[Sammelthread] AMD K7 - Sockel A (462)

[Fusseltuch]

Willst du viel, spül' mit Pril ! 😛

 

[digitalbath]

Im Nostalgieluxx gibt einen running gag:

damit soll es am besten funktionieren.

 

[Tzk]

Jetzt ganz ohne Witz:
Ich nehme einen Schweinsborstenpinsel, Spüliwasser und dann gehts ab. Danach spüle ich mit 70% IPA hinterher und lasse mindestens 24h trocknen, damit Rückstände unter BGA Chips (RAM, Chipsatz) verduften können. Testlauf und geht

Und ja, ich hab extra dafür Fit gekauft, weil das ein innerer Zwang war

Edit:
Die ganz harten packen Boards auch in die Spülmaschine. Der8auer zum Beispiel …

 

[stunned_guy]

Und immer schön mein Bild für den Gag misshandeln
Ich finde den Running Gag immer wieder klasse !

 

[the_patchelor]

Fit ist doch für Anfänger

Mein schmutziges Geheimnis:
-mit Wattestäbchen oder Mini Bürste/Pinsel und Isopropanol besonders hartnäckige Verschmutzungen lösen (klebrig, schmierig)
-wenn zuvor gelötet, dann auch mal punktuell Platinenspülung/Flux Entferner oder Elektronikspülung
-Dann Duschen: mit ein paar Spritzern Würth Fettlöser, Wasserhahn und Pinsel

Reicht bei 90% der Sachen für ein glänzendes Ergebnis…. Haben aber hier auch kaum Kalk im Wasser, ergo wenig „Wasserflecken“

- mit Kompressor und Ausblaspistole gründlich trocken pusten (vermeidet auch Wasserflecken), auch unter Bauteilen und in Erweiterung Slots. Aufpassen bei Aufklebern…
-dann noch Warmluft, entweder Badlüfter, Heizgebläse, Heizkörper, Werkstattheizung, Infrarot Strahler oder ein paar Stunden in die Sonne legen…(ohne Blütenstaub…)

Damit habe ich in der letzten Zeit die besten Erfahrungen gemacht. Isoprophanol oder Elektronikwäsche Spülungen gehen bei 20 Mainboards nicht nur irgendwann ins Geld (selbst wann man Isoprophanol im 10liter Gebinde kauft) sondern „laugt“ auch oft Flussmittelreste/Lötzinn aus und dann hat man so grau weiße Schleier auf der Platine und an jeglichen Kontakten.

Für spezielle Fälle nehm ich auch mal Vorwaschpray oder Badreiniger mit Wasser. Aber auch da aufpassen. Reagieren teilweise schnell mit Metallen (teilweise gewollt oder auch nicht) und waschen auch je nach Alter und Lötverfahren, das Lötzinn aus. Dann muss aufwendig teilweise mehrfach mit Fettlöser nachgewachsen werden.



So sieht der eigentlich aus, ohne ausgeblichenes Etikett.


Ach ja, Kontaktspray bitte nicht zum reinigen verwenden, das ergibt oft einen Film auf den Platinen, bindet Staub und läßt sich schön schwer wieder entfernen…

 

[R-998]

Grüße an alle, sagen Sie mir bitte, wie Sie so hohe Prozessorbusfrequenzen erreichen, welche Modifikationen sind erforderlich, um einen Bus von mindestens 240 MHz zu erreichen? Mein Barton 2500+-Prozessor weigert sich, mit einem Bus über 217 MHz zu arbeiten, unabhängig von seiner Frequenz, die Spannung wird auch auf 1,95 V erhöht (die Platine schaltet sich unter Last oben aus). Bitte sagen Sie mir, was getan werden kann, um einen höheren Bus zu erreichen. Mein Mobo ist Ep8RDA3I. Sorry for my Google Translate

 

[WMDK]

Bitte sagen Sie mir, was getan werden kann, um einen höheren Bus zu erreichen.

