MEMORY Voltage Scaling, please contribute...

[STEvil]

showing no bandwidth difference between cas 2.5 and 3 may show the board is not actually changing from 2.5 to 3.. its really up in the air though.

In windows going from cas 2 to 2.5 shows a small performance hit, barely noticeable. Going from 2.5 to 3 is nearly double but still barely noticeable.. so we'll have to wait on some results from someone else or some numbers from you if you get around to it later.

Thanks!

[enzoR]

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
- 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.



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..

VX is Winbond UTT, winbond doenst mean it can do TRP 1, TRCD-r or TRCD-w 1 and cannot do CAS3. Heck my CH-5 does CAS3. this stuff is manufactured on 0.11 together with infineon. Finally Bigtoe already said "The secreat is out"

[STEvil]

I have yet to meet any BH/CH 5/6 that cant get TRP to 1. Havent seen a single stick of VX do it yet.

Got a link to where he said that? I know its out too, but he was referring to its on the web, not that it is winbond based if it is the same one I remember.


What is your system, can you provide cas 2 vs 2.5 vs 3 benchmarks and a picture of your ram?

[enok]

what memtest are u using? have u guys tried the last memtest86+ 1.40?

now my UTT results:

3.1v - 244mhz - PASSING all tests of v1.40, only giving one error always on same bit at #7/#8
3.2v - ...
3.3v - ...
3.4v - ...
3.5v - ...

1T 1:1 2-2-2 , default vdd

my AN7 can only handle till 244 fsb... not too bad.

[HermS]

Heres the results from my DFi NF3

3.3v - 254mhz
3.4v - ...
3.5v - 264mhz

1T 1:1 2-2-2

Will do a more thorough test when my NF4 arrives.

[enzoR]

I have yet to meet any BH/CH 5/6 that cant get TRP to 1. Havent seen a single stick of VX do it yet.

Got a link to where he said that? I know its out too, but he was referring to its on the web, not that it is winbond based if it is the same one I remember.


What is your system, can you provide cas 2 vs 2.5 vs 3 benchmarks and a picture of your ram?

but does this UTT also do TRP1?

[STEvil]

I dont know, nobody has even tried as far as I can tell and I dont have any to test.

[googles]

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

same here...ive never had errors occur until the 98/99% mark on test 5

[enzoR]

I dont know, nobody has even tried as far as I can tell and I dont have any to test.

if it didnt then i guess that would be 100% proof VX is UTT right?

[Rabbi_NZ]

Herms, how much memory are you using... 2x256 or 2x512?

I have added both of the results from Herms and enok to the first post. Thanks for the feedback

STEvil,
how many people have tried running those settings on VX?
how many people have tried running those settings on CH/BH?
maybe it's jus a case of you having good BH/CH and poor VX?
do you have any idea what chips MIGHT be on VX if not winbond-based?

From what I've gathered UTT is based on the same process as CH, all be it improved... BUT, the chips are not speed binned by Winbond themselves, they are just sold off as UnTesTed chips.
That's why TwinMOS speed bins the chips as 60D/50D/44D themselves... 44D is, in effect, CH44.

[EnJoY]

Both my BH5 and BH6 could not do TRP1 without getting a lot of errors. And both of those sets did 265 in my A64 and ran every other timings as tight as possible. So,,,I think trp1 just varies per stick, not per type (ie BH5, CH5, BH6, UTT, etc)

[HermS]

Rabbi, its just one stick of 256Mb of AA4T 50D, unfortunately one of the pair I ordered was D.O.A, should have a replacement next week.

*edit* doh just saw you wanted 2 x 512 results, sorry to confuse things Rab.

[Rabbi_NZ]

Sorry if I mentioned I only wanted 2x512... I didnt mean to.
Your results are appreciated AND useful

[aCidbAbY]

currently getting 260mhz 2-2-2-10 1T memtest86 stable @3.56-3.6vdimm.

