[Sammelthread] AMD K7 - Sockel A (462)

[Sparksnl]

Wow. If only it didn't have half the cache missing, that would be a great little cpu.



Now that's the usual picture you'd expect with most Thortons. High voltages for relatively low overclocks. I've even seen some which needed 1.85V for 2200MHz already.

I know, my third one is more like that...

I will look at the cache issue. Are these not laser cut and for that it should be impossible to fix?
The good one is a week 51 cpu.

 

[Tzk]

No clue if these are laser cut. I'd fill the L2 Bridge gaps + paint it on a bad clocking cpu and try it. However not every Thornton got a fully working L2 Cache, so it comes down to luck if a) the cpu is a good clocker and b) the cache is fully working. A quick google search showed some success stories for the Thornton -> Barton mod, so in theory it works.

 

[Sparksnl]

I was thinking about doing that. Try with the lesser god.
But I will use it for L9 bridges.

 

[Sparksnl]

Digitalbath EBED BIOS test with 2x512MB Corsair TCCD:


Boot test/quick test

 

[The Sandman]

Stress it?

 

[Tzk]

Tref 0016? Is that value accurate?

 

[Sparksnl]

Tref 0016? Is that value accurate?

It is on auto so maybe.

I first want to look of this is max fsb. But I can try and test it, hopes are not high though...

 

[digitalbath]

Strange, I hope the 16 is not BIOS related. As far as I remember, I never used tREF=16 for my option.rom. Maybe it is just the tool?

Edit:
For TCCD dimms, I would use a value somewhere between 2000 and 3000.

 

[Sparksnl]

With these settings it is not PI 1M stable at 260MHz, I expected this because this bord comes unmodded and max is 260MHz...

edit: when setting it by hand in bios



So not tool related?

Edit 2: actually... it is booting with TREF @0516 as well...



Edit 3: The good news is that 2x256MB TCCD boots a little higher (+1MHz so 261MHz). Not much of a difference but still.

Edit 4: PI1M @260MHz FSB with 2x256MB is successful! (PI32M is not)

 

[digitalbath]

tREF 0516 and 2592 are settings from the BIOS? If yes, then the option.rom should work normally.

Maybe I shoud explain how the additional settings work. The System starts with the default values in tRC, tRFC, tREF, DS, SR, alpha timings and romsips. Right before the system boot from hard drive the option.rom starts as the last module in BIOS. The option.rom reads the additional settings you set in BIOS. According to this the option.rom change the values you set before.
If you set "auto" in the additional settings, the option.rom skips the change for the active setting and jumps to the next setting.

So, if tREF is set to "0516 cycles" and boots with 0516 the option.rom should work properly. I am not sure why you get a 0016 value with a "auto" setting in BIOS. Maybe it is related to the tool?

could you make two screenshots like this one ?:

one after a bootup with tREF="auto" and one after you start the tool?

 

[Tzk]

Of course it will boot with low tref values. That causes lots of (unneccesary) refreshes of the dram cells of the memory. That‘ll prevent flipped bits, but also cause a huge impact on bandwidth.

Default at 200mhz is 1560 cycles. You can usually raise it to 2560 at 250mhz or even 3120 above 270mhz

 

[The Sandman]

260-MHz TCCD 32M kann meines auch nicht. Rumtumnen, bis es klappt vielleicht. Instabiler Krampf.... lassen sie das
BH-5= Love

 

[digitalbath]

@The Sandman
...heißt ja nicht, dass es nicht geht .
BH-5@2-2-2-5/250MHz sind wohl schneller. Da hast du recht.

 

[Tzk]

NICE!

264Mhz 32M mit 2x512mb TCCD sind ne Ansage.

 

[The Sandman]

lobby low
ciao
aber sonst richtig geil +++Weil AsRock

 

[digitalbath]

@Tzk
Danke. Das schlimme ist ja: Soviel wie das ASUS bei den TCCD Riegel rum krampft, so wenig interessiert dem popligen ASRock K7NF2-RAID das ganze. Für die TCCD Riegel brauchte ich noch nichtmal mit den DS/SR rumärgern oder irgendwelche alpha-timings bis an den Anschlag drehen.

lobby low
ciao

mit 264MHz wahrscheinlich nicht. Mit 260MHz könnte ich es vielleicht noch hin bekommen.

