MEMORY Voltage Scaling, please contribute...

[Rabbi_NZ]

This thread is dedicated to collecting UTT/VX/NEW-BH5 related data... especially how they scale with voltage.

To contribute please do the following:
* Set timings to 1T-2-2-2 (I run 1T-2-2-2-8-7-15, but that is just for reference)
* Find the max MEMTEST STABLE MHz at 3.1v, 3.2v, 3.3v, 3.4v, 3.5v, 3.6v.
* MEMTEST STABLE (for this thread) = 25+ loops of test #5
* Where possible... also give chip dates & PCB revision.


*********************VX/UTT Memory-Timing Tips*********************
* i865/i875 TWEAKER can be found here
* A64 TWEAKER v0.31 can be found here
* nForce2 Tweaker 0.25b can be found here



DFI nF2 Infinity , Mobile 2600+ , 2x512 TwinMOS UTT 50D (uwackme)


DFI nF3 754 LanPartyUT , Mobile 3400+ , 2x512 OCZ VX (rabbi)


DFI nF3 754 LanPartyUT , 3000+ , 2x512 OCZ VX (aCidbAbY)


DFI nF4 939, 3500+ Winny , 2x512 TwinMOS UTT 04514/05044 (esdee)


Abit i865 AS8, 560J? , 2x512 TwinMOS UTT (Hipro5)

Best is - as for ABIT based mobos - and IF someones mobo/Rams can "handle" that, to put : 2 - 5 - 2 - 2 , 32min , StreetRacer , Enchanced , 4 , Enable , Enable as for 1:1 and 2 - 5 - 2 - 2 , 32min , Auto , 5 , Enable , Enable as for 5:4......and as for 3:2 , 2 - 5 - 2 - 2 , Auto , Encanced or Normal , 6 or 7 , Enable , Enabe or Disable.......

************************************************** *********


********************AMD Socket 462 Results********************

enok (2x512, TwinMOS UTT 50D 0451 KO-6633, Abit nF2 AN7):
3.1v = 240 MHz
3.2v = N/A
3.3v = N/A
3.4v = N/A
3.5v = N/A
3.6v = N/A

TEDY (2x512, TwinMOS UTT 0503 44D KO6633, DFI nF2 Infinity)
3.1v = 235 MHz
3.2v = 240 MHz
3.3v = 245 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

mrlobber(1x512, OCZ VX, DFI nF2 Infinity)
3.1v = N/A
3.2v = 243 MHz
3.3v = 247 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

godsa(1x512, TwinMOS UTT 50D 0503 KO-6633, DFI nF2 LanPartyB):
3.1v = 254 MHz
3.2v = 257 MHz
3.3v = 261 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

EMU(2x512, TwinMOS UTT 50D, DFI nF2 Infinity):
3.0v = 198 MHz
3.1v = 217 MHz
3.2v = 228 MHz
3.3v = 243 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

Fr3ak(2x512, TwinMOS SP UTT, Shuttle nF2 SN45G):
3.1v = N/A
3.2v = 245 MHz
3.3v = N/A
3.4v = N/A
3.5v = N/A
3.6v = N/A

uwackme (1x512, TwinMOS UTT 50D, DFI nF2 Infinity):
3.1v = 240 MHz
3.2v = 245 MHz
3.3v = 245 MHz
3.4v = 245 MHz
3.5v = N/A
3.6v = N/A

uwackme (2x512, TwinMOS SP UTT 50D, DFI nF2 Infinity):
3.1v = 215 MHz
3.2v = 230 MHz
3.3v = N/A
3.4v = N/A
3.5v = 250
3.6v = N/A

3.1v = N/A
3.2v = N/A
3.3v = N/A
3.4v = N/A
3.5v = 245
3.6v = N/A

************************************************** *********


********************AMD Socket 754 Results********************

Rabbi (2x512, OCZ VX, DFI nF3 754):
3.1v = 242 MHz
3.2v = 259 MHz
3.3v = 263 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

HermS (1x256, TwinMOS UTT, DFI nF3 754):
3.1v = N/A
3.2v = N/A
3.3v = 254 MHz
3.4v = N/A
3.5v = 264 MHz
3.6v = N/A

