lowest underclock? Quite low that, it crashed very soon after the 1M run due to changing volts past 0.850/1.038. Power for that idle and load differed 2W AC.

Anyway, I've tested very slow, observantly and hard. I's true, my 4th core is the weak link and by far. Any time my system freezes or reboots, it's the 4th core spiking and getting stuck at full load. Luckily the 4th core is not needed for much at all. When I tested stability with high NB volts and higher CPU volts not gaming/lengthy testing on just core3 but on all others, for over 15hrs now, I've had no failure or error. This is what I'm running since then.

[KTE]

Xbit Labs article:

In the end of March – beginning of April AMD is going to introduce its Perseus platform offering great stability, manageability, security and low power consumption for commercial environments. Perseus platform is regarded as a primary competitor to Intel's vPro platform. It will include quad-, triple-, and dual-core processors based on AMD's K10 and K8 architectures, integrated chipsets from the 780 series announced yesterday and optional graphics cards from R6XX family working in Hybrid Graphics mode. It is important to specifically stress Hardcastle technology dealing with security and management that should strengthen Perseus positions in the commercial market even more.

After that in the first half of April AMD is going to introduce a few quad-core Phenom processor models based on B3 core stepping and free from the notorious TLB bug. We should see Phenom 9650 and Phenom 9550 working at 2.3 and 2.2GHz frequencies respectively. At the same time the first triple-core Phenom 8600 and Phenom 8400 processor should also come out. They will at first be based on B2 core stepping and work at 2.3 and 2.1GHz frequencies respectively. All these four new processors will feature 95W TDP.

The second half of April will be remarkable due to upcoming Cartwheel platform launch that should represent mainstream performance, immersive HD and upgradeability. This platform is in fact available since the launch of 780 chipset series, because it uses mainboards based on these chipsets, processors and optional R6XX graphics cards. However, the key feature of the upcoming Cartwheel platform is in fact AMD LIVE! Technology support offering media-centric computing experience that connects and extends the digital world by means of special software.

In May AMD is going to continue expanding the frequency potential of their quad- and triple-core processors. They are expected to launch quad-core Phenom 9850 with 2.5GHz frequency and 125W TDP and Phenom 9750 with 2.4GHz nominal frequency and 95W TDP. New triple-core CPUs, just like their quad-core counterparts will start using B3 core stepping. As for the particular CPU models, we should see Phenom 8750, Phenom 8650 and Phenom 8450 with 95W TDP and 2.4, 2.3 and 2.1GHz respective frequencies.

The new mobile Puma platform is scheduled to be launched in June. It is expected to ensure best-in-class experience and performance-per-watt for consumer/commercial notebook users. Puma will include a new "Griffin" dual-core processor, an RS780M chipset with DirectX 10-class integrated graphics, a 55nm M8x mobile graphics processor, and support for the PC management software. The platform will also feature PowerXpress technology which allows the system to switch between external and integrated graphics when changing between AC and DC power without the need for a reboot. Also, Hybrid CrossfireX allows the integrated graphics hardware to work in tandem with the separate graphics controller to offer more power for gamers.

Catalyst 8.3 for Vista 32b: http://downloads.guru3d.com/AMD-ATI-...load-1871.html

Supposed to be the revolutionary driver, CrossFireX, Hybrid Tech, etc.

[madfaze]

which phenom has a good batch?the one ive been testing was 0743 FBPW... is this a good batch?

[KTE]

0743 are very good, should do 2.75G benchable at least.

[madfaze]

0743 are very good, should do 2.75G benchable at least.

really?! stock vcore?!

cool...


not tested so far cause i had an isue with our board specially on the link width of pcie..sometimes it read out x4 only (instead of x16)...do you happen to experiece that too?!

[justapost]

On a side note, while underclocking using MSR values, have a look at this using decimal multiplier and DID. Very very very slow, took long to save a basic 200KB image, about 45 seconds.
Lowest underclock? Quite low that, it crashed very soon after the 1M run due to changing volts past 0.850/1.038. Power for that idle and load differed 2W AC.

Thanks for the update, thought higher DID's than one (2x) did not work. Tried 600MHz with success. 300MHz also works but it's too slow.
For the record, 0.85V VID was lowest stable for you, no matter what freq?

Update:
broke 100W ground here

96.3W AC idle

100MHz (0x48009500)


Can switch between 100MHz/0.625V and 2.6GHz/1.3625V via reg c0010062h without problems.

UPDATE: 500MHZ/0.7V is somewhat comparable to by old celeron laptop at 600MHz, still onyl 97W AC consumption in idle. I like the higher DID's.