[Rabbi_NZ]

currently getting 260mhz 2-2-2-10 1T memtest86 stable @3.56-3.6vdimm.

Would you mind running tests from 3.1v-3.6v please aCidbAbY?
I'm trying to see how these UTT chips scale with voltage...

[STEvil]

if it didnt then i guess that would be 100% proof VX is UTT right?

No, it would show that UTT may not be capable of what the "old" winbond chips are.

STEvil,
how many people have tried running those settings on VX?

Myself so far, two a7n8x-dlx with about 4 bioses, an nF7 with about 8 bioses, and a DFI Lanparty-UT with 3 different ones. They would not POST at those settings.

how many people have tried running those settings on CH/BH?

Here is a year old thread... there are more examples in the extreme overclocking section and I think I may have seen one or two here before too although nobody pays them much attention it seems. http://www.xtremesystems.org/forums/...ad.php?t=27206

maybe it's jus a case of you having good BH/CH and poor VX?

My VX seem to keep up with every other stick of VX out there, give or take 5mhz depending on the memory controller my CPU had when they were tested on the 754 setup.

do you have any idea what chips MIGHT be on VX if not winbond-based?

If I did I wouldnt be allowed to say. It was stated once or twice that they were not winbond based in the VX thread iirc, though..

[Rabbi_NZ]

If I did I wouldnt be allowed to say. It was stated once or twice that they were not winbond based in the VX thread iirc, though..

Minor technicality, or straight out lie in my opinion.
They may have been refering to the fact they purchased them from Infineon seeing as they buy a LOT of the UTT chips or maybe they just lied to us.
Plain and simple, I see NO arguement that proves these are NOT Winbond chips.
NO OTHER chips I have seen have the 2 dimples on top, and NO OTHER chips I have ever used, or heard of even, react to voltage or low latencies like Winbond chips.
The odd TCCD module and very select few Micron based ones have, but that is very rare... and they DO HAVE DIMPLES on top.

I realise you may not be allowed to say, even if you do know, but unless something concrete is put forward they will remain UTT in my mind.


Anyway, I didnt want to start a debate in this thread, just wanted to get voltage scaling data.

STEvil,
if you still have your VX or some UTT would you mind posting your experiences between 3.1v-3.6v?

Cheers for a good debate. No ill feelings are meant if I come across that way.

Rabb

[aCidbAbY]

ok so what you want me to do is see what clocks i can get with 3.1v - 3.5v

[STEvil]

I already did, but I only tested 3.2v+.

255 @ 3.2v, 260 @ 3.3v but unstable

Have the VX at 2.9v in my a7n8x right now, 2-3-2-11.


Samsung TCB3 acted like BH-6 back in the early days..

[Rabbi_NZ]

STEvil & aCidbAbY,
if you get a chance could you do this please:

Find where the memory is ERROR FREE in MemTest (+ or 86) Test #5 for 25-50 loops (50 preferable but I realise it takes a while to test.
Find the max MHz your RAM will do at 2-2-2 1T with 3.1v-3.6v under those conditions... ERROR FREE.

Thanks

[aCidbAbY]

what do you need for tras? 10,8,6?

[Rabbi_NZ]

Tras & Trc make no difference in terms of max clock for me, but Trfc should be 15.

For reference only, I run 2-2-2-10-7-15

[enzoR]

this UTT is actually infineon... made on winbond machines. stateing that VX isnt winbond could still mean its UTT.

[Rabbi_NZ]

this UTT is actually infineon...

Winbond still exist, they sell "most" of the chips to Infineon, but not all.

made on winbond machines. stateing that VX isnt winbond could still mean its UTT.

agree 100%

[trans am]

Highest possible memtest86 stable:
On single channel DFI board, my VX does 264 2-2-2-10 2x512

Twinmos SP UTT does 261 with 2x256
one of the 256 sticks can max out at 267mhz.
best performance occurs @ 3.4-3.5v