[Sammelthread] Sockel 939 / 940 / 754 - Thread

[Neo52]

gingen die FX nicht bis 64??

 

[MusicIsMyLife]

gingen die FX nicht bis 64??

Sockel 940 war glaube ich maximal FX-53, bei Sockel 939 war es der FX-60, bei Sockel AM2 der FX-62 und bei "QuadFather" mit Sockel 1207 war es der FX-74.

 

[stunned_guy]

Sockel 940: FX-51, FX-53
Sockel 939 : FX-53, FX-55, FX-57, FX-60
Sockel AM2: FX-62
Sockel 1207: FX-72, FX-74

Sockel 939 hatte die größte Vielfalt, auch bei den Kernen: Clawhammer, San Diego, Toledo

 

[Neo52]

@stunned_guy: schon alle da? 🤪

EDIT: Hab das gerade mit dem X2-6400 verwechselt

 

[stunned_guy]

Ja, bis auf die Sockel 1207, da habe ich kein Interesse dran.

 

[Sparksnl]

You gotta be kidding… you bought it for 25 Bucks?

Yes indeed. Let’s see what it can do when I get it. Somebody did not do his or her homework.

 

[maxpower1984]

Sockel 1207: FX-72, FX-74

den FX70 gabs wohl auch noch

 

[stunned_guy]

Der ist so alt, den hatte ich gar nicht mehr auf dem Schirm.

 

[Sparksnl]

While looking for better timings I just noticed I got my TCCD dimm running at 394MHz with only 3.0v (for the C3 max MHz compo). If I remember correctly this dimm does not scale on voltage that much, but I can try...

Edit: Which Ultra D bios is good for TCCD and had a working setting for running dor example FSB 200MHz and mem 250MHz?
I noticed LD329 and 623 don't have a working version.

Edit 2: The FX-57 was in today. It was completely unstable at stock settings, even when raising voltage a bit. Lucky... a BIOS reset seems to have solved it.

 

[Sparksnl]

I tested BIOS 704-BTA for max MHz with C3. This BIOS seems to be working well with the 05/04 memory/fsb divider.
Too bad I still can not go beyond 394/395MHz.

 

[Tzk]

Have you tried setting drive strength (both of them), DQS Skew and the asnyc setting? Also byte granularity? Maybe there's still a bit of headroom left?

 

[Sparksnl]

I tried drive strenght. There is actually only 1 setting possible, others crash.
I tried both DQS settings as well, this did not help.
I can still test DRAM Data Drive Strength.

 

[Sparksnl]

I think the main problem is that this dimm runs 394MHz @2.9-3.2v. But does not scale above 2.9/3.0v.

 

[Sparksnl]

After swapping CPU I was able to find 12 TCCD dimms that boot 390MHz C3 from which 2 are 256MHz, 10 are 512MB versions.
So my 512MB's were limited by the A64 memory controller.
I was now able to run PI 1M with 2x 256MB @c3 and 397MHz (reset at printscreen).

 

[Sparksnl]

Bump to place 3 for DDR1 memory: https://hwbot.org/competition/HWBOT TEAM CUP 2021 - MEMORY/stage/5285_memory_frequency_-_ddr1
https://hwbot.org/submission/4812674_sparks.nl_memory_frequency_ddr_sd_ram_414.1_mhz

I can and will go higher. No more time today. Still binning step by step.

 

[Sparksnl]

Corsair CMX512 3200XL @3.4v running 419.6MHz
https://hwbot.org/submission/4813876_sparks.nl_memory_frequency_ddr_sd_ram_419.6_mhz

Now to look for a better memory controller since I have more then 1 dimm which does this exact speed (2 and maybe 3).

 

[Sparksnl]

I have been binning all of my A64's for max C3 runs (41 cpu's). After finding that 1 cpu which is capable of doing 419.6MHz for C3 and 512MB dimms I now have 4 more.
Please note that not all cpu's were capable of doing C3 max frequency with 512MB dimms but did work at 400+MHz with 256MB dimms. So it is important to bin with good 512MB dimms.
So now for more runs with my best 2 dimms. (later...).