Jupiler (1x512 & 2x512, TwinMOS UTT 44D, DFI nF3 754):
3.1v = 242 MHz @ (1x512) , 230+ MHz @ (2x512)
3.2v = 250 MHz @ (1x512)
3.3v = 253 MHz @ (1x512) , 246 MHz @ (2x512)
3.4v = 255 MHz @ (1x512)
3.5v = N/A
3.6v = N/A

aCidbAbY (2x512, OCZ VX, DFI nF3 754)
3.1v = N/A
3.2v = 250 MHz
3.3v = 256 MHz
3.4v = N/A
3.5v = N/A
3.6v = 260 MHz

3.1v = N/A
3.2v = 255 MHz
3.3v = 257 MHz
3.4v = 260 MHz
3.5v = N/A
3.6v = N/A

v3n (2x512, OCZ VX, DFI nF3 754)
3.1v = N/A
3.2v = N/A
3.3v = 250 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

************************************************** *********


********************AMD Socket 939 Results********************

conrad.maranan (2x512, OCZ VX, MSI Neo2 nF3 939):
2.8v = 220MHz
2.9v = 220 MHz
3.0v = 225 MHz
3.1v = 235 MHz
3.2v = 250 MHz
3.3v = 255 MHz
3.4v = 263 MHz
3.5v = N/A
3.6v = N/A

jikdoc (2x512, OCZ VX, MSI Neo2 nF3 939)
3.1v = N/A
3.2v = 250 MHz
3.3v = 253 MHz
3.4v = 257 MHz
3.5v = 258 MHz
3.6v = 260 MHz

Garrett(2x512, TwinMOS SP UTT 44D, MSI Neo2 nF3 939)
3.1v = N/A
3.2v = N/A
3.3v = 257 MHz
3.4v = N/A
3.5v = N/A
3.6v = N/A

MarilynMX(2x512, OCZ VX, MSI Neo2 nF3 939)
3.1v = N/A
3.2v = N/A
3.3v = ~240 MHz
3.4v = 249 MHz
3.5v = N/A
3.6v = 258 MHz
3.7v = 261 MHz
3.8v = 264 MHz
3.9v = 266 MHz

mrlobber(2x512, OCZ VX, DFI nF4 SLI-D)
3.1v = N/A
3.2v = N/A
3.3v = 258 MHz
3.4v = 262 MHz
3.5v = N/A
3.6v = N/A

bobmanfoo(2x512, OCZ VX, MSI Neo2 nF3 939)
3.1v = N/A
3.2v = 242 MHz
3.3v = 257 MHz
3.4v = 263 MHz
3.5v = 266 MHz
3.6v = 268 MHz

odb (2x512, OCZ VX, MSI Neo2 nF3 939):
3.1v = N/A
3.2v = 245 MHz
3.3v = 252 MHz
3.4v = 255 MHz
3.5v = N/A
3.6v = N/A

arj(2x512, OCZ VX, MSI Neo2 nF3 939):
3.1v = 209 MHz
3.2v = 222 MHz
3.3v = 242 MHz
3.4v = 259 MHz
3.5v = 261 MHz
3.6v = 263 MHz

Wingz(2x512, OCZ VX, MSI Neo2 nF3 939):
3.1v = 238 MHz
3.2v = 241 MHz
3.3v = 255 MHz
3.4v = 265 MHz
3.5v = N/A
3.6v = 275 MHz (1x512mb "good" stick only)

neuvas(2x512, TwinMOS SP UTT 44D ,MSI Neo2 nF3 939):
3.1v = N/A
3.2v = N/A
3.3v = N/A
3.4v = N/A
3.5v = 263 MHz
3.6v = N/A

esdee(2x512, TwinMOS UTT 04514/05044,DFI nF4 939):
3.1V = 255
3.2V = 259
3.3V = 263
3.4V = 267
3.5V = 270
3.6v = N/A

TEDY (1x512, TwinMOS NEW-BH5, DFI nF4 Ultra-D)
3.0v = 240 MHz
3.1v = 245 MHz
3.2v = 250 MHz
3.3v = 253 MHz
3.4v = 257 MHz
3.5v = 260 MHz

3.0v = 240 MHz
3.1v = 246 MHz
3.2v = 251 MHz
3.3v = 254 MHz
3.4v = 258 MHz
3.5v = 262 MHz
3.6v = 265 MHz

