Phenom 9500 w/ MSI K9A2 Platinum

[KTE]

That's half of it at least. Why I'm after a detailed screenshot is in important detail you'll miss typing and easier in pics.

These:

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).

If you can check your RAM Drive Strength using EVEREST then do that too.

[KTE]

3DMark'06 CPU:
XP 32b, Phenom at 2652MHz low NB/HT, loose timings, 2k6 CPU score:



Use it as a comparative as it's totally untweaked. I was testing low volts stability there with a low card.

And to update on Vista X64:
A user here is running 257x11 on 32b and 255x11 64b in Vista with Phenom and the MSI RD790 daily stable. ASUS RD790 needed more volts to run the same oc.
BUT, this user experienced behavior the same as my chips have done. Before his chip could not oc past 240HT at all, completely impossible for it to, would reboot with instability. Now, it goes much past fully stable, after many weeks of use.

[madfaze]

havin hard time to set my mem timings to 4-4-4-12 had a four sticks of corsair mems, 9500 phenom set my vdimm already @ 2.1v NB to 1.2v...

my system wont boot at windows either...

[KTE]

Try only 2 sticks madfaze.

[madfaze]

Try only 2 sticks madfaze.

ok try it later when i got home


how about the NB volt?should i adjust it too...pretty weird it also hangs during POST and adjusting it in the bios...


highest i could get so far 228 fsb at stock vcore...

[KTE]

OK then.

Any of you guys miss this? http://biz.yahoo.com/bw/080304/20080304005451.html?.v=1



Deneb and Shanghai are up and running multiple OS and stress apps perfectly, shown at CeBIT. Didn't I tell you they have the chips working early Feb.?
And yet, one sided trolls obviously were hurt there.

Now, no hype. They just have a chip running, nothing much you can say about that but that they've made one function the basic logic. Still doesn't mean nothing about the timeline.

[informal]

Great news,i saw it in NEws section.Still we have some powerful spintel trolls to bash AMD even for this .
My guess is Q3 for shipping parts .Q4 for top freq. bin.

Anyhow,just found this here:


2 additional speed bins of Phenoms (B3) come in April and May .I guess these will be 2.4 and 2.5 in April,followed by 2.6 and 2.8GHz(AM2+ FX part?).Note the plural used in both instances on the roadmap ,so it means at least 2 speed bins per month,and since we have now 2.3Ghz i guess it means the speeds will be the ones i wrote.Let's hope they are,it would imply more easier 3Ghz mark for OCing.

[KTE]

Well... they're 2.4G and the 2.6G parts that we know of, then the FX parts late Q2/Q3. They may launch two FX models, one followed by another and that depends on many things including B3 tweaks and TDP gains. FX82 and FX84.

They do have Agena FX 2P too BTW, just search and you'll see it has been shown for long. But it was scrapped for mass production till the late steppings or 45nm, not the least because they didn't have the resources to continue with all the problems to fix K10h, tight schedule and 55nm/45nm efforts. So far, they've turned things drastically around in GPU, chipsets and IGP, CPU to go.

[informal]

Yeah i know it was shown since last year,but i was thinking about common "poor man's version" of FX ,for AM2+.
2P is too narrow (profit wise) for them to be concentrated on it.With Deneb FX they can finally address this segment since the chip will be a lot smaller(and faster,IPC and freq. wise).
I can't wait for Shanghai revision,that will be my next upgrade .Till then i'm perfectly happy with my X2 .

[dr_drache]

ok, i did as you asked KTE, all those windows open, and I had crystalCPU open to show that the registers are 18, not 10 as if it was not TLB patched...

(all images are resized to 1024/600)

first one without patching :


2nd screen shot, only thing changed are the MSR's

[Suosaaski]

That's half of it at least. Why I'm after a detailed screenshot is in important detail you'll miss typing and easier in pics.

These:

If you can check your RAM Drive Strength using EVEREST then do that too.

Here's with BIOS version 1.1:
http://img116.imageshack.us/img116/1...arbenchvo5.jpg

And here with P0H:
http://img186.imageshack.us/img186/2...enchp0hja1.jpg

Forgot the SPD tab, will get new screenshots with it and I'll also download the trial for Everest to show the RAM Drive Strength when I get back home.

All tests done at default settings. The OS version is Vista Ultimate 32-bit with Service Pack 1.

PS. Will check the CrystalCPUID values too later today.

[Suosaaski]

Received a new BIOS version from MSI (P0J) after asking for a modified P0H that disables the fix for all cores instead of just the first. I'll test it later today.

[KTE]

Latest EVEREST: http://www.lavalys.com/beta/everestu...ori5j1htcv.zip

Sorts out random clock/temperature jumps and now shows Phenom 9500/9600 separate to Phenom 9600 BE. Top result I included just to show you a difference of 11x200 NB multi (top one) and the rest are all 9x NB multi.



790FX board popularity so far... price sells, MSI MB sold out in many places >



MSI K9A2 Plat. V2 is just a down-featured revision of V1. It is cheaper and has a lower MSRP than the V1 too, as I had expected.

ok, i did as you asked KTE, all those windows open, and I had crystalCPU open to show that the registers are 18, not 10 as if it was not TLB patched...

(all images are resized to 1024/600)

first one without patching :
2nd screen shot, only thing changed are the MSR's

Very well. Now that perf. is on par with what I expect. Vista is poor in many things and SP1 is actually much slower for memory/IMC intensive tasks than no SP1 in Vista. The guys getting very poor performance compared to XP is because of their SP1 upgrade and because of Vista itself.

