[Sammelthread] Sockel 939 / 940 / 754 - Thread

[Sparksnl]

New in. A very late batch. Strange that weeks (10 to 15?) after San Diego FX-55 came, these were still being made.
Let’s put it to the test if low batch numbers are better.

 

[stunned_guy]

Sehr spätes Stepping. Ich hatte eine ähnliche Woche am Start, die lief aber sehr bescheiden und nicht annähernd so gut wie ältere Clawhammer. Versuche mal dein Glück

 

[Sparksnl]

Well my 0512 cpu scaled not well as well (on cold). I just need to know, if the cpu which was completely in the middle when it was born and does not run well I have a good idea which periods are best.

 

[Sparksnl]

I ran my FX-53 (CAAGC 0426TPMW) on cold today. No coldbug at all with 1.84v (1.824v) @-190C:
HWBot
Validation

 

[stunned_guy]

Ist ja auch ein "Coldhammer" ähmm Clawhammer
Nice !

 

[Sparksnl]

Yes indeed, a Coldhammer 🥶

Edit 22:53: I found something while running this cpu. While not being able to boot a higher multi then 17x to XP, I was able to set 17.5x and 18.0x in XP.
This could be a great help with limited FSB speeds (<220MHz at full pot). I noticed that the best FX-55 records used FSB to up speeds. This does not seem to be the best way to run them.

 

[Sparksnl]

Sehr spätes Stepping. Ich hatte eine ähnliche Woche am Start, die lief aber sehr bescheiden und nicht annähernd so gut wie ältere Clawhammer. Versuche mal dein Glück

Cold bug post with 1.64v - 1.88v at-147C - 160C. It sometimes posts at -160C but never when it coldbugs at -180C to -190C. It needs to go back to -147C then.
I needed to boot to XP at -160C and cool down to -190C afterwards.
Max speed was 3717MHz @ -160C / 1.81v. -190C ran 3616MHz as a max. So I stopped my session (was not going to be better as my other one, uche: below: ).

 

[Sparksnl]

BANG!!! --> FX-55 Clawhammer @4126.93MHz

https://hwbot.org/submission/5529625

New number 1 score!
😁 😁 😁 😁 😁 😁 😁 😁 😁 😁 😁 😁 😁

Photo of my 4118MHz score.

I had to end the session because of moist. I still had 1.5L LN2 left and wanted to run PI 1M.

 

[stunned_guy]

Mit welchem Multiplikator hast du gebootet ? Ich nehme mal irgendwas zwischen 18 und 19,5 an oder ?
Das ist rekordverdächtig, sehr schön, ich freue mich für dich

 

[Sparksnl]

I booted with 17x, why push things. I already knew it did 3.9 and I knew there was at least 100MHz more in it (my last session was sort of ruined).
It took me a while to find best voltage and speed to up by 1 and the order to set fsb and multi.

 

[Sparksnl]

I did some testing with my Opteron 254, 256 and 154. The idea was to find the cpu which could run PI 1M at max MHz at a CPUz voltage of 1.424-1.432v. The best CPU should also have an as high as possible Max FSB on air at 5x CPU multi.

Typical is that my one and only Opteron 154 came out as best. I tested that 154 because I know what it can do on cold.

I can now safely say that:
- Opteron x56 is a bin of the better x54.
- Opteron x54 will never do 4GHz with B1.
- Week 0611 is not the absolute best bin.
- I am very lucky with choosing my only 154. Let’s just say I chose wisely…

The reason I test max fsb is for cold capability.

Ps: The good news is that I will be selling some 254’s for cheap.

 

[Raiden Zero]

0617FPMW and 0629FPMW look promising. X15 might just be enough for them to hit 4G at like -60c. My best 254 did around 3.4GHz benchable on water (1.5-1.55 iirc) but was BCLK limited at 285 on single stage.

 

[Sparksnl]

I think that my Opteron 256 1462241E60231 is my best cpu. It is almost on par with my 154 and it indeed has x15 multi.
It did PI 1M with 1.5v (1.512-1.520v) at 3315MHz. My best cpu did 3349MHz, but that is too close compared to it’s sub zero potential (max fsb of 400 on air).
Those 0629 scaled to 3326 and 3349 (14931G60036).

3.4 benchable is really good on air. So it needs ln2 or dice.

 

[Der Namenlose]

Mein angekündigter FX-55 kam am Donnerstag an.
Zustand nicht gut, aber bei weitem nicht so schlimm wie befürchtet:



Der fehlende Pin ist lt. Pinout wohl ein NC_AL, wenn ich das richtig verstehe sind NC Pins nicht in aktiver Verwendung, das würde ja Hoffnung machen.
Sollte der Pin doch nötig sein liegt er ja immerhin außen am Rand, da steigen die Chancen auf Reparatur.

 

[bschicht86]

Also was ich da schon gesehen habe, schaut der doch fast Fabrikneu aus.

 

[Der Namenlose]

Hast ja recht, fehlt nur ein Pin, keine Wollmäuse, Wärmeleitpastenklümpchen oder gar andere undefinierbare Substanzen zwischen den Pins - quasi NOS

 

[Sparksnl]

Too bad it is delivered lime this. When you pay an amount like this you expect it to be in a good state.

 

[Sparksnl]

Hi All, I was just wondering if anyone of you needs an Ok air only San Diego 4000+.
I have a KACAE 0544GPAW batch which can go but does not need to.
I have 2 better air ones (3.3-3.4 range) and one which is golden. So it won’t be used here.

I place it here because it does not need to go and if it goes I want it to go to you guys.

 

[Sparksnl]

New in: AMD Athlon 64 (X2) ES pre production 4400+

It came from Germany. From any of you?