That is very little information for your rather broad question, so let's try to set a stage for this...

The maximum achievable bus frequency depends on several factors. Obviously the first thing is the chipset which comes in different qualities depending on where on the wafer the chip was located during production. Same goes for the CPU, they also come in wide variations of quality depending on the wafer location which can be identified by the stepping. Since it's a Barton core, there are four major stepping categories:

I = best quality, high potential
A = second best quality, average to high potential
K = mediocre quality, at best average, usually below that
P = worst quality, it runs at what it is spec'd for, don't expect more

Have a look at the last line on the black label on your CPU and tell us what it reads. It should be something in the format of "AQYHA 0349 XPMW" for example. Really important is also the production date, which in the previous example is "0349", so the 49th week of 2003. It's important because it tells is wether your CPU is superlocked or the multiplier can be changed which is really helpful for further testing because as it is right now, you have three problems at once:

1.) You don't know what speed your chipset can run at
2.) You don't know what speed your CPU can run at
3.) You don't know what speed your RAM can run at

So when you say your system doesn't work at speeds above FSB217 you don't really know what causes this problem because it could be any of the three listed or a mix of things.

At 217MHz FSB, your CPU (if it's locked at multiplier 11) it would run with 217x11 = 2387 MHz. Now, just because you are throwing 1.95 volts at it (which is rather high already) doesn't mean it has to be stable at that speed because as I said, there are different qualities. Each CPU at some point runs into a "wall" where it just won't scale up in speed anymore with more voltage applied or just won't be stable anymore no matter how high you set the voltage. I've had CPUs of really bad quality which didn't even make 2.3GHz stable at a reasonable voltage and then again others which do 2.5GHz at default voltage of 1.65V easily.

My point is, if you don't know you CPUs limit, you can't cross out the CPU in testing and if your system fails at 217MHz it could be the CPU, it could be the board or it could be the RAM.

Speaking of which... What RAM are you using and what speed is it rated for at the given timings? Like do you have BH5, TCCD, CE5, D43...? If you are overclocking be sure to check your memory speed dividers. Ideally they should be 1:1 because only then you really get a performance boost out of it but it also means that your ram needs to be tested to run at that speed. 217MHz FSB would result in DDR433 so you need to check if your RAM can do it and what timings it runs at because you might have to lower them from maybe 2-2-2-5 to 2.5-4-3-7 to allow the RAM to run at higher speeds and / or even increase VDimm if needed.

Let's get back to the CPU once more. So what I would do is I'd check if you can lower the multiplier to like 6 or 8 for example and see if the frequency changes accordingly or if the system simply fails to boot then. If it does work and you can change the multiplier, try to max out the cpu first so you know what clock it can run at. Start at default voltage (1.65V) and set it to 11x200 (2200MHz / 3200+), put it through Prime95 and 3DMark2001 to test it every time you change the clock speed. If it does 11x200 @ 1.65V, change it to 11,5x200 @1.65V (2300MHz) and run the same tests again. You do that for every step until at one point a test fails. You then head back to the BIOS and increase VCore by +0,025V and check again, does it run through @ 1.675V? If not, rinse and repeat with 1.700V, then 1.725V until it's stable again. However, if you realize that even two or three voltage steps of +0,025V don't make any difference and the system still fails, you're facing the beginning of that earlier said "wall" and each increase in frequency will call for even more voltage, making it not worthwhile to proceed beyond a certain speed. If it's a crooked CPU with a locked multiplier and it just doesn't like high speeds at all, there's nothing you could do now other than buy a new CPU with a better stepping. If it can't make high speeds, you won't get high bus speeds because they are tied together with locked CPUs.

Anyway, if you then know that your CPU for example does 2.5GHz @ 1.775V stable you now know, that when the system fails with FSB217, it's most likely not the CPU which is causing an issue here. Of course there are CPUs which don't like high FSBs but 217 is nothing I'd consider high by any means. That narrows our search down to the two remaining items, so let's talk about RAM again.

