Hardware Canucks

Hardware Canucks (https://www.hardwarecanucks.com/forum/)
-   Overclocking, Tweaking and Benchmarking (https://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/)
-   -   SS phase change powered E6850 / Ballistix DDR3 (https://www.hardwarecanucks.com/forum/overclocking-tweaking-benchmarking/3193-ss-phase-change-powered-e6850-ballistix-ddr3.html)

3oh6 November 8, 2007 01:46 PM

SS phase change powered E6850 / Ballistix DDR3
time to liven the benchmarking section up here a little bit.

i have had my head buried in work, reviews, and classes for what seems like a lifetime. the other night i decided that sleep was not required and a benching session with the new E6850 i recently treated myself to was just what the doctor ordered for a little relaxation. needless to say that when i decided on this course of action at 12:30 at night as i was riding home from work, dry ice was out of the question so i simply hooked up the phase and figured, "perhaps i will get some decent clocks"...hehe, yeah, decent :biggrin:

i had no intentions of benching this much but when the little Intel monster started purring i couldn't stop until i had to take off back to work...thats right, a solid 12-13 hour bench session and all i ran was 32M SPi. if that isn't an addiction to 32M SPi, i don't know what is. here is the list of hardware used:

MB:          Asus P5K3-Dlx Wifi (vMCH mod)
CPU:        Intel C2D E6850 (L724A479)
CPU Cooling: Chilly 1 single stage phase change
RAM:        Crucial Ballistix PC3-12800 8-8-8 
GPU:        Biostar 8600GTS 512MB DDR3
NB Cooling:  Noctua NC-U6
SB Cooling:  Thermalright HR-05-SLI
PWM Cooling: Naked
PSU:        Silverstone Zeus 560W
HD:          Seagate SATAII 80GB 8MB NCQ
OS:          Windows XP Pro SP2

Ambient Temperature: 16-17C - yeah i was wearing a hoody and a toque

i apologize in advance for no photos of the setup but it was dark and i only like shooting with natural light so nothing but screens. it is basically the combination of the two setups below. the cooling on the motherboard in the second image, and the cooling for the CPU in the first image.

Click for full size...

enough babbling though, here are some results:

PB sub-13min 32M:
3535.1MHz SPi 32M = 12m 59.266s

the goal of this little challenge is to get sub-13min with as little CPU frequency as possible. so basically to get 12min 59.XXX at as low a CPU clock as you can get. this is a very popular challenge amongst SuperPi tweakers as most guys can play because the frequency required is usually 'only' just above 3.6GHz. this time below is one of the better results out there in this challenge...i am quite proud of this result.
click for full size including memset...

4500MHz SPi 32M = 10m 36.000
this run was basically a warm up...i was simply wanting to see if 4.5GHz was going to be possible at that is pretty much the limit of my E6600 with this cooling. lets just say i didn't run any slower than this the whole night. as soon as this run finished, i knew it was going to be a looong fun night because the voltage used was quite low and thinking back, it probably didn't even need that much.
click for full size including memset...

4639MHz / DDR3-2062 SPi 32M = 10m 16.750
the next major goal was to beat my personal best 32M SPi time of 10min 17 once i new this CPU was going to clock as high as it did. i managed to beat my previous personal best time but at almost 200MHz less (you can find that result here), which is absolutely insane. DDR3 in the right setup is completely lethal in SPi and all benchmarks. in everyday use, not for the money yet, but for benching...absolutely. the only catch is that this is where the system topped out with these settings. i will hopefully get some better cooling on the northbridge, chilled water or dry ice, for the next run and that will potentially get me a little more with high clocked memory and the tight chipset latencies of PL6.
click for full size including memset...

Did someone order 14K read bandwith in Everest? How about some stoopid tight latency?
after seeing the bandwidth almost top 14K in Everest, i had to push it that extra little bit as i had never seen 14K read in Everest before that i could recall. i have now :P
click for full size including memset...

