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

[Audigex]

Thats exactly what I thought Stulid. This comp I wanted to be about system and memory tweaking and not CPU power. hence the 4ghz which I honestly though most could achieve.

a lot of other forums do 5ghz but I deemed that too elitist. I dont think 4.oghz on AMD or anything SB or beyond is difficult at all. I want people to enter!!!

A agree with your idea - but 4GHz is still out of reach for many members here: many of us can't afford to upgrade often enough for that. Don't get me wrong, this is a free prize and OcUK have every right to give it away how you want - I have no right to free stuff and I'm quite happy to buy my upgrades if I'm not lucky enough to win them... but a semblance of inclusiveness and fairness in the contest would make it feel a lot less like OcUK is only interested in those spending money for the latest, greatest kit.

I'd suggest, that most people are members of two groups - either the "constant upgrade" crowd who will have 4Ghz capable kit, or those with 3-6 year old machines just starting to check out their next upgrade.

To me, a more interesting competition would've been a few smaller prizes for different categories: a 2GHz dual-core contest etc - although tbh, it's still going to be the newer Intel chips that win, as they've got better clock-for-clock performance than older Intel or AMD chips.

A nice basic idea, but flawed at the most fundamental level - and tbh, OcUK know enough about computers to be aware of how intrinsically flawed it is: even without an unobtainable-for-many 4GHz limit, per-clock performance is much higher with newer architectures than old.

As mentioned above, anyone who can realistically win this is going to have an i7, probably one of the newer ones (Haswell, or at least Ivybridge) and have upgraded recently.... making it a contest for the elite or those lucky enough to have done a big upgrade recently.

The only way this could be realistically fair would be to split it into dual and quad-core (I'll even let you ignore those of us, like me, with tri-core chips), for Core2Duo, the various generations of the 'Core i' chips, and each of the major AMD Architechtures (AM2, AM3, FM1, FM2). It would've needed a lot of prizes (I make it 8 AMD and 8 Intel, ish, for 4 generations of both quad and dual core), but at least most active members could participate on equal terms.

The fact that 4 of the top 5 times were done with the exact CPU which is to be given away as the prize, frankly, says it all. I somehow can't see a Q6600 breaking into the top 10.

 

[Duff-Man]

What you need to test guys if different size wazza for your system and different pauses between copies and when you start the benchmark. The problem really is it differs as each system config. changes.

Interesting Duff Man. You wanna play Pi sometime with me let me know. Did you use MAXmem?? 640??? 620??? I will test your style for sure.

Maxmem was 610, and VM size was 375Mb. (not 100% sure on VM size but I think that's right - will check my notes when I get home). I had maxmem at 600Mb for quite a while, but 610Mb seemed to reduce the number of late stage failures I got.

100% agreed about the waza needing optimisation... some file sizes didn't do much of anything, even copying the same total amount across. Lots of small files was rubbish for me. Also the optimal wait time definitely changed depending on the type of file transfer. I don't claim to have found the best solution, but it worked fairly well for my setup Always hard to tell what's working well though because any two runs using the exact same procedure seem to vary by half a second or so anyway.



edit - Not sure I'd have a chance playing pi with you I only have a couple of sticks of RAM, and I have no experience with LN2. Plus I'm still a noob at RAM tuning. One day I might have a crack at cold though, it looks ****ing awesome Pots of steaming liquid rising from the motherboard = instant science-boner

My next computing type project will be a watercooled case though. I've got a few ideas for gathering data within a WC loop (lots of sensors), and I have an outline for a CAD model so I can get a prototype laser-cut. Using a raspberry pi I should be able to compute all sorts of thermal efficiency quotients, and use them to balance fan noise against heat load, and to track any efficiency drops (like warning of a dust buildup or waterblock clogging etc).

After that's done, maybe I will have the balls to try cold!

 

[PCZ]

?! Your times were without any copy waza?!

I thought you were using all the super-secret tricks that nobody posts about? Plenty of posts about copy waza in other super pi threads on the web.


Still - very impressive if that score was without any waza. I'm sure I couldn't have matched your time without it... probably would have been 1.0 to 1.5 seconds behind you.

Lucky for you then.

 

[Frozennova]

Lucky for me too,

My times were between 3 and 5 seconds quicker from memory with waza.

Then another 3 seconds for ram drive.