Let's say your system fails at 217MHz and the RAM runs with the 1:1 FSBRAM ratio with timings of 2,5-3-3-7. First thing to do now is try to relax the timings as much as possible, so for example set them to 3-4-4-11 and try again. If your system boots now, you have an indicator that tells you that you either need more voltage on the RAM to run them at 2,5-3-3-7 with anything above FSB217 or if you don't want to increase the voltage, you need to relax the timings. Anyway it tells you that the problem would most likely have been the RAM if you can cross out the CPU already which is tested to 2.5GHz in this example. If relaxing the timings and increasing VDimm don't give you a single MHz more bus frequency, we should look at the mainboard now.

There are two main reasons for a board not overclocking as you'd like it to.
1.) The BIOS limits the overclocking by having too tight internal timings on the memory controller and all that communication going on
2.) The board design isn't made for massive overclocking / caps are degraded
...or of course... a combination of both.

If it's only the first, there's hope. Our regarded users @digitalbath and @Tzk do have a brilliant understanding of how to modify BIOSes and have shown great competence in creating special BIOS versions adapted to all the testing and investigating the NF2 chipset quirks over the last 2-3 years. They actually made such big leaps in BIOS modding that now almost every NF2 board can be brought to FSB250+ – even the crappy ones you would have never expected to be able to achieve such speeds.

So, if your CPU does higher speeds and your RAM doesn't cause any trouble as well, it might just be the BIOS that is not optimized for overclocking thus limiting your maximum achievable bus speed. But always remember, make sure to cross out all other components before that or you simply can't tell what is causing issues here and you'd be fishing in muddy waters. Also if everything else checks out try to increase the chipset voltage a bit if your BIOS offers that option, like +0,1V and see if that changes anything.

Last but not least, there are two more things that could limit your success.

1.) Degenerated capacitors
2.) Overall insufficient design for overclocking

So first of all, take a good look all your capacitors on the mainboard and check for any bulging or leaning massively to one side, pushed out rubber bottoms, obvious leaks underneath or on top and everything that visually counts towards a possible defect. Why I'm saying that is that many old mainboards, ESPECIALLY Epox mainboards usually do have several bad capacitors on it either bulging heavily or already spitting electrolyte. In either case those caps then don't do what they're supposed to do anymore which is filtering, buffering and flattening power supplied to specific components which can cause stability issues or even burned out hardware when a catastrophic failure occurs. Since overclocking introduces – spoken in a simple way – more noise / stress to all the affected capacitors, that might just push them over the edge to either total failure or unsatisfying behavior in general, causing aforementioned issues like instability and the like.

One other thing you have to keep in mind is how your board is designed. With the 8RDA3I they clearly didn't have the enthusiasts on their radar since afaik this board only comes with a two phase VRM layout. That is fine for the average midrange board not for being heavily beaten for the last MHz because when it really counts, you'd want at least a three phase design for better and more stable power supply for the CPU. Since I've read that you applied 1.95V to the CPU and given the fact that this is not an enthusiast mainboard, you might have already pushed it into an overcurrent protection or some other intended failsafe mechanism. Speaking of which, how did you cool your CPU while it was seeing 1.95V? Please try to start low as already described above because 11x217 is by no means a frequency that justifies such a high VCore.

As you can see there are quite a few things to check and cover and we need a lot more information from your side to specifically address certain issues at hand because this vague question of yours can only be vaguely answered like I did with considering a few of the most important factors but nothing in great detail until we know exactly what we're looking at, what your approach so far was and what hardware specifically you're using as well as the condition it is in.

Hopefully that got you an idea on how to proceed and provide additional details for further assessment or help by other users.

 

[Tzk]

Very thorough explanation, @WMDK.