PB Overall 32M SPi:
4768MHz SPi 32M = 10m 10.891

the last of my runs is my new personal best 32M SPi time and i still haven't come close to the top frequency i can bench at with me E6600 under dry ice so when i do a proper session with dry ice and my new best friend (E6850), this time will drop like a rock. based on these phase results, 5GHz benchable on dry ice shouldn't be a problem which means i will definitely get that infamous sub-10min run that i have wanted for sooooo long.
click for full size including memset...

as always, don't hesitate to ask any questions or post comments. some of this may be way to geeky for most guys here but i know some of the fellas here like to see some over-the-top overclocking and benching. also, don't worry if you don't understand my obsession with benching 32M SPi. some guys like to bench 3D, some guys like to race their cars on the track or the 1/4 mile...my main hobby is overclocking and benching 32M SuperPi :thumb:

3.0charlie November 8, 2007 02:30 PM

I'm lost for words. Those are insane benches. And I taught that I had a good bench with a 16:08 dual-32Mb SuperPi run....

Good job, and I sure hope you had good times!


Babrbarossa November 8, 2007 05:11 PM

Amazing- do you think that RAID and/or raptors would trim that back much more? I know that they're overrated, but I'd like to know by how much.

enaberif November 8, 2007 05:26 PM

Hahahah look at that CPU voltage .. 1.6v!

Prof. Dr. Silver November 8, 2007 05:55 PM

That is just sweet! I might get me one of those CPU's also! And I have to start saving for one of those Phat Phase Changers!! LOL If I get one of those bench-sickness-moments...in the winter I just insulate my HDD's and put my rig outside! Let it freeze up and then start benching....That is how I managed to get a D 930 up to 4.5GHz...LOL

Babrbarossa November 8, 2007 08:57 PM

hehe- yeah, just throw a little anti-freeze in your water loop, and away you go

3oh6 November 9, 2007 01:02 AM

thanks guys


Originally Posted by Babrbarossa (Post 25537)
Amazing- do you think that RAID and/or raptors would trim that back much more? I know that they're overrated, but I'd like to know by how much.

with the tweaked setup i run my SPi with, raptors wouldn't do a damn thing...i don't think. with an un-tweaked setup, Raptors may help the slightest bit with 32M benching but even then i couldn't be sure.


Originally Posted by enaberif (Post 25541)
Hahahah look at that CPU voltage .. 1.6v!

1.6v is nothing, the processor is at -40C. i didn't think a vCORE mod was going to be necessary for phase benching but i think i might be able to get more out of it over 1.7v, even on phase.

i am still waiting on word about my copper CPU pot but if it doesn't look to be anytime soon, then i might just have a session with the aluminum pot i have now to see if i can get that sub-10min 32M. this chip produces a lot less heat than my E6600 so the temp swings with the aluminum pot at 1.75v might be small enough to get 32M ran at around 4.9GHz me thinks.

Eldonko November 9, 2007 11:21 AM

Still waiting on the F1? Great runs btw, vcore and/or vdroop will crack that 10 min mark for ya.

3oh6 November 9, 2007 11:35 AM


Originally Posted by Eldonko (Post 25608)
Still waiting on the F1? Great runs btw, vcore and/or vdroop will crack that 10 min mark for ya.

yeah...Vince is a busy guy so its understandable, what with traveling the globe benchmarking and all :biggrin:

oh, and vdroop??? whats this vdroop you speak of? voltage damper enabled is rock solid with my board. some guys boards still droop but with C2D, it doesn't drop a bit...with a quad, perhaps. i will definitely be doing the vcore before my next session, and perhaps i will be ballsy enough to do the vdimm as well :thumb:

Eldonko November 9, 2007 11:46 AM

Interesting. DDR2 P5K vdroop is the worst in any board I have ever saw (w/ a quad). 1.665 BIOS vcore = 1.37v real load. WTF. Basically you have to do the mods to do any real benching, vcore BIOS max is 1.7v, which is probably barely over 1.4v real load. :sad: This is from P5K and P5KC, wonder if the DX is better..

All times are GMT -7. The time now is 03:09 PM.