 

[8 Pack]

A agree with your idea - but 4GHz is still out of reach for many members here: many of us can't afford to upgrade often enough for that. Don't get me wrong, this is a free prize and OcUK have every right to give it away how you want - I have no right to free stuff and I'm quite happy to buy my upgrades if I'm not lucky enough to win them... but a semblance of inclusiveness and fairness in the contest would make it feel a lot less like OcUK is only interested in those spending money for the latest, greatest kit.

I'd suggest, that most people are members of two groups - either the "constant upgrade" crowd who will have 4Ghz capable kit, or those with 3-6 year old machines just starting to check out their next upgrade.

To me, a more interesting competition would've been a few smaller prizes for different categories: a 2GHz dual-core contest etc - although tbh, it's still going to be the newer Intel chips that win, as they've got better clock-for-clock performance than older Intel or AMD chips.

A nice basic idea, but flawed at the most fundamental level - and tbh, OcUK know enough about computers to be aware of how intrinsically flawed it is: even without an unobtainable-for-many 4GHz limit, per-clock performance is much higher with newer architectures than old.

As mentioned above, anyone who can realistically win this is going to have an i7, probably one of the newer ones (Haswell, or at least Ivybridge) and have upgraded recently.... making it a contest for the elite or those lucky enough to have done a big upgrade recently.

The only way this could be realistically fair would be to split it into dual and quad-core (I'll even let you ignore those of us, like me, with tri-core chips), for Core2Duo, the various generations of the 'Core i' chips, and each of the major AMD Architechtures (AM2, AM3, FM1, FM2). It would've needed a lot of prizes (I make it 8 AMD and 8 Intel, ish, for 4 generations of both quad and dual core), but at least most active members could participate on equal terms.

The fact that 4 of the top 5 times were done with the exact CPU which is to be given away as the prize, frankly, says it all. I somehow can't see a Q6600 breaking into the top 10.

I will think of a new comp to encompass more hardware next time. To be fair a lot have still entered this and done well and learned a lot. But I except what your saying.

 

[Audigex]

Sounds good - and as I said, I appreciate and applaud the effort/principle, even the execution... it's just the detail which needs some work to avoid accidental elitism.

 

[dfour]

yeah I think you should have excluded all times on haswell chips so the haswell went to someone who didnt already own one.

@8pack any idea why tightening the times on my rig made the times worse?? Could it have been better to go for more speed with loser timings than just tighter timings??

 

[BileMonkey]

FWIW, I had a go at this, got to low 1:30s with a 4670K and quickly realised I was never going to to win, as to pull a truly competitive time would have required much bigger cajones when pulling the voltage lever. But what I did learn in the process was that my RAM would run happily at stock voltage at 9-9-9-27-175, which, if I'd bought the same type of RAM with that as the XMP, would have cost me over 100 quid more. So in a way, I feel like a won a decent memory upgrade which probably benefits my system speed as much as the 4770K prize would have for most apps. So many thanks to 8Pack and the rest of you guys

 

[Duff-Man]

FWIW, I had a go at this, got to low 1:30s with a 4670K and quickly realised I was never going to to win, as to pull a truly competitive time would have required much bigger cajones when pulling the voltage lever. But what I did learn in the process was that my RAM would run happily at stock voltage at 9-9-9-27-175, which, if I'd bought the same type of RAM with that as the XMP, would have cost me over 100 quid more. So in a way, I feel like a won a decent memory upgrade which probably benefits my system speed as much as the 4770K prize would have for most apps. So many thanks to 8Pack and the rest of you guys

You should have posted a time anyway! You would have been 3rd or 4th - very respectable. Top 4670k time probably...

Glad you learned more about your RAM though. What speed are you running 9-9-9-27 at? And what sticks are they?

 

[Gzero]

yeah I think you should have excluded all times on haswell chips so the haswell went to someone who didnt already own one.

@8pack any idea why tightening the times on my rig made the times worse?? Could it have been better to go for more speed with loser timings than just tighter timings??

Did you do a quick memtest with the tighter timings, I've always found that the error rate goes up if you go past a certain point when overclocking that it is detrimental to your times/performance.

 

[BileMonkey]

You should have posted a time anyway! You would have been 3rd or 4th - very respectable. Top 4670k time probably...

Glad you learned more about your RAM though. What speed are you running 9-9-9-27 at? And what sticks are they?