I'd like to expand on the factor limiting the board, which is the chipset, the bios and the chipset timings. As long as a stock bios is used this heavily depends on the default fsb of the cpu. The bios loads different chipset timings ("romsips") for 100, 133, 166 and 200mhz fsb cpus. Usually cpus with lower default fsb will hit a hard fsb wall, sometimes as low as 193Mhz (for 100/133). Theres two solutions to this: a) trick the bios into thinking that the cpu has a higher default fsb and b) using a modded bios which already has this trick built in. a) is called the L12 mod or wire trick while b) is called soft L12 mod. I usually suggest going for b), as it is easily reversable and does only depend on the bios the board is currently running on. Bonus: modded bioses got different chipset timings which will allow for a higher fsb, if all parts are capable.

If you tell us which board you're running, we can suggest a mod bios or even create one specifically for your board and needs

 

[digitalbath]

@WMDK
well explained and worth to read!

If you tell us which board you're running, we can suggest a mod bios or even create one specifically for your board and needs

he wrote that he use an Epox 8RDA3I.

As far as I know, there are two versions of 8RDA3I boards. A PRO version and a non PRO version. Both boards require separate BIOSes. You can find them here.
I've modded a BIOS version for the non pro Version of 8RDA3I in the past, but I've never modded a version for the pro version.

 

[R-998]

I = best quality, high potential
A = second best quality, average to high potential
K = mediocre quality, at best average, usually below that
P = worst quality, it runs at what it is spec'd for, don't expect more

Have a look at the last line on the black label on your CPU and tell us what it reads. It should be something in the format of "AQYHA 0349 XPMW" for example. Really important is also the production date, which in the previous example is "0349", so the 49th week of 2003. It's important because it tells is wether your CPU is superlocked or the multiplier can be changed which is really helpful for further testing because as it is right now, you have three problems at once:

1.) You don't know what speed your chipset can run at
2.) You don't know what speed your CPU can run at
3.) You don't know what speed your RAM can run at

Thank you for the detailed answer, my processor, judging by the CPU Z, has index A and should overclock well, and this is so, at a frequency of 2.2 GHz it works without increasing voltage (SuperPi 32M passes), I already flashed the BIOS from Digitalbath (find in biosbude.de) but I don’t know what means romsip and alpha timings and what positions should they be set to, my RAM is Hynix DT-D43 and it is able to work like DDR540 on an A64 system, it is not a bottleneck, it works in 1:1 mode, my processor is blocked and cannot change the multiplier (4th week of 2004), with bus 217, the processor also remains stable and passes 3dmark2001 and SuperPi, when overclocked further, it simply does not load the system, when overclocked from under Windows using Clockgen it also becomes unstable. My Ep8RDA3I Is not PRO. I use Zalman 9500LED , cooling is not a problem. Thanks.

 

[funkflix]

As far as i remember you can't see the I/A/K/P etc. in CPU-Z, you need to dismantle your Cooler to see the full Batch.

Maybe some other Clocks are coupled to the FSB? AGP?

 

[R-998]

As far as i remember you can't see the I/A/K/P etc. in CPU-Z, you need to dismantle your Cooler to see the full Batch.

Maybe some other Clocks are coupled to the FSB? AGP?

AXDA2500DKV4D
AQYHA 0414VPMV. My videocard is MX440 64M and it works fine with 89mhz AGP (Tested in Slot-1 mobo)

 

[funkflix]

Until the Pros will reply again here... Have you ever tried different Slots for the RAM and/or just Singlechannel? S462 and S939 can act totally different with some kind of ICs.

 

[R-998]

tried different Slots for the RAM and/or just Singlechannel

I tried singlechannel , but dualchannel gives more performance and a single channel does not make sense, and I did not notice a noticeable increase in the bus frequency

 

[funkflix]

You just need to figure out the thing stops you at 217MHz, that is why I told you to test different setups. Sure, DC is what you want.

 

[digitalbath]

AGP / PCI clock shouldn't be the issue.

@R-998
FSB frequency of 217 results (multi of 11) to 2387MHz. So you're sure, that the CPU is not at its limit?