 

[Sparksnl]

Final score with the hardware I have at this moment: Corsair CMX512-3200XL @422.3MHz 3.5v
HWBOT
Validation

I used my FX-55 San Diego. A nice note on this cpu is that when it came, it had almost all pins bent. When fixing it 1 pin broke but it seems to be running just fine.
Nice to see it was worth the effort.

 

[The Sandman]

Sick clocks, as always

 

[Sparksnl]

Please note that this tccd runs 7.1ns on this speed. BH5 needs to run 281MHz c2 for this.

 

[Tzk]

If only tccd would survive this abuse at 3.5V just as bh5 does... still a huge achievement. Well done.

Is this v1.1 XL or v1.2?
Edit: saw the Hwbot photo. It is 0528 v1.2.

Also interesting that this E4 San Diego does so well at high clocks.

 

[Sparksnl]

Indeed it is XL v1.2. I would run it for hours if possible. This specific dimm was less good at 3.4v but it scales really well from 3.4v to 3.5v
I also noticed that all 3 good dimms did not scale to 3.6v.

Actually I have more San Diego’s which do good. They seem to be good for this.

 

[The Sandman]

Wouldn't be surprised if they lost their CL2.5 ability after the voltage torture.
TCCD is degrading ultra fast. Can you do a 32M test run or maybe even Goldmemory after you are done with em?

 

[Sparksnl]

I sure can. Just c2.5 or c2 as well?
I checked 200MHz C2 as booting afterwards.

 

[Sparksnl]

Wouldn't be surprised if they lost their CL2.5 ability after the voltage torture.
TCCD is degrading ultra fast. Can you do a 32M test run or maybe even Goldmemory after you are done with em?

I quick tested the dimm which I used most (binning and best run). This dimm can still boot and run c2.5-4-4-8 1T at 2.9v and PI 4M. It can also still boot 320MHz C2.5 @2.9v. (I knew it never did more then 305-310MHz PI 1M): 0528

I never tested my second best and most used dimm for c2.5. But it boots c2.5 @2.9v and 320MHz just fine: 0518
I also tested my third best dimm, I never maxed it out for c2,5. This dimm also boots @2.9v and 320MHz c2.5. Besides that it runs PI 4M as well: 0538

Just to be sure I also tested my 256MB dimm which only scaled up to 3.2v for c3.I never tested it for C2.5 as well. It still boots c2-2-2-5 1t @ 222MHz (and c3). However, C2.5 does not seem to boot at 200Mhz: 0451

I checked my favorite tccd dc kit which I use for socket A as well. I could boot it at 320MHz. That is a little better then I expected: 2x 0538

 

[The Sandman]

Lucky you. My OCZ can't do CL2.5 anymore. They lock up during boot or 32M.
But I was able to complete Memtest at 300-MHz on the Abit AV8. Goldmem found many errors at 200-MHz CL2.5. Everything was fine at CL2.

 

[Sparksnl]

It is just the 256MB dimm for now. Could it be that the better scaling TCCD ic's can also take more voltage for longer periods?

 

[The Sandman]

We don't know. I'm just spreading the warning. OCZ claimed on XS that they got TCCD Chips directly from Samsung which were already unable to run CL2.5 and did behave like Winbond aka voltage lovers.

For further chit chat ram talks please lets move over to the DDR thread

 

[Tzk]

@Sparksnl
Could you add the chip weeks above, of you know them? Maybe there‘s a correlation between scaling, voltage and dying at high volts?!

 

[Sparksnl]

@Sparksnl
Could you add the chip weeks above, of you know them? Maybe there‘s a correlation between scaling, voltage and dying at high volts?!

Done

I also noticed that after my CB2003 run which I did yesterday (Opteron 180) it is now dead. So now from 3.3GHz cpu back to 3.2GHz Opteron 180 (I had a backup cpu).