************************************************** **********


*************************Intel Results*************************

hipro5(1x512, TwinMOS UTT, Abit i865PE AS8):
3.1v = N/A
3.2V = 239 MHz
3.3V = 259 MHz
3.35 = 261 MHz
3.5V = 264 MHz
3.6V = 268 MHz

3.1v = N/A
3.2V = 217 MHz
3.3V = 239 MHz
3.35V = 252 MHz
3.5V = 255 MHz
3.6V = 257 MHz

bwan(2x512, TwinMOS SP UTT 44D 0503,Abit i865PE AS8):
3.1v = N/A
3.2v = 200 MHz
3.3v = 225 MHz
3.4v = 250 MHz
3.5v = 253 MHz
3.6v = 255 MHz

kindnuguz(2x512, OCZ VX,Abit i875 IC7):
3.1v = N/A
3.2v = 238 MHz
3.3v = N/A
3.4v = 250 MHz
3.5v = N/A
3.6v = N/A

************************************************** *********



AMD Socket 462 AVERAGES
3.1v = 237 MHz (5 Entries)
3.2v = 243 MHz (6 Entries)
3.3v = 248 MHz (5 Entries)
3.4v = 245 MHz (1 Entry)
3.5v = N/A
3.6v = N/A

AMD Socket 754 AVERAGES
3.1v = 242 MHz (2 Entries)
3.2v = 254 MHz (4 Entries)
3.3v = 256 MHz (6 Entries)
3.4v = 258 MHz (2 Entries)
3.5v = 264 MHz (1 Entries)
3.6v = 260 MHz (1 Entries)

AMD Socket 939 AVERAGES
3.1v = 234 MHz (4 Entries)
3.2v = 244 MHz (7 Entries)
3.3v = 255 MHz (9 Entries)
3.4v = 260 MHz (9 Entries)
3.5v = 265 MHz (5 Entries)
3.6v = 265 MHz (5 Entries)

Intel AVERAGES
3.1v = N/A
3.2v = 224 MHz (4 Entries)
3.3v = 241 MHz (3 Entries)
3.4v = 253 MHz (4 Entries)
3.5v = 257 MHz (3 Entries)
3.6v = 260 MHz (3 Entries)



***** uwackme's bh-5 burn-in guide thread at DFI-Street.COM *****

***** uwackme's bh-5 burn-in guide as Word document - hosted by NiCKE^*****


Cheers

[TEDY]

oh you getting nice 260 with only 3.3v ?

that is great... CPC on ? 1T ?

Did anyone get any decent ocing with them on DFI NF2 ?

[EnJoY]

Rabbi, with your 3.3v rail at it's current setting, with the 4.0vdimm bios you should be able to get 3.4-3.5volts. Why aren't you doing that?

[Rabbi_NZ]

Rabbi, with your 3.3v rail at it's current setting, with the 4.0vdimm bios you should be able to get 3.4-3.5volts. Why aren't you doing that?

My 3.3vRail = 3.6v with DMM when not loaded... in Smart Gaudian software it is actually read at 3.47v-3.5v... and Im pretty sure the 3.3vRail needs to be at least 1.5v higher than the vDimm.
I tried running it at 3.4v and it was only at that level unloaded... in Sisoft sandra it actually dropped to 3.3v so my 3.3vRail needs to be cranked up to get anymore vDimm, and I dont want to mod this PSU.

Rabbi, with your 3.3v rail at it's current setting, with the 4.0vdimm bios you should be able to get 3.4-3.5volts. Why aren't you doing oh you getting nice 260 with only 3.3v ?

that is great... CPC on ? 1T ?

Did anyone get any decent ocing with them on DFI NF2 ?that?

1T 2-2-2-10-7-15 3.3v

[EnJoY]

I had mine on my OCZ upt o 3.8, I suggest you try that and set it to 3.4volts. If that doesn't work, somethings wrong with your board.

[Rabbi_NZ]

I had mine on my OCZ upt o 3.8, I suggest you try that and set it to 3.4volts. If that doesn't work, somethings wrong with your board.

My 3.3vRail = 3.6v with DMM when not loaded... ...my 3.3vRail needs to be cranked up to get anymore vDimm, and I dont want to mod this PSU.

