狂少's CDT-tweak

[wittekakker]

A programmer for Electrically Erasable Programmable Read-Only Memory, the hardware chip where the component BIOS is saved on.

[s7e9h3n]

A programmer for Electrically Erasable Programmable Read-Only Memory, the hardware chip where the component BIOS is saved on.

Or where the SPD is stored on memory sticks

[hipro5]

Ha, ha, ha...I was testing 32M at 3600MHz with the Yorkfield.....
I had to go though coz my wife was expecting me to eat so I'll continue tomorrow.....
The score was 12.08min without Copy Waza or anything.....
York IS FAST....

[over@locker886]

Ha, ha, ha...I was testing 32M at 3600MHz with the Yorkfield.....
I had to go though coz my wife was expecting me to eat so I'll continue tomorrow.....
The score was 12.08min without Copy Waza or anything.....
York IS FAST....

[VictorWang]

oh, i'm not in a good mood today 'coz of 1st time i got a THREATEN here on XS by someone big.

about opb's tweak.
yes, it's complicated and may not be that easy to fully understand.
same here, i also got tons of questions about that and totally shocked by the performance he shows.
but, i will not doubt him but will try myself to see if i can catch up with him.
if not, i'll ask him in person for some hints.
this is the way what I think is the right way we should go in such situation.
that's not easy to find and post such a tweak.

[blossa]

oh, i'm not in a good mood today 'coz of 1st time i got a THREATEN here on XS by someone big.

about opb's tweak.
yes, it's complicated and may not be that easy to fully understand.
same here, i also got tons of questions about that and totally shocked by the performance he shows.
but, i will not doubt him but will try myself to see if i can catch up with him.
if not, i'll ask him in person for some hints.
this is the way what I think is the right way we should go in such situation.
that's not easy to find and post such a tweak.

THAT, does not sound good.

About the tweak, I think I have the same 'issue' as eva2000. Atleast I think it was he... I almost get no gain regardless of what tweak I use. I have not tested this tweak yet but I will test it regardless of the result other has archived.

For me it does not matter if someone can run SP32M @ 3.6GHz 12 minutes flat with Yorkfield or if someone can get 10 seconds better time with a tweak. I greatfully try to learn from every posted tweak so that I myself maybe can get a better time. (Thanks for sharing! )

[Supertim0r]

I have an example of a 32m run done last month.

It's not CDT-Tweak, its other/older tweak. Basically 2 ".bat" and 2 ".reg"



It desactivate all services, clean temp and few others things I ignore.

NOTE : I'M EXTREMELY BAD AT SPI / TWEAKING

first run is without OPB tweaks, almost stock windows

13m21s

second run is with OPB tweaks, same stock windows

13m04s

I've tested 3 run of each and the results were similars

It's done in single channel (was waiting other memory)

[Onepagebook]

Ha, ha, ha...I was testing 32M at 3600MHz with the Yorkfield.....
I had to go though coz my wife was expecting me to eat so I'll continue tomorrow.....
The score was 12.08min without Copy Waza or anything.....
York IS FAST....

lol master, next time don't type 12m08s like 12.08,
it scares me..
the best time for spi 1m from yorky is like this:
no tweak no driver(only intel inf installed)

[Zeus]

Nice one Kevin!

These Yorks are in another league, unbeatable with current CPU's.

What's the trick to get performance level 5 to work?

With a divider i can get as low as level 6, running 1:1 won't let me run lower than level 7.

[Onepagebook]

Nice one Kevin!

These Yorks are in another league, unbeatable with current CPU's.

What's the trick to get performance level 5 to work?

With a divider i can get as low as level 6, running 1:1 won't let me run lower than level 7.

what about run your strap down to 1066 and take the 4:5

[Zeus]

I have an example of a 32m run done last month.

It's not CDT-Tweak, its other/older tweak. Basically 2 ".bat" and 2 ".reg"



It desactivate all services, clean temp and few others things I ignore.

NOTE : I'M EXTREMELY BAD AT SPI / TWEAKING

first run is without OPB tweaks, almost stock windows

13m21s

second run is with OPB tweaks, same stock windows

13m04s

I've tested 3 run of each and the results were similars

It's done in single channel (was waiting other memory)

Thanks for sharing.

I have no idea if there are others like teamjapan for example that are already using something similar but if not...

Now if we take this tweak together with the CDT tweak plus one super finetuned memory subsystem we might just see some very nice results.

