[Team Latvia] closing in on Superpi WR's...

[mrlobber]

... yet not quite there.

Dropped the CPU in Sam's mailbox for a nice surprise in the morning :D

http://img401.imageshack.us/img401/4760/dsc09499dd8.th.jpg (http://img401.imageshack.us/img401/4760/dsc09499dd8.jpg)

So this means Sam & mrlobber (Team Latvia) just came back :D

Setup:


QX9650
P5K3-Deluxe fully modded / Foxconn ES BlackOps for 3D
Vikingpot 1.2 (LN2)
2x1Gb HyperX 14400 (D9GTR)
unknown PCI VGA / Asus 9800GX2 for 3D
FSP 600W value PSU
Samsung 40Gb
41L of LN2
"a little bit tweaked" windowsXP (Sam's job)


Coldboot around -105C, idle coldbug around -130C, load around -145C.

So let's move over to the results :D

Specially for 1M - BIOS downflashed to 0501

We had a score even closer to No_Name's WR (7.438) ... but rig crashed so that one doesn't count.

So how about 3rd place 1M :D

http://img133.imageshack.us/img133/6165/7469iz5.th.png (http://img133.imageshack.us/img133/6165/7469iz5.png)

And, even better, a 2nd place 32M? :cool:

http://img165.imageshack.us/img165/3953/803531si5.th.png (http://img165.imageshack.us/img165/3953/803531si5.png)

This one was a lot harder, mobo was not much 32M stable with multis above x12, and getting that 32M stability above 5.7G was a heck of a challenge

It actually took us ~8hrs and ~60 :p: failed attempts to get the 32M calculated properly... the CPU was absolutely maxed out on the board. So unless somebody suggests us a board which would gave 100 Mhz more on CPU, let's say MemoryExtreme Team Italy can still feel safe up there on the 32M hill :)

Then Sam went for a couple of hours long break to go to school and pass his state English exam (very needed actually to understand what's written in the forums :D)

Came back -> switched board to Sam's BlackOps -> added 9800GX2 -> let's bench 3D... with only 3L of LN2 left :D

Now this time BlackOps didn't want to run properly at extreme clocks, we had to boot at 333x13 and then work it up with clockgen ... pathetic :(

Nevertheless, score number 5 on hwbot... card still on air :banana:

http://img232.imageshack.us/img232/2796/111613gy7.th.png (http://img232.imageshack.us/img232/2796/111613gy7.png)

FM link: http://service.futuremark.com/compare?2k1=9382561

With more LN2, 112k or even 113k might be possible ;)

Now Sam just went for a 100k aircooled attempt this night but I said "enough", so enjoy what we've done above, and that's it :up:

[Cooper]

These are some very impressive numbers. Congratulations guys :clap:

[stunned_guy]

Awesome results !!!! Congrats.
That 1M time is very close to no_name...and you have around 150 mhz less than him.

[nachthymnen]

OMG:eek:
That's an amazing session guys:clap:
Really nice 1M and 32M:eek::clap:
Sorry for BlackOps results:(
I'm going to buy a retail Blackops...:rolleyes:

Congratulations:up:

[hiwa]

:clap::clap::clap: WOW nice 1 and 32 time congrats ;)

[massman]

Thanks for sharing all cpu info, guys!!

Make sure to crack it next week ;)

[gandalfone]

P35 ddr3 is very performing spi board

Great results

Congratz:clap:

[Webster]

Wow, great 1M and 32M times!:up:

[M.Beier]

As I told Sammy boy:
Luck, pfff ;)

Nah lookin' good :)
Can I somehow convince you guys NOT to try 32M again, you kinda interfear with my plans, you know? ;)

EDIT:
How about if I dont write on the OCX chat for a full week? Will that do? :p:

[N3RO]

Amazing PP in this 1M run ;)

Great job :)

[T_M]

Incredible guys, really amazing with a quad and your efficiencies.
That CH is mad, what were you clocks for that test?

[Oliver]

Sweet works guys... Good to see the cpu arrived ;)...

Sweet pp on the runs and also great 3d01 run...

Keep tearing it up...

[The Shooter]

Nice work:up: To bad for no WR in SPi 1M.