[KTE]

There's a new BIOS, not sure about what it brings. Will try it and see: A7376ACI.101

really?! stock vcore?!

cool...

Air, if cooled, it will bench 2.7G stock volts. I tried two of them, both did, others had some which did the best so far too, some above 3G.

not tested so far cause i had an isue with our board specially on the link width of pcie..sometimes it read out x4 only (instead of x16)...do you happen to experiece that too?!

Nope, first time hearing it. Set PCIe speed to Auto in BIOS and see what happens.

Which PCIe are you using BTW? Use the one closest to the CPU.

Thanks for the update, thought higher DID's than one (2x) did not work. Tried 600MHz with success. 300MHz also works but it's too slow.
For the record, 0.85V VID was lowest stable for you, no matter what freq?

I haven't had any time to try things, only tried 2-3. I was at 207MHz HT and so couldn't try lower, had much work open so not sure. I just left volts as they were for previous 800MHz CnQ setting
I read some ignorant individual online saying AMD Phenom has problems that it can only go to 800MHz power saving, and it's obvious he had no clue what he was talking about, so we proved him wrong.

400MHz is fully stable, even 150MHz was fully stable switching between ~150MHz to 2691MHz. I never tried lower yet.
I think we're making very good ground here Achim, excellent tweakers chip its become.
It looks to be too complicated a chip for too many out there to get most out of though.

You're right, below 400MHz gets slow, jitters and stutters. 400MHz seems like best CnQ option to run and that's what I've been running since my last post. I also have 3rd core downclocked (13x core 0,1,2 and 12.5x core3). This is what I'm running for a day now:




Enable CnQ:

Update:
broke 100W ground here
http://www.abload.de/image.php?img=96acc2b.jpg
96.3W AC idle

100MHz (0x48009500)
http://www.abload.de/image.php?img=100dh0.jpg

Can switch between 100MHz/0.625V and 2.6GHz/1.3625V via reg c0010062h without problems.

UPDATE: 500MHZ/0.7V is somewhat comparable to by old celeron laptop at 600MHz, still onyl 97W AC consumption in idle. I like the higher DID's.

Nice, glad you got below 100W, finally

I am idling 93W AC with the above CnQ config (2xHDD,1xOptical,2xfans). Same as 800MHz/0.850V-1800/1.038. Which means, the power consumption for CPU at the 1GHz mark idling is negligible and most of the AC power is the rest of the system. Phenom at 200MHz/0.600V idle cannot be consuming much power at all, it won't even be 10W DC IMHO.

Quick calc: 95*((200/2300)*(0.600²/1.232²))
= 1.959W TDP

So CPU alone will have very low power idle at those settings, the rest will be nearly all other system component power. That's why CPU power is not going down much and neither is CPU temperature.

That's kinda incredible that it can run stable from 100-2700MHz, no core that I know has done that or given the user the ability to run CnQ in such a way. BTW I have the Abit 770 and the GBT Odin since a few hours now. Going to set it up morning, too busy now before sleep.

[Suosaaski]

There's a new BIOS, not sure about what it brings. Will try it and see: A7376ACI.101

It's supposedly 1.3 with option for TLB-fix disable. I asked for MSI with an option in BIOS for it and they uploaded that BIOS to their FTP-server a few days ago.

I am not sure what it is based on though, my guess is 1.3.

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.

That bolded one...

[madfaze]

Nope, first time hearing it. Set PCIe speed to Auto in BIOS and see what happens.

Which PCIe are you using BTW? Use the one closest to the CPU.

yup i always use the first mezzanine the one nearest to pcie, have to reseated to bring x16..well its my first time too, wish they had a locked in it...

[KTE]

Guys, some needed updates and news

Switch to EVEREST Ultimate Build 1315, it fixes wrong clocks/multi detection issue for Phenom: http://www.lavalys.com/beta/everestu...s6wdlkn2cx.zip

Switch to Catalyst 8.3 (with Southbridge drivers), quad-fire working fully and many many improvements featured: http://game.amd.com/us-en/drivers_catalyst.aspx

Latest AMD CPU Info: http://www.amd.com/us-en/Processors/..._15259,00.html

AMD has supplied a Phenom with TLB Fix working in BIOS compatibility list for AM2+ and AM2 motherboards, details: http://techreport.com/discussions.x/14288

Here's the list AMD provided from their own testing and that of MB MFGs:



I have the Abit AX78 here so I can check that to confirm what the MB MFG says. All those run Phenom perfectly at stock too by AMD and/or MB MFG testing.

It's getting better and better, as expected, AMD is supporting us well.

[madfaze]

I have the Abit AX78 here so I can check that to confirm what the MB MFG says. All those run Phenom perfectly at stock too by AMD and/or MB MFG testing.