 

[Der Namenlose]

Of course - and there I was wondering who'd buy a 4400+ so quickly at that price

Edit:
I don't think it came from anyone in the forum, the listing was from a commercial seller.

 

[Sparksnl]

Well this must be one of the first consumer multicore cpu’s in week and year. So when waiting long enough thisone is for museums.
And €28 was not extremely high as well. Not for a cpu this early.

Edit: I actually waited a day or so.

 

[Der Namenlose]

Maybe I missed how long it was online already when I saw it, not as much of a regular on ebay as I used to be .
I did put it on the watchlist and wanted to link to it here, but by then it was gone, which I really didn't expect.

Curious to see what it will do for you - somehow I wouldn't expect those early Dual Cores to clock very well.
In terms of the early X2 CPUs without that X2 designation in the top row, I had an X2 4400+ and my only X2 4800+ (0538) also doesn't have it.

 

[Sparksnl]

I bought it as a collectors item. I will test it for benching but it will not run fast for OC.

Edit: I quick tested it and it is not that bad.

 

[Der Namenlose]

Heute mal versucht den FX-55 zu richten, ein zweiter Pin sitzt arg locker, der macht bei nächster Gelegenheit sicher auch einen Abflug.
Jetzt geht das gute Stück immerhin mit etwas liebevollem Druck in den Sockel und macht auf dem günstigen Testboard, was er soll.



Nächster Schritt dann DFI und gucken, was so geht.
Aber das steht erstmal recht weit hinten auf der Prioritätenliste ...

Edit:
Jetzt mal das seit Ewigkeiten mit Venus Mod hier liegende NF4 Expert entstaubt, ein wenig RAM muss weg (keine Sorge, nichts spannendes )
und ich will ja auch mal testen, was die neuen TCCD so leisten - DFI hat mich freundlich dran erinnert, dass das mit Plug&Play hier nix wird .

90min und etwas Trial and Error später läuft die Kiste - eventuell frisst es mit Preset dann auch Samsung.
Muss sowieso mal gucken, welches BIOS da drauf ist, ggf. gibt's da Änderungsbedarf ^^ .

 

[Der Namenlose]

Beim Ausbauen des FX-55 hat sich nun wie erwartet auch der zweite Pin verabschiedet, sollte aber ebenfalls ein nicht belegter sein, wenn ich mich recht erinnere
Immerhin hab ich bei dem Kollegen dann keine Scham auch mal unter Trockeneis (oder so ich es mal hinbekomme LN2) etwas Feuer zu geben - hoffentlich hat er auch Potenzial .

@Sparksnl since (at least by perception) you're working a lot with the NF4 Expert - do you use it with TCCD or mostly with Winbond?

Did some testing yesterday and mine absolutely did not like any TCCD modules, after spending some time with google this doesn't seem to be unusual.
I switched my BIOS to the E704-2BT (BigToe Modbios for the Expert), which certainly did help, got to 300 2.5-4-4-8 1M, but 32M stability still is an issue.
Apparently there were Merlin Modbioses for the Expert as well, but I did not find any working mirrors, didn't find any on BiosBude either.

I will cross check with other modules which I already tested on my Ultra-D and maybe another CPU (currently using an Opteron 154), but figured it might be worth asking for insights here.
My NF4 Ultra-D certainly is less picky with Samsung sticks and I have most settings dialed in to where I get solid results, maybe I'll switch back to that for testing the new DIMMs from redux.

 

[Sparksnl]

You should put it on cold. Try a -90C to -110C range. I expect it to do well in that range.

I use AT6 mainly, 2x256MB for finding the proper speed etc. After that I swap to 2x512MB BH5. I should try and test TCCD on high speeds for efficiency against BH5.
I use the NF4ED06 BIOS. It has less issues then E704-2BT. I noticed it missed some multipliers and caused other issues.

I would think that the Ultra D could at least be just as good. But it has so many bios choices and the Dr. Expert works fine with that bios.

 

[Der Namenlose]

Thanks for your input, might try the NF4ED06 as well then. Would be interesting to see what your experience with TCCD is on the Expert.
My goal currently is to max out the TCCD, mostly at CL2.5 - performance for 2D benchmarks and Winbond binning is secondary for now.

I'll do some more testing in the coming days to see if I'll switch back to the Ultra-D for Samsung binning or try to figure things out with the Expert.

 

[Sparksnl]

If needed I can run my 256’s at 280-290C2 and my 512’s at 250-260 C2 (on air). So I did not take any effort to run C2.5 yet.
I did notice that some IMC ‘s seem to perform worse on cold with 512’s.

Some more tips.
- When going on cold at first keep cpu/pll multi at 2.0x or 2.5x. That is best for CB boots most f the times.
- When going below 0C, set 1.64v or so. That helps with Coldbug post issues on some cpu’s.

 

[Der Namenlose]

For 2D Benchmarks I would probably go with Winbond as well, my best so far run similar to yours on air (260-270 C2 with 512MB sticks).
Running TCCD at CL2.5 for clocks of 300 and more is just a habit and has in the past been the most fun to me in terms of DDR1 testing.
If I was less hesitant to push the voltages too far, I'd try going for 400 CL3, but first I want to see what they all can do at CL2.5 .

Thanks for the tips on running on cold, I do remember the part about the LDT multiplier at 2.0x / 2.5x, but the voltage part is new to me, will definitely keep that in mind.
While I hope to test on Dice sooner rather than later, I doubt I'll get around to that anytime before July - we'll see.

 

[Sparksnl]

That is it probably. I don’t do 3D benching that much. Well and for sure not on older platforms. Only PCI and AGP is interesting for 3D there for me.