[c1rcu1tburn3r]

Great work guys...:up: I like to see unlocked multipler at work :D
I think you can go higher, try different setup till you'll kill the wr ;)

[hey]

unbelievable pp for that frequency guys :eek: keep pushing ;)

[xwards]

Congratulations guys :clap:
keep pushing :up:

[jimmyz]

Sick runs guys, Big numbers, you should use larger images though those thumbnails don't do this thread justice.

[qballe]

killer-chip ... compliments :clap::D

[before]

44.8k pp 1M... I never thought it was possible at those clocks. Well, it has to be the fastest clocks for clocks 1M run ever, hasn't it? :)

[M3kk]

Very nice scores, bros :D.

[Zannema]

Really impressive score :up: :up:
Congratz

[gianni-gt]

thats tweaking in perfection. ;) congrats!

[Bullet92]

you said little bit tweaked :D
congrats guys. :up:

tick-tack - time to break no_name down :)

[George_o/c]

:eek: guys ! :shocked: Great scores, that is the most incredible 1M run I've ever seen ... That's actually the first time I see a performance product < 45k :eek: Breathtaking efficiency, seems that with a little more MHz you can kick ass ... :up:

And what's with your rams at 32M ? 976MHz @ 7-6-2-6-15 :eek: What rams are you using and what Vdimm ? :)

The 3D 01 is great too, what clocks for the CH run ?

I'm sure we are going to have some 1M competition soon ... You guys can for sure crush the WR ;)

[lukas69]

Really amazing score congratz for 1M and 32M
:clap:

[makaka]

congrats , super pi 32 mb at 5855 mhz is just awsome

[kiwi]

44.8k pp 1M... I never thought it was possible at those clocks. Well, it has to be the fastest clocks for clocks 1M run ever, hasn't it? :)

With pre 45nm bios it is possible as you see ;)

[before]

With pre 45nm bios it is possible as you see ;)

Do you think that I dunno how pre-45nm BIOS improve 1M time? lol ;) I never pretend to be a terrific tweaker, just someone interested in tweaking (do you think that I've waited for monteboy's P965 strap app to play with MCHBAR offset registers?). Ok; after looking into MCHBAR/MSR etc, it's true that I've not succeeded to figure out why pre-45nm BIOS are performing faster than post-45nm BIOS for Spi1M. Maybe I'll, at least I hope so... :p:

Sub 45k pp at 6GHz even with a 45nm BIOS or a well-tunned OS is something that I've never expected in a reasonable way... That's what I wished to say. Nothing about saying that I don't believe in it, or something; TBH I really don't care about that. That's not my business. ;)

[kiwi]

I know that you know about pre 45nm but why surprised? Just because nobody else did it or only a few did it doesn't mean it can't be done :) Take a simple statement like no 2 extreme setups are same, hell even with the same hardware it might not even be possible to repeat it :)

And as far as I know SP runs were made on a very uniquely cut/optimized/tweaked OS that really works (well I saw that it does work) better than ordinary tweaked OS

[M.Beier]

And what's with your rams at 32M ? 976MHz @ 7-6-2-6-15 :eek: What rams are you using and what Vdimm ? :)

His D9GTR, of that Im pretty sure... They're about same as mine...
But hey - D9JNL is the best for benching ppl say :rofl:

[T_M]

kiwi etc, before was not in any way quetioning the result. It is clear he was just suprised that the efficiency can still be had at such high clocks (as per previous discussions, efficiency is often thought to drop as you go higher).

[mrlobber]

Those wondering about 970+ 7-6-5... oh well, we run HyperX 14400 @ Vmem = 2.40 , Vmch=1.85V for 1M and >2V for 32M. And this kit can do 900Mhz 7-6-5 32M @ 1.9 real Vdimm - D9GTR power for sure :D

3d2k1 - all tests at same freq (with 3L of LN2 you don't have much time to optimize :D) ... looks like GPU bottleneck is there already and without more "exotic" cooling significant gains might not be able to be made... whatever, run order for that 01 run: CH-LL-LH-DL-DH-N-CL

For 1M run, there actually exists a "small CW" for 1M as well ;) which helps gain around 0.03-0.045secs if done right. And even without that, with a carefully cleaned OS you can get a 45000-45100 PP ratio even on high CPU clocks :)

[dduckquack]

nice efficiency on those runs guys!

