8 Packs 4ghz 32m Challenge win full Haswell 4770K Bundle.

PCZ said:
Well done Duff Man, nice play on the last minute submission.
Click to expand...

I was tweaking right up until the end... I really thought you were going to pull out another amazing time and pip me to the finish line, so I wanted the best score I could! Frankly I was expecting a slew of last-minute Haswell entries from all corners, but they never came...

I have onboard LAN disabled for all my benching runs, so I couldn't update easily - all had to be done by pen-drive. To upload a pic and continue benching I need to:

a) Reboot XP in "normal startup" from msconfig (so a USB stick can be read)
b) load pic onto USB
c) power down and switch HDDs over
d) reboot into Win 7 and upload picture
e) power down and switch HDDs back over
f) reboot to XP
g) set diagnostic startup and maxmem (etc) and reboot once more

... takes too much time when you're against a deadline, especially since the damn board only sees the IDE SSD about one time in three. I could check the progress in the thread on my phone easily enough, but couldn't really upload scores.

I had a 7:21 score for a while, but only got under 7:21 right at the very last run I had time for.


acme15 said:
I'll look out for an i7 4770K, Z87 motherboard, and some Corsair 2133MHz RAM on the MM in the next few days ;)

*shakes head*
Click to expand...

None of the kit will be going on MM, or anywhere else :)

It will be used to run large-scale simulations for the geological sequestration of CO2, using the PFLOTRAN open-source code.

Well, more accurately, I'll probably keep the motherboard and CPU for myself, and my existing Haswell kit will go to running simulations above :D
 
Fair enough, well deserved. I imagine you spent a gruelling amount of time in the BIOS :p

Please god make the next competition more sensible/fair/level/competitive though 8Pack :rolleyes:
 
Good to see the hardware being put to good use Duff-Man.

Next competition needs to be a bit more open to all to enter 8 Pack.

Without being negative as I had a lot of fun while competeting, it would have been nice for myself, Kei or somebody with a Gen1 i5/i7 to have won the prize as they would have benefitted the upgrade greatly
 
Frozennova
I think you missed the point of the competition.
It was more a showcase for haswell than a competition.
 
8 Pack said:
Great work Duff Man and to all who entered the comp. Many really good submissions and some great system tweaking going on which was what the comp was geared towards making your system efficient for the mhz.

Top Dog great Ivy work with DIMMS on Air.

Some amazing commitment to AMD and getting the best out of your systems on a bench that does not favour AMD at all!!!

Duff Man I will PM you tomorrow AM about the prize when I get back into work. Now to think about the next comp hmmmmmmmmm!!!
Click to expand...

Thanks 8 Pack :D

I had a lot of fun with this competition. I've never really taken memory tweaking seriously before, and certainly not OS tweaking, so I learned a LOT from this competition.

I didn't have a huge amount of time for runs, but I spent a scary amount of time scouring the web for info! I must have read through every available pi thread on the net, not to mention a whole host of info about how memory is accessed at a low level (and so what the timings mean).


Thanks for running this competition 8 Pack, and I'm very grateful for the kit :) I do agree with Frozennova though - if you are planning a new competition, then if it's at all possible to formulate something that has a prize for AMD, or for older-gen systems, then it might generate even more interest. I have no idea what that could be of course!
 
Not at all, a couple of posts back I said I was more than happy to come away with the knowledge I gained through participating.

I'm not going to lie, it's a little annoying that you needed Haswell to win Haswell but ocuk didn't have to do the competition or give such an awesome prize. So hats off to them in that respect.

I'm not going to moan like some people did when the competition started as it's great that ocuk looks after the community like they do.

But as this was ultimately about teaching people the dark art of memory overcooking there would have been a lot more people getting involved if more people had a chance of winning.

I took from it exactly what 8 Pack was trying to teach, if I was only in it to win I wouldn't have entered
 
Ai good effort to all those envolved, as long as people have enjoyed themselves and taken valuable knowledge away from participating then its a win win for everyone. I have a spare bedroom to clear up now lol, 3 pc's in bits, ram sticks everywhere. I'm still going to try to get sub 9 mins on my 920. :). Good effort all.
 
Well done to everyone that entered and managed to improve their times.

Technically 8 Pack's rules didn't say OcUK staff couldn't enter (just that they had to be forum members (which you'd need to be to post your entery surely?) so surely 8 Pack won? :D

I didn't enter, but I'd be interested to know what sort of real world improvements people get with their improved RAM speeds? For example how much faster is video encoding? How many more FPS do you get in games? How much quicker to files 'zip'/compress and decompress?
 
I think in terms of gaming the improvements are small but they are there. I might do some testing using different memory speeds along with the Asus ROG bench as that's quick and easy to bench.
 