What if OPB has even more tricks like these up his sleeve?

lol

[Zeus]

That's why we need you around man, i've been asking all over what that strap function in bios was, nobody seemed to know.

You were spot on:

[Onepagebook]

That's why we need you around man, i've been asking all over what that strap function in bios was, nobody seemed to know.

You were spot on:

you can't cheat like that

**don't do 32m on the 3120t from the TREF...that's gonna waste too much.. you can try 16384 on the 32m

[jimmyz]

Wow, good to see you post here again OPB. I hope you keep it up. I am not a Pi bencher so alot of this is slightly over my head, but that is how we learn.

[elmor]

I'm currently not seeing any gain from the CDT-IV tweak after reinstalling the OS, need to do some investigation on what's different from the other one.

The gain Supertim0r is seeing is probably mainly from LSC = 1, however after testing with SPI 8M I see a 0.2s gain constantly with OPBs regtweaks and the cleaner.

[Galaren]

Well im just glad to see one the greats back, furthermore since i havent been playing around with my intel system im glad to see Tref mentioned, that has got to be the most mysterious timing...

Edit: i remember some test i did on A64
0016= 200MHz(7.8us) 11x230MHz 1:1 Latency 159ns Memory Read 404Mb/s (Everest Ultimate Ed.2006)
0032= 200MHz(1.95us) 11x230MHz 1:1 Latency 41.1ns Memory Read 4820Mb/s (Everest Ultimate Ed.2006)
0064= 200MHz(7.8us) 11x230MHz 1:1 Latency 40.2ns Memory Read 6452Mb/s (Everest Ultimate Ed.2006)
3120= 200MHz(15.6us) 11x230MHz 1:1 Latency 38.8ns Memory Read 7120Mb/s (Everest Ultimate Ed.2006)
4672= 166MHz(7.8us) 11x230MHz 1:1 Latency 38.8ns Memory Read 7125Mb/s (Everest Ultimate Ed.2006)

[Zeus]

I'm currently not seeing any gain from the CDT-IV tweak after reinstalling the OS, need to do some investigation on what's different from the other one.

The gain Supertim0r is seeing is probably mainly from LSC = 1, however after testing with SPI 8M I see a 0.2s gain constantly with OPBs regtweaks and the cleaner.

Rather weird question maybe but did you run windows update with your new install?
I have a feeling this might have something to do with CW and CDT not working properly, don't know for sure though, it's just a hunch.

[elmor]

Rather weird question maybe but did you run windows update with your new install?
I have a feeling this might have something to do with CW and CDT not working properly, don't know for sure though, it's just a hunch.

No Windows Update.

[eva2000]

I'm currently not seeing any gain from the CDT-IV tweak after reinstalling the OS, need to do some investigation on what's different from the other one.

The gain Supertim0r is seeing is probably mainly from LSC = 1, however after testing with SPI 8M I see a 0.2s gain constantly with OPBs regtweaks and the cleaner.

Yeah i'm also doing a fresh windows install with fresh unattended compiled winxp pro sp2 cd to see if it helps

[mrlobber]

I'm currently not seeing any gain from the CDT-IV tweak after reinstalling the OS, need to do some investigation on what's different from the other one.

That's what my theory is, the CDT tweak works significantly better than a good copy-waza only on some special OS conditions which many of us haven't been able to duplicate yet In the last days I've tested like a couple of dozen of different copy-waza methods on the same OS and identical hardware configuration... all of them, including CDT (the part I know about it at the moment), have a difference of maximum 3 seconds at 3600Mhz.

[hipro5]

lol master, next time don't type 12m08s like 12.08,
it scares me..
the best time for spi 1m from yorky is like this:
no tweak no driver(only intel inf installed)

We here in Hellas, pronounce it: 12.08m....

Yes York is fast.....

[BeardyMan]

hehe ddr3 is known to be faster then ddr2 in superpi 1M

[Yoxxy]

Hipro, what memory voltage is that? 2.28v?

[kiwi]

Guys, this tweak definitely works! I tested it myself and shaved of 19 seconds vs regular run and 15 seconds vs copy waza. I didn't even bother copying 3 times and using maxmem (was just testing around )

Here are my times:

Regular:
13m 21.781

Copy-waza:
13m 17.453

CDT as discribed by elmor but without maxmem and I copied file 3 times (D>C) + 1 time (D>D) + 1 time (C>D)

13m 02.625

Might do even better if I did maxmem + 3 times copy


See attached images

[NiCKE^]

Not bad at all!