 

[Sparksnl]

tREF 0516 and 2592 are settings from the BIOS? If yes, then the option.rom should work normally.

Maybe I shoud explain how the additional settings work. The System starts with the default values in tRC, tRFC, tREF, DS, SR, alpha timings and romsips. Right before the system boot from hard drive the option.rom starts as the last module in BIOS. The option.rom reads the additional settings you set in BIOS. According to this the option.rom change the values you set before.
If you set "auto" in the additional settings, the option.rom skips the change for the active setting and jumps to the next setting.

So, if tREF is set to "0516 cycles" and boots with 0516 the option.rom should work properly. I am not sure why you get a 0016 value with a "auto" setting in BIOS. Maybe it is related to the tool?

could you make two screenshots like this one ?:
...
one after a bootup with tREF="auto" and one after you start the tool?

Where can I find this tool?

Edit: Just as a reminder for myself I need EBED for screenshots.
Edit 2: Ah is this RWEverything?

 

[digitalbath]

here.

edit.
yes, this tool is RWEverything. A nice tool for reading PCI register, CMOS register, DIMM SPD, ....

 

[The Sandman]

Infra war schon lange nicht mehr on...
braucht mal wieder nen Hammerschlag lol !

 

[Sparksnl]

I will try this versio because what I have crashes...

 

[digitalbath]

Crashes when you load RWE? Strange. I never had crashes with RWE.

 

[Sparksnl]

Crashes when you load RWE? Strange. I never had crashes with RWE.

It is typical, I have it on most systems.
Can it be a hardware incompatibility like my hd4350 pci?

 

[digitalbath]

I don't know. Maybe Windows itself? I use it only on Win XP.

You can also make a screenshot with wpcredit. It is able to read PCI register. When you start the program, press the greed button (select device), then select "Bus#0 Dev#0 Func#1"
You can find wpcredit here.

 

[Tzk]

I found that sometimes strange things happen after a few lockups or bluescreens... i‘ve had a windows setup which would crash on every cpu-z start. Reinstalled, works perfectly fine now.

 

[Sparksnl]

I could try, but reïnstall... I back up for this reason. But since I have more platforms with this problem I have some sort off issue.

I do have more to come... I feel some cold sessions comming in the next months. I ordered something nice.

 

[Sparksnl]

I just tested to start RW Everything with: HD4350 PCI - Crash Windows
I just tested to start RW Everything with: Readon 9700 AGP - Starting just fine.

Something in the driver must be causing this. A diagnostic start was not enough.

Edit: RWEverything screens for EBED

TREF is manual 2560



TREFT is auto in BIOS.

 

[digitalbath]

many thanks!
That is what I wanted to see. The good news: the BIOS and the option.rom works 100% correct. The bad news: the tool not.

explanation:
If you look on the second screen, you see the values at 60=1A and 61=06. tREF is 16bit, that mean you get together the hexadecimal value = 061Ah (h means it is a hexadecimal value). In decimal the value is 1562cycles (7,8ns x 200MHz), which is a perfect value for a 200MHz DIMM.
For the first screen you get the hex values 0A00h. In decimal it is the destinated value tREF=2560 cycles

 

[Tzk]

So this is a bug? if yes, then @I.nfraR.ed should have a look at it, maybe he can fix it. I guess the 00 in register 60h causes these issues.

 

[Sparksnl]

I sent him a private message on hwbot about it.

 

[stunned_guy]

Die HWBot Seite lädt wieder so schnell wie früher. Die hatte sich die knapp letzten 1,5 Jahre so dermaßen langsam geladen, dass ich die Lust daran verloren hatte. Sofort ist meine Lust wieder da, muss ich doch gleich mal ein paar Benchmarks machen und hochladen Mal schauen, was mit einem Asus A7N8X machbar ist.