I dont do a lot of gaming but going from 4ghz with stock settings on ht-link, northbridge and my old ocz, compared to tweaked HT-link, northbridge and my current mushkin ram I see about 10 FPS gain in the hunter, although using onboard GPU using low settings.
 
Tweaking my setup to get better times actually made me realise I could get more performance out of my system than I thought.

All it needed was a better cooler for the processor.

Previously I was running my i5 760 at stock 2.93Ghz / 3.33Ghz with turbo boost with the Arctic Freezer 7 Pro.

Bought myself a Noctua NH-C12P SE14 and I can now run my i5 760 at 3.6Ghz / 4.1Ghz with turbo boost and it's silent.

So this comp actually saved me money as I don't have the urge to upgrade now.

My graphics card though will probably be the next thing.
 
Congrats Duff-Man, your tweaking paid off in the end, great time

Duff-Man said:
What waza were you guys using? I ended up using a 768mb+512mb manual copy, with a 25s wait before engaging pi.

I tried alsorts of stuff that didn't help... Closing explorer, closing svc host, RAM drive. None of them seemed to help. A well set up copy waza is worth a good 4-5 seconds though...
Click to expand...

My wazza folder was three 632MB files 'rar'-ed into one big 1896mb archive on D:/

I had tried a maxmem of 600mb but found that 640mb worked best for my system with a pagefile of 512-512 on D:/

8 Pack said:
Top Dog great Ivy work with DIMMS on Air.
Click to expand...

Thanks 8pack, I love this ram, I've had it for years and been using it for AMD, X58, sandy bridge and now this Ivy setup

Here are my sub-timings if anyone is interested, great competition again 8pack, thoroughly enjoyed it and learned a lot from it :)

7m44922a_zps0b2ee4bf.jpg
 
Last edited:
That's a great time for ivy Topdog, wazza must have shaved off a fair chunk of time. Fastest I could go without wazza was 7:52.8.
 
topdog555 said:
My wazza folder was three 632MB files 'rar'-ed into one big 1896mb archive on D:/

I had tried a maxmem of 600mb but found that 640mb worked best for my system with a pagefile of 512-512 on D:/
Click to expand...

Interesting... I actually meant a 1536Mb and a 768Mb file (not 512 and 768). I didn't try .rar-ing them up at all though...

My maxmem was 610Mb in the end. I found that I got FAR fewer late-stage crashes with the extra 10Mb of memory. Also, for some reason, disabling two of the CPUs in BIOS gave me better stability (fewer late-stage failures), and slightly better performance when setting affinity to one of the CPUs.

I tried allocating different amounts of memory for a RAM drive, but couldn't gain anything from it. If anything it was always a few tenths slower than using the SSD. I tried the waza to the RAM drive, and also between the original partitions, but no dice.

My page-file was 375Mb. I played around with different values but didn't see a huge difference.

... I'm not surprised you've kept that RAM for a long time, it's some pretty damn impressive stuff!




In terms of the tweaking:

I always found it hard to tell exactly which changes were improving times and which were not, when it comes to the minor OS tweaks. Individual runs of pi seem to vary by as much as 0.8s, so you need to do a *lot* of runs to figure out a mean and variance for the results, and get a probability for whether the times have improved. This can make things really tedious, running the same case over and over then rebooting. Also it's best done at memory settings that will certainly pass (you do *not* want to be failing pi half the time when doing this!). I definitely spent more time on this than actually pushing the time itself - I only really brought everything together in the last couple of days.

Copy waza and a stripped copy of XP are definitely the biggest improvements by far - worth at least 4 seconds each when set up properly. Everything else is just fractions of a second (other than RAM tweaking of course). OS tweaking is definitely more time consuming and "open ended" than the RAM tweaking.

As far as the RAM goes, I was surprised by how much tuning the tertiaries helped - certainly more than the secondaries. Wish I'd tackled it earlier on actually.
 
Last edited:
1Deep said:
That's a great time for ivy Topdog, wazza must have shaved off a fair chunk of time. Fastest I could go without wazza was 7:52.8.
Click to expand...

Thanks 1Deep, I also made a video of the run just in case someone doubted it ;)

Duff-Man said:
I'm not surprised you've kept that RAM for a long time, it's some pretty damn impressive stuff!
Click to expand...

I've just found the receipt for the RAM, they were purchased new in Sept 2009, they will run at 1140mhz 7-10-7 20 1T on the sandybridge, a good investment I think


Duff-Man said:
In terms of the tweaking:
a stripped copy of XP are definitely the biggest improvements by far - worth at least 4 seconds each when set up properly. Everything else is just fractions of a second (other than RAM tweaking of course). OS tweaking is definitely more time consuming and "open ended" than the RAM tweaking.
Click to expand...

I totally agree with you Duff-Man, my XP is a stripped version dedicated to Super Pi, the image size is only 104,000 kb
 
You must log in or register to reply here.
Back
Top Bottom