...

[STEvil]

try measuring the mosfet, mine was overvolting by .1v

You should try running CAS 1.5, too... and getting TRCD and TRP down to 1 (note: VX cant get TRCD-r and TRP down to 1 since it is not winbond chips, but UTT probably can because it is Winbond chips).

[Rabbi_NZ]

STEvil,
1) I ran 10x270 ~3.3v 1.5-2-2-0-7-15 on SuperPI no worries... but the performance diff wasn't even noticeable in benchmarks compared with 2-2-2-10.

2) I will be shocked if TwinMOS UTT and OCZ VX dont use the same chips... what gives you the idea they dont dude? What else could VX possibly be?
People have taken spreaders off and actually seen the two dimples on their VX chips..

3) Measured at mosfet and under load in Sisoft Memory Bench it was bout ~3.33v which is very close to the loaded voltage when vDimm is set to 3.3v... i think my 3.3vRail simply needs more voltage.
Also, remember that reading comes before the big resistor bank and more traces so it's possible the value drops even further after that... in fact, it probably does. Just because it's overvolting there doesn't mean the memory is getting that much voltage.

[MarilynMX]

I dont know how to define memtest stable since I'm running 258/259 at 3.6V first 50 loops no errors but avg 2~6 errors per 100 loops after #50. Also my memory controler limits the HTT up to 245 I think? I can't use office under 250x10 or 255x10 but rock solid on 240x11 wtf?

[Rabbi_NZ]

Thankyou MarilynMX

Sounds like it may be a heat issue for you if the first 50 tests of #5 never give errors.
With 260 I can pass #5 all day long no problem with 3.3v but #7 gives a single error every 10 passes or so... I think it is a memory controller issue too to be honest. I can Prime all day at 260 aswell, and I can run SuperPI 32MB up to 268 2x512 so it's def wierd.

Would you mind testing you RAM from 3.1v to 3.6v please?
I would be very interested in your results.

[STEvil]

STEvil,
1) I ran 10x270 ~3.3v 1.5-2-2-0-7-15 on SuperPI no worries... but the performance diff wasn't even noticeable in benchmarks compared with 2-2-2-10.

Thats why I suggested TRP 1 and TRCD-r 1 also.

2) I will be shocked if TwinMOS UTT and OCZ VX dont use the same chips... what gives you the idea they dont dude? What else could VX possibly be?
People have taken spreaders off and actually seen the two dimples on their VX chips..

Things that give me the idea UTT is not found on VX:
- VX cannot do TRP 1. Winbond can
- VX cannot do TRCD-r or TRCD-w of 1. Windbond can.
- VX can do CAS3, Winbond cannot (have not seen reliable example as of yet)
- Dimples come with the package type, not a feature of Winbond chips only.

3) Measured at mosfet and under load in Sisoft Memory Bench it was bout ~3.33v which is very close to the loaded voltage when vDimm is set to 3.3v... i think my 3.3vRail simply needs more voltage.
Also, remember that reading comes before the big resistor bank and more traces so it's possible the value drops even further after that... in fact, it probably does. Just because it's overvolting there doesn't mean the memory is getting that much voltage.

Measure at the DIMM slot then, mine was the same there as it was at the mosfet and there was no droop detectable by any of the three multimeters I own..

[Rabbi_NZ]

Thats why I suggested TRP 1 and TRCD-r 1 also.


Things that give me the idea UTT is not found on VX:
- VX cannot do TRP 1. Winbond can RABBI: "will try tonight, are you sure UTT can?"
- VX cannot do TRCD-r or TRCD-w of 1. Windbond can. RABBI: "will try tonight, are you sure UTT can?"
- VX can do CAS3, Winbond cannot (have not seen reliable example as of yet) RABBI: "UTT can with no problems"
- Dimples come with the package type, not a feature of Winbond chips only. RABBI: "name one other chip type that uses that package"

I'm not trying to start an arguement with you STEvil, I've seen you round here long enough to know you're no idiot... but I DO think VX is UTT.