[before]

I know that you know about pre 45nm but why surprised? Just because nobody else did it or only a few did it doesn't mean it can't be done :) Take a simple statement like no 2 extreme setups are same, hell even with the same hardware it might not even be possible to repeat it :)

Can't really agree here. :) There's no magic; things can be explained and obviously reapeated in a major way; the remaining is the small part due to bug.

Dunno if it's still actual (probably not; would be nice cause I'd love to think that it could be improved), but at the time I did it, I've got the fastest clocks for clocks 65nm/P965 1M ran over 5GHz. Why? Tuned-OS? Right. But OS explained 'only' let's say 25%... MCHBAR chipset tuning was the key! IMHO nobody, apparts from the great JP :up: and I was playing with MemoryBaseAddress to improve benchmark performance at that time. I may be wrong; and that's obviously unverifiable cause 'Monteboy's app' or 'Eva2000's JP Strap Tweaks' have bring this method to public. And that's nice, now everybody can tune his/her system, without being familiar with MCH, offsets, registers, hexadecimal, etc. Monteboy has done a great app, and George a very nice synthesis (eva-thread style! amazing).

And as far as I know SP runs were made on a very uniquely cut/optimized/tweaked OS that really works (well I saw that it does work) better than ordinary tweaked OS

Well my OSes are well-tuned (not perfectly; that's right) and that's why I always used to get fast times with A64, C2D, etc. BTW the benefit from tuning Windows is definitely important but overrated... at one point, you need to tune something else (MCHBAR for example), change hardware, or install another OS (Linux is faster than XP/2k3 :D but you won't even get your windows to speed up Spi like Linux... don't even think about it) lol

Anyway, maybe I'm wrong... I've never pretended to be someone, to be the guard of Reality, etc. That's not supposed to be the truth, but just my own opinion after years of tweaking and Pi benching.

If it exists an OS tweak to get sub-45pp at 6GHz; and well why not? I offer to teach a bit of what I know about MCHBAR tuning in exchange. :yepp: In private instead of public if required.

[kiwi]

Maybe SP is buggy who knows. How do you explain that you get advantage only in SP and not any other application and not even in 32M?

Anyway, here (http://images.people.overclockers.ru/148252.png) is one screen to show how optimized that OS is :) It's not even the best time but try that with such settings if you can ;) But this is discussion out of this topic

[M.Beier]

To explain the fanatism of the benchmarking...

Lets dig up a picture of Sam from the chat...
http://img124.imageshack.us/img124/6457/img4171yp1.jpg

[Monstru]

Very good result, extreme tweaking...Well done guys!

[before]

Maybe SP is buggy who knows. How do you explain that you get advantage only in SP and not any other application and not even in 32M?

A bit of what I've already said; I dunno what happens. BTW I highly suspect the key is about cache instructions.

Anyway, here (http://images.people.overclockers.ru/148252.png) is one screen to show how optimized that OS is :) It's not even the best time but try that with such settings if you can ;) But this is discussion out of this topic

No disrespect; but here I don't see how getting sub-45pp @6GHz. Anyway, give me the same hardware, and well maybe you'll be surprised of what I can do. :p: No let's be serious, I didn't want to put myself or anybody else into an utopic challenge with nothing to expect at the end. In fact, I don't care much about results than about how to get them. Sorry, just telling me (or anybody else with experience in windows tuning) that it has been achieved thx to a fine tweaked-windows is a bit weak.

Don't get me wrong, I'm NOT saying that it's UNTRUE. I can I've missed the right OS tweak even after years of tweaking Windows but I've tried so many things, investigating the registry, looking at the API calls during Spi (thx to my bro koda :up:); etc.

Now; if, like I hope, it really exists a Windows-tweak to severely improve Spi 1M (I mean to save about 300 to 500pp @6GHz); come one don't be shy, I won't eat you; just let's start a discussion in private. I'm offering to teach about MCHBAR tuning, and you won't be disappointed, you'll be able to change timings that you've never heard of.

[Horus´]

A bit of what I've already said; I dunno what happens. BTW I highly suspect the key is about cache instructions.



No disrespect; but here I don't see how getting sub-45pp @6GHz. Anyway, give me the same hardware, and well maybe you'll be surprised of what I can do. :p: No let's be serious, I didn't want to put myself or anybody else into an utopic challenge with nothing to expect at the end. In fact, I don't care much about results than about how to get them. Sorry, just telling me (or anybody else with experience in windows tuning) that it has been achieved thx to a fine tweaked-windows is a bit weak.