Which BIOS did you choose? I would start with the 619XT BIOS. 619XT is the name of the integrated romsips. Romsips are internal chipset timings, that Tzk mentioned in his post before. The romsips tables are essential if you want to overclock via FSB bus. The romsips from official BIOSes usually limits the FSB bus OC to ~220MHz. That’s why mod BIOSes with modded romsips tables (DFI origin) helps to overclock.

In my mod BIOSes, I’ve included an option for selecting a „romsip preset“. The BIOS will change the included chipset timings after start to the selected timings. There are no general chipset timings that will work on every board. Most boards needs different timings to perform. So, I would start to try every romsip preset in the BIOS. One or some presets will probably work good, most average and some really bad.

For Hynix DT-D43 Chips I would select these timings:
tCL:2,5
tRCD:3
tRP:3
tRAS:8
tRC:11-13
tRFC:14-16
alpha timings: 3-4-5-4-3-4-5 preset.
drive strength : 3-5
slew rate : 7-9
tREF: 25xx - 36xx

 

[R-998]

FSB frequency of 217 results (multi of 11) to 2387MHz. So you're sure, that the CPU is not at its limit?

I choose 619XT. Well, this frequency seems to me too low to stop, especially since "A processors" are capable of much higher performance, judging by the HWBOT. I will try the timings you suggested, now the memory works like 2.5-3-3-5 and it works fine (2 circles of Memtest86 without errors), is it worth looking for romsips just by trying everything in turn? What is the best way to do this, increase the bus frequency from BIOS or from under Windows using Clockgen?

 

[digitalbath]

Well, this frequency seems to me too low to stop, especially since "A processors" are capable of much higher performance, judging by the HWBOT.

the "average" Barton CPU, as I see, will perform something like this:
1,65V / 2200MHz
1,85V / 2400MHz
I have worse Bartons then this, though.

I will try the timings you suggested, now the memory works like 2.5-3-3-5 and it works fine

wow, low tRAS timings. I only use tRAS=5 for low latencies like 2-2-2-5. For CL2,5 timigs I use most tRAS=7/8

is it worth looking for romsips just by trying everything in turn?

yes, just select a preset and test it.

What is the best way to do this, increase the bus frequency from BIOS or from under Windows using Clockgen?

I clock in most of the time in windows. If it crashes, it will not corrupt the OS.
There is a special tweaker for Nforce2 from Infrared. I would use it. If you don't have it I could look for a download link some time later.

 

[R-998]

the "average" Barton CPU, as I see, will perform something like this:
1,65V / 2200MHz
1,85V / 2400MHz
I have worse Bartons then this, though.

So it is, straight 1 in 1 ( hwbot rec ), it's the same processor, only a different board (I killed it when trying to get more with a volt mod, the bios probably flew because it always starts with a different postcode), while I used much less cooling and it coped (2.4 is not stable, 2.38 is stable). Yes, please look for this tweaker, I don't have it

 

[Tzk]

Yes, please look for this tweaker, I don't have it

Releases · irusanov/nForce2-xtreme-tweaker-vcl

Contribute to irusanov/nForce2-xtreme-tweaker-vcl development by creating an account on GitHub.

github.com

 

[R-998]

Now I tried to turn everything on and overclock on this epox (before that, I did everything with this CPU on the Msi K7N2 delta l) after it failed, I plugged it into Epox, assembled it, turned it on, the system comes to loading windows and instantly reboots, with any BIOS settings this happens, does this mean that along with that motherboard I damaged the processor as well?

 

[digitalbath]

Now I tried to turn everything on and overclock on this epox (before that, I did everything with this CPU on the Msi K7N2 delta l) after it failed, I plugged it into Epox, assembled it, turned it on, the system comes to loading windows and instantly reboots, with any BIOS settings this happens, does this mean that along with that motherboard I damaged the processor as well?