Thanks man. The sticks are Corsair Dom. Platinums rated 1833MHz @ 10-11-10-30 1.5v. Out of the box I think they were running t225, or it might have been more, it was really damn high anyway. At stock voltage all four DIMMs run 1833 9-9-9-27-175 totally stable. The best two DIMMs ran 2133 @ 9-10-9-112 with 1.65v under XP64 and ran Pi fine, but wouldn't load Windows 8 with the same settings. The other pair managed 10-12-10-112 a 1.6something and would be coaxed no higher with 1.65. Both ran command rate at 2T. There was probably more left in them, but I couldn't bring myself to go above 1.65v. Especially as I was basically trawling the web and trying setting that other people had got stable because I had **** all clue myself. At least at the start. I have half a clue now maybe. Best time was 7:34 something, so probably not the fastest 4670K but pretty happy to be close. When I saw people posting 7:2x I knew I had no chance. That 7:18 just baffles my mind.

 

[Rofflay]

10. Raikiri : 7m 57.285s - i5 2500k "Kingston" 1909Mhz 10-11-11-32 1T

I missed this due to forum inactivity so i just set my 3570k multi to a flat 4ghz and ran prime on Windows 7 without this wazza sorcery and got 8m 11.584s on my 2133 kingston kit clocked at 2400mhz 11-12-11-30-1T.

Not bad i think and with a bit of tweaking i could get in at just under 8m.No chance of 7m 40s scores though im not killing my ram at 2600mhz lol.Also without trawling all the pages how the hell did raikiri get that score rofl? And what did Wazza get? I think i can take him but Duffman is out of my league lol.



Cheers and congrats to all

 

[PCZ]

We are not sure how he did it.
Given the memory speed and OS he was using the time he posted is not realistic.

Using a 2500k with much faster memory and lower timings I got down to 8m 03s using XP.


On IVY everyone got stuck at around 8 minutes using W7.
Only when changing OS to a lightweight XP with the right tweaks do the 7 40s become possible.

 

[Duff-Man]

The loot arrived

Thanks, 8 Pack - you're the best!

As I said before, this won't be going to waste. It'll be used to build a new machine to run large-scale scientific simulations (on the storage of CO2 in underground reservoirs mostly). Might need to put 32Gb in the new machine though, which will mean 8Gb sticks - we'll see.

I'll be swapping the CPU over with my current 4770k chip though, since it's a really terrible clocker. Looking forward to seeing what I can squeeze out of this new one! Just need to get some isopropyl swabs first to clean off my heatsink etc.

The CPU cooler is immense... Very impressive indeed (Zalman CNPS12X). Looking at the reviews, it's a toss up whether I'll get better cooling from this, or from my current Corsair H70. I may try both and see for myself


Thanks again for running this competition 8 Pack - I had a blast running pi, learned a lot, and this loot is just the icing on the cake!

 

[setter]

Nice un Duff Man, well done. You put an awful amount of dedication into this thread/competition. I knew i was never gonna be near the top ten, but a thread such as this makes you learn a bit more about your kit and how to get the best out of it. And more so, a great natured thread, no bickering or fanboy banter. Graphics card forum take note.

 

[Subliminal Aura]

Nice one duff.... Thanks for the update. Congrats

 

[1Deep]

Awesome stuff there Duff-Man

I will be going haswell next week btw, managed to sell my ivy gear and just deciding on the mobo as there's so much choice. That gd65 mobo looks real nice though I must say....

ps: That cooler box is HUGE!

 

[setter]

Awesome stuff there Duff-Man

I will be going haswell next week btw, managed to sell my ivy gear and just deciding on the mobo as there's so much choice. That gd65 mobo looks real nice though I must say....

ps: That cooler box is HUGE!

Same here, ill be keeping my sammy green ram and going for a 4770k with the rog maximus hero board.

 

[1Deep]

I would prefer an Asus board as they have always done well for me, never had a problem with them. The maximus hero does look very good, I'm torn between that and the gene, decisions.......

 

[setter]

I would prefer an Asus board as they have always done well for me, never had a problem with them. The maximus hero does look very good, I'm torn between that and the gene, decisions.......

Im a big fan of asus boards too mate, owned so many over the years. The hero looks very similair in layout to my current board. As im running sli i like to keep my audio card above the top gpu. Both cases i have access to are full tower e atx, (lian li a70f/shinobi xl) so the gene would look a bit lost in them.