Don't get me wrong, I'm NOT saying that it's UNTRUE. I can I've missed the right OS tweak even after years of tweaking Windows but I've tried so many things, investigating the registry, looking at the API calls during Spi (thx to my bro koda :up:); etc.

Now; if, like I hope, it really exists a Windows-tweak to severely improve Spi 1M (I mean to save about 300 to 500pp @6GHz); come one don't be shy, I won't eat you; just let's start a discussion in private. I'm offering to teach about MCHBAR tuning, and you won't be disappointed, you'll be able to change timings that you've never heared of.
Wasn't kingpins result done without the prebios? and that is already 45.5K PP...

[before]

Who are you? :)

I didn't bring someone's name in particular so please don't register here to start doing such BS. Please, that's not fair. (Anyway, I don't think Vince's score was done without pre-45nm BIOS... ;) But that's not the subject).

I feel that I don't need to write again what I've already written. Did I say the score is impossible? No; Can you read what I said? Yes; Can you understand it? Oh well yes, I apology for my English but whoever you are, I know you're not stupid... :)

Please, let me quote that... and believe me, I really think that it's possible to me to ignore the right method... How can I be sure to know everything? :shrug:

Now; if, like I hope, it really exists a Windows-tweak to severely improve Spi 1M (I mean to save about 300 to 500pp @6GHz); come one don't be shy, I won't eat you; just let's start a discussion in private. I'm offering to teach about MCHBAR tuning, and you won't be disappointed, you'll be able to change timings that you've never heared of.

[Horus´]

Who are you? :)

I didn't bring someone's name in particular so please don't register here to start doing such BS. Please, that's not fair. (Anyway, I don't think Vince's score was done without pre-45nm BIOS... ;) But that's not the subject).

I feel that I don't need to write again what I've already written. Did I say the score is impossible? No; Can you read what I said? Yes; Can you understand it? Oh well yes, I apology for my English but whoever you are, I know you're not stupid... :)

Please, let me quote that... and believe me, I really think that it's possible to me to ignore the right method... How can I be sure to know everything? :shrug:
Nope you didn't, it however feels like that when reading your posts ;)
Your english is just fine.:D

[before]

Mate, I'm just saying that I can't do it by myself with my actual knowledge... :) If possible; I'd love to; simple as that.

Despite of what I know I don't see how to do it by OS tuning (and if I'm not wrong about who you are; I'm sure you know that I'm not a noob on that matter). So, isn't it understandable that I'm bit surprised?

:)

[Horus´]

Mate, I'm just saying that I can't do it by myself with my actual knowledge... :) If possible; I'd love to; simple as that.

Despite of what I know I don't see how to do it by OS tuning (and if I'm not wrong about who you are; I'm sure you know that I'm not a noob on that matter). So, isn't it understandable that I'm bit surprised?

:)
I know, and i can feel your frustation about this since you probably spend more hours on tweaking pi then the entire forum togheter.(:D)
Just a question , you don't have to answer it, what is your score at 6ghz? Are you that far of then?:shrug:

[mrlobber]

Horus, I don't know who you are, but please, could we put this discussion aside as it seems starting to go the wrong way slowly...

Xavier, my highest respect to you, and I'd gladly like to discuss many things with you :) Unfortunately then, there seem to be tweaks some overclockers like to guard as much as their own life. A couple of the tweaks used in our runs are Sam's own and even though we're benching together on my hardware, his tweaked OS (and even more, the core of it has been given exclusively to him by another guy who isn't an overclocker but has immense knowledge about OS internal structure and processes) carries many secrets to me, which he politely :p: refuses explaining to me to the full extent. And I'm used to accepting that since contrary to him, while he's still at high school and has a ton of free time to begin, my family life and other everyday duties would never allow me spending as much time for tweaking as he has done in past 4 years, like for example, trying whichever windows theme / icon combinations possible for every bench out there (2D & 3D) etc... :(

[George_o/c]

Just a question , you don't have to answer it, what is your score at 6ghz? Are you that far of then?:shrug:

I want to add something ... It's not that common seeing QX9650s benching SPi at 6GHz ... :p:

[TheKarmakazi]

This is a VERY nice run congratulations guys. Whatever tweaks you used, they are sick. Did Bill Gates tweak your OS? :D