I don't think so.
APIC setting?
Did you try other FSB speed? something like FSB 100?
CPU Interface disable?
clear CMOS?

probably this BIOS Version doesn't work for this board.

 

[R-998]

APIC setting?

tried turning it off, didn't help , other fsb are tried too , cpu interface also doesn't affect anything. 2.05v can damage Barton CPU without load? With Thorton 2000+ board works fine

 

[Tzk]

Does the epox feed the cpu off 12v? And if no, does your psu have a strong 5V rail?

Edit:
Nevermind, should be 12V. It has a 4pin connector.

Edit2:
Sometimes NF2 systems just become unstable after they locked up or froze. Then it helps to force it into booting fail-safe (100x6). Usually you can do this by booting for ex. Without ram and then again with ram.

And regarding your Hynix. Are those 256mb sticks? The nf2 does like those a lot while it absolutely hates other sizes. Hynix in general are a good choice for the nf2, especially the mentioned 256mb sticks as bt-d43 or dt-d43.

 

[R-998]

And regarding your Hynix. Are those 256mb sticks? The nf2 does like those a lot while it absolutely hates other sizes. Hynix in general are a good choice for the nf2, especially the mentioned 256mb sticks as bt-d43 or dt-d43.

Yes , 256mb per 1 , now I’m more interested in whether my processor is damaged, and it seems that it’s not, or not completely, since I plugged it into the board on the KT266A (i modify bios to attach barton microcodes) and it works at a lower frequency, but it works fine, despite the fact that on Epox at the same frequency the system cannot boot any of my installed windows. screen of work

 

[Tzk]

Forgot to share my thoughts on this… my guess is that Barton’s degrade quite fast at or above 2V. I usually stick to 1.85V on my best CPUs and 1.95V on the ones I don’t care about much.

I got some Barton’s which max out at 2.4ghz 1.85V and even a a7n8x board which hits a Fsb wall at 220mhz.

Also Bad caps max cause issues. Note that a bad cap sometimes doesn’t bulge, but just fails without any visible issue.

 

[R-998]

Forgot to share my thoughts on this… my guess is that Barton’s degrade quite fast at or above 2V. I usually stick to 1.85V on my best CPUs and 1.95V on the ones I don’t care about much.

My MSI K7N2 Delta L did not work even for 5 minutes at 2.05v, I think the processor could not be damaged or degraded in such a short amount of time. That is, do you think that this is the focus of my processor and not the motherboard and its bus frequency? Is there any other way to increase the frequency in this case? And how do they even take records on this stone at 3 GHz and higher if it is afraid of overvoltage and the boards for it die so easily?

 

[Tzk]

For 3ghz+ you need two things:
1. A really good cpu which does at least 2.5ghz at 1.65V or less
2. cold. You won’t reach this on ambient cooling. A single stage, dice or ln2 is needed for more than validation shots.

I got a cpu which is quite good and does 2.7ghz 24/7 stable on watercooling at about 1.85V. However it can’t reach 2.8ghz at reasonable volts. These CPUs have become rare though.

What the exact issue with the epox is… no clue. You‘d have to test this with another cpu (preferably unlocked) to be able to test for higher Fsb at low cpu clocks. And maybe with other ram sticks or in single channel, just to rule them out.

Oh and the nf2 got agp and pci lock, so that isn’t an issue.

 

[R-998]

These CPUs have become rare though.

Are these successful samples or Barton processors in general? It's just that ordinary Barton processors cost us, for example, less than 4 bucks apiece, but how do you even distinguish successful from unsuccessful, except for the A or K index?

 

[Tzk]

Golden Samples, so not all Barton’s will do this. Usually the best ones are IQYHA or IDYHA steppings, preferably XP-m 2600+ FQQ4C (1.45V vcore), because those are unlocked. Golden ones are for example from weeks 0351, 0352 or 0401. mine is a 0352 mpmw.

However you can buy these for like 20-40 bucks a piece instead of 5-10 for a random Barton. If you manage to find one of course