[sjohnson]

Agree, TwinMOS SP (UTT, right?) does cas 3 NP. Don't know about the other values, might just have to try them

[Rabbi_NZ]

Agree, TwinMOS SP (UTT, right?) does cas 3 NP. Don't know about the other values, might just have to try them

If you could try it on TwinMOS UTT it would be very much appreciated sjohnson... also, if you could see how they scale between 3.1v-3.6v as mentioned in 1st post it would be great.

[EnJoY]

You say you don't want to mod your psu bet yet you alread have the 3.3 at 3.6? And PC Power & Cooling allows up to any voltage on all the rails, so they are already adjustable from the factory. Am I wrong? Sorry, not trying to pick a fight, I'm just confused.

[Rabbi_NZ]

It has pots from factory and turning them does not void warrantee, but they only go +/- 10% which limits my 3.3vRail to ~3.6v.

Once hipro5 finishes his DDR Maximizer I will buy that and give my RAM more juice

[EnJoY]

Ooooh, I see. Well best of luck with that then, the DDR Maximizer does look awesome.

[sjohnson]

If I get the time to try volts lower than 3.3, I'll post them as well.

Two 256M sticks TMSP here, unmodded Infinity NF2.

#1: 253 FSB SC, 3.3v, 2-2-2-11 all aggressive error free 24 hours memtest #5
#2: 257 FSB SC, 3.3v, 2-2-2-11 all aggressive error free 24 hours memtest #5

Both, 240 FSB DC, 3.3v, 2-2-2-11 all aggressive error free 24 hours memtest #5

BIG difference in SC & DC, I'm still playing around with BIOS' to find one with less difference. BTW, it makes no difference how loose I set memory timings, those are the max error-free speeds.

[Rabbi_NZ]

BIG difference in SC & DC, I'm still playing around with BIOS' to find one with less difference. BTW, it makes no difference how loose I set memory timings, those are the max error-free speeds.

Same for me sjohnson... there are only a few timings that change stability and it's none of the "BIG 4" timings.
Trc, Tref, and a few others can kill performance... usually once a stable speed is found I can set all timings to absolute max and it will remain stable.

[STEvil]

sjohnson, can you provide 2-2-2-11 vs. 2.5-2-2-11 vs 3-2-2-11 for the UTT to show the bios or SPD are not interferring?

[sjohnson]

What are you asking? Tests at those speeds? Sorry, I don't have a windows build ATM, my testing is all in memtest86.

At 2-2-2-11 set in BIOS and showing same in memtest86, memtest86 shows memory bandwidth of 1807 MB/s@240 DC

At 2.5-2-2-11 set in BIOS and showing same in memtest86, memtest86 shows memory bandwidth of 1708 MB/s@240 DC

At 3-2-2-11 set in BIOS and showing same in memtest86, memtest86 shows memory bandwidth of 1708 MB/s@240 DC

That's the best I can give at this time. I'm not up on the memory math, does the reported bandwidth at cl 2.5 and 3 of 1708 have any significance?

I'm pretty sure that my BH-5 won't post at CL 3, so I don't *think* anything is amiss but I'll admit I don't know for sure!

(Edit) whoa!, at CL3 I'm getting tons of errors. Going to try CL 2.5 to see what I get there. Back soon with another edit

(2nd Edit) CL3 gave 2006 errors in the first run of test 5, at the 99% level. CL 2.5 is error free after two loops. I'll try dropping the FSB at CL3 and see what happens.

(3rd Edit) 3-2-2-11 is error free at 210 FSB, 215 gave 68 errors at 99% on test 5. There may be a magic set of settings that give better results with CL3, but I'm not trying to find out, at least not tonight.

[Skip]

i had thought test 5 always gave the errors at the end, at least that is the only time i will ever get them.

also trfc is a very important timing, going from 16, the usual, to 12-11 or 10 really helps my bandwidth scores.

[sjohnson]

The end of test 5 involves relocating the running memtest program so as to test all the memory. So, it's a double-relocation plus the program has to continue to execute.

At least, that's the way it was described to me to explain the 99% error tendency at high FSB. If I push the memory high enough, I'll generate errors before 99%, though.

[reject]

i thought that the program just didnt list erros till the end so as not to interfere with the test

[sjohnson]

Nah, it shows the errors as they occur. The program is running in memory not being tested so isn't affected by the errors in memory under test.

memtest86 can err out itself, that's when you get stack errors, weird behavior and crashes.