It's getting better and better, as expected, AMD is supporting us well.

how was AX78 doin?

[KTE]

how was AX78 doin?

http://www.xtremesystems.org/forums/...6&postcount=50

Can't put my own X4 into it because I'm testing some software/tweaks on it yet in the MSI. Tested it using someone else's 96BE.

[Kayin]

Can you shed some light on Vista 64 and this board's not wanting to play nice together? I need 4+ gigs of RAM, but I can't afford a crashing computer either. Will the new drivers do anything?

[justapost]

I read some ignorant individual online saying AMD Phenom has problems that it can only go to 800MHz power saving, and it's obvious he had no clue what he was talking about, so we proved him wrong.

Ah looking at your screenies 0.625 also works for you.

400MHz is fully stable, even 150MHz was fully stable switching between ~150MHz to 2691MHz. I never tried lower yet.
I think we're making very good ground here Achim, excellent tweakers chip its become.

The high power consumption due to overclocking in idle always bothered me. With a tweaked ps-1 this is no problem anymore.
p-state tweaking is also neat for testing individual cores, even with cnq. As soon as you remove the load you are save.
Gonna try to enable more than two p-states this weekend if i find the time, also have a small gui app in the works, might take a week, can work an hour max per day on that.

I am idling 93W AC with the above CnQ config (2xHDD,1xOptical,2xfans). Same as 800MHz/0.850V-1800/1.038. Which means, the power consumption for CPU at the 1GHz mark idling is negligible and most of the AC power is the rest of the system. Phenom at 200MHz/0.600V idle cannot be consuming much power at all, it won't even be 10W DC IMHO.

Quick calc: 95*((200/2300)*(0.600²/1.232²))
= 1.959W TDP

So CPU alone will have very low power idle at those settings, the rest will be nearly all other system component power. That's why CPU power is not going down much and neither is CPU temperature.

Now with those low cpu TDP the difference in AC shrinks, you have one hd more that might make 5W max difference. Gigabyte PSU should arive here next week.

That's kinda incredible that it can run stable from 100-2700MHz, no core that I know has done that or given the user the ability to run CnQ in such a way. BTW I have the Abit 770 and the GBT Odin since a few hours now. Going to set it up morning, too busy now before sleep.

Wonder in what degree the cpu will suffer from higher voltage and temperature differences over time.
But having different p-state profiles for manual switching will come in handy here.

[KTE]

Guys, you'll have to excuse me with time because my systems have all failed to boot. No idea why. It happened when I tried changing PSU.

Phenom is not booting at all in MSI or Abit board with 3 different PSUs, same problem as before when you switch power off (defaults), stuck in reboot cycle. Spent 4 hours on it.

Gigabyte PSU is totally DOA on all boards. Tried 4 boards, MEM/CPU doesn't even start.

Two C2 systems also no boot with those 3 PSUs but luckily an older P4 worked with the Antec and the Seasonic which I'm posting off right now. Spent hours troubleshooting this.

At least the P4 idles 55W AC.

Achim: Download GBT 550W PSU guide here, there is no guide but on CD in box and you might find trouble installing it (it has many things): http://america.giga-byte.com/FileLis...s800a-d1_e.pdf

Kayin: Check the new BIOS, it should have added compatibility with Vista 64b for many RAMs. I'll check later on what is officially supported. The problem is mainly BIOS for support.

[Mathos]

Well, my initial findings just in the last few mins after updating to P0j seem to be good. The disable option for the TLB fix works. Was getting around 1596KB/s on winrar at stock speeds, and 1684KB/s while running the cores at 2.6ghz. If I could set the NB/IMC up to 2.4 like I had with 1.1b3 I could probably get back above 2000Kb/s. I notice the HT can be set now, I'm wondering, does changing that also raise the NB/IMC multi to match?

OCing the cores seems to be going a lot smoother, even with C&Q enabled in bios. Switched the multi up to 13x, disabled spread spectrum and boot into windows fine, started up everest beta1316 without issues too. Before when I was OCing with 1.1b3 it would stall for a second at the loading drivers part then continue. But this one has given me no issue so far. Even left cpu voltage at default in bios, Everest is reading at 1.24v core, which is stock when not in idle mode. But it's showing the VID as 1.250v in AMD pow mon. Among other good news things, the new beta of Everest actually reads my voltages properly, so my 12v rail is show at 12.67v like it does in bios instead of 8.1v like it was before.

I think what I'm gonna do is bug MSI's customer support thing, and see if I can get em to put together a bios that allows for NB/IMC clock and voltage changes like the ones that had the p-states section. Since apparently thats how that ACI.101 bios came about.