Here's with BIOS version 1.1:
And here with P0H:

All tests done at default settings. The OS version is Vista Ultimate 32-bit with Service Pack 1.

PS. Will check the CrystalCPUID values too later today.

OK. Thanks. That's definitely patched.

Received a new BIOS version from MSI (P0J) after asking for a modified P0H that disables the fix for all cores instead of just the first. I'll test it later today.

Can you host the file somewhere please and link it back here.
I'm the one who told them how to disable it when they didn't know and yet I don't even get the BIOS.

[Suosaaski]

OK. Thanks. That's definitely patched.
Can you host the file somewhere please and link it back here.
I'm the one who told them how to disable it when they didn't know and yet I don't even get the BIOS.

Sorry, I don't have a place to host it.

It's on their FTP server now, you just need to ask for a password via http://ocss.msi.com.tw/ or I can send it to you via e-mail if you wan't.

[KTE]

Sorry, I don't have a place to host it.

It's on their FTP server now, you just need to ask for a password via http://ocss.msi.com.tw/ or I can send it to you via e-mail if you wan't.

You can host it here for free: http://www.yourfilehost.com/flash_upload.php

Then link back

Waiting for their reply will take too long or PM me the password if you have it.

[Suosaaski]

You can host it here for free: http://www.yourfilehost.com/flash_upload.php

Then link back

Waiting for their reply will take too long or PM me the password if you have it.

I'll check it out.

[madfaze]

@Suosaaski

did you already installed the BIOS?


are experiencing any BSOD?


some peeps @ MSI Forums are experiencing BSODs


check it here

[Suosaaski]

@Suosaaski

did you already installed the BIOS?


are experiencing any BSOD?


some peeps @ MSI Forums are experiencing BSODs


check it here

Nope, not yet. That seems to... suck. I wonder if I'll install it or not.

[KTE]

They give no details how or what they did. They can easily boot using UBCD4Win or Debug Mode and check the debug logs to see what is wrong. It's not necessarily the BIOS according to that info.

Secondly, that BIOS incase you didn't now is an AMD release BIOS for testing purposes only, dubbed "Performance BIOS".

I'm flashing it in ~20 mins.

[madfaze]

some of my test runs at stock vcore, bios p0h





image 1

image 2

image 3

image 4

image 5


validation




do i have to lower HT speed?

[Suosaaski]

Secondly, that BIOS incase you didn't now is an AMD release BIOS for testing purposes only, dubbed "Performance BIOS".

I'm flashing it in ~20 mins.

Yup, I knew that. I have the P0H for daily use, works fine so far.

Let us know how it works for you.

[madfaze]

how was the new bios KTE?

[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.

Bad news (BIOS P0J):
The first tests: First core ends 10, the rest 18
The second values: First core ends 20, the rest 22

So, that means that for P0J it is disabled for the first core only, too. Am I right?

PS. Need to re-check, seems AOD messes with it a bit, at least changing from "red" to green results in the first core reporting 18 and 22 too...

PPS. Slightly off topic (but still within the topic of the thread), I do like the fact that with P0J (don't remember how it was with P0H) I don't have to mess with P-States anymore, as there is simply the option for cpu multiplier... might try overclocking via bios now...

PPPS. Doesn't seem to do anything wether I set it to enabled or disabled ("AMD's TLB-fix") via BIOS, it reads 18 and 22 for every core, and with AOD at "red" only the first core reads 10 and 20. AOD at "green" or "yellow" reads 18 and 22 for all cores.

[madfaze]

Bad news (BIOS P0J):
The first tests: First core ends 10, the rest 18
The second values: First core ends 20, the rest 22

So, that means that for P0J it is disabled for the first core only, too. Am I right?

PS. Need to re-check, seems AOD messes with it a bit, at least changing from "red" to green results in the first core reporting 18 and 22 too...

PPS. Slightly off topic (but still within the topic of the thread), I do like the fact that with P0J (don't remember how it was with P0H) I don't have to mess with P-States anymore, as there is simply the option for cpu multiplier... might try overclocking via bios now...

have you upload the bios file that KTE posted?so that we can have a copy too...

[KTE]

Tested BIOS P0J. The best performance BIOS is 1.13 BETA and then P0H after manual disabling using the tool I linked in the first post, so let's compare with it at exact same clocks, background, timings and so on.

V1.13:


VP0J:


Bugs fixed
*HT can now be chosen in BIOS and set.
*All RAM dividers and timings work fine.
*TLB patch is completely enabled/disabled with the BIOS option.
*Checked the registers, all are changed.

Runs fine, I booted and have been running my 2665 settings since about 40mins now. I told you don't believe rumors.

Problems
*One hex value in the C001_1023 MSR is Ah instead of 2h now. This might be what causes lower performance (I'll check).
*As you can see above, the memory/IMC performs lower, slower all round than 1.13 or any other BIOS where you disable patch manually. In the below screenshot my older 2665 run is exactly the same in every setting as my new 2665 run (shown as 2666). The old is BIOS P0H ran just before flashing new BIOS and the new is BIOS P0J both without the patch (checked registers manually).



Big difference here and in latencies even though I actually had a lower MaxAsyncLatency value here than in the older P0H 2665 run.

So, it is slower, offers lower performance than older BIOSes but has the TLB patch option to disable properly within the BIOS. Still should work flawless for many.

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?