Data Compression PC

[encode]

Installed the Corsair H60. What can I say - I was just amazed by its performance! Even at lowest fan speed it keeps CPU cold even at full load! Now I limited by motherboard performance only- Z68 mini-ITX by ASRock has some overclocking limitations - the VCore voltage is always auto and I can't change it. During high load the VCore voltage automatically falls down, forcing CPU to lower its frequency, even if CPU is really cold. At 1 core full load it lowers voltage very slow, but at 4 cores fully loaded the voltage falls down really fast...
Anyway, what I've got now is a cold running CPU at 4.6 GHz! But the most important part for me is that now PC is really silent at ANY load!
Check out some pictures! (The first picture is a today's view from my window - November in Russia)

[encode]

Found the reason of throttling. The Chipset/North Bridge tiny heatsink is just extremely hot most of the time! According to forums this is the reason of fast CPU speed drop-down. Will try to install a custom North Bridge cooler...

[encode]

Installed Zalman ZM-NB32K North Bridge cooler. Found that, in fact, I can find anything in Moscow!
Got MUCH more stable overclock! However, throttling still happens after much more longer time though! The rest thing I must do something with is the MOSFET. This Mini-ITX has no MOSFET heatsinks... And such thing is the only left that produce heat...

Check out some pictures of the new cooler and some comparison to pre-installed one and another Zalman North Bridge cooler!

[encode]

Check out some pictures of my PC. Especially take a look at custom motherboard cooling - MOSFETs and Chokes are cooled with Enzotech heatsing, Chipset cooled with active cooler from Deepcool (with Mini Kaze fan from Scythe)

[encode]

Upgraded my case to SUGO SG08. It is the most irrational update ever, I guess. Previously I had SUGO SG07, now I have both. SG08 is much better - it has thick and sexy aluminum front panel, USB 3.0 headers and better PSU fan.

Also I have both Core i7-2600 and Core i7-2600K CPUs. Tested their IGPs on 3DMark Vantage:

Intel Core i7-2600 -> P1417 3DMarks
Intel Core i7-2600K OC -> P2303 3DMarks

Anyway, I just get prepared to Ivy Bridge!

Within a few months I will step by step upgrade my system to:

Corsair Dominator GT/G.Skill Ripjaws X or Z - 2133 MHz or 2400 MHz CL9 memory (two 4 or 8 GB modules)

ASRock H77M-ITX or Z77 ITX from ASUS. In addition, EVGA designing some very interesting - enthusiast Z77 ITX board.

Main reason for upgrading MoBo - current ITX boards have weak power circuit. New ASRock and ASUS boards look great - fully featured CPU power design (4+4 pin) and nice heatsinks. As a note, my current Z68 ITX board use just one 4 pin for CPU power.

Intel Core i7-3770K (Ivy Bridge)

It's much like a Core i7-2700K developed in 22nm tech process. 22nm means it's much cooler, power efficient and must have great OC potential!

Most likely I'll get Gainward GeForce GTX 680 4GB

480 GB SSD Corsair Force GT

[Karhunen]

Any thoughts on Heat Pipes ? Too bad mineral oil is not sufficient to cool current CPUs, I prefer simple passive cooling methods.

[encode]

The best way to go is water/liquid cooling. Even simple Corsair H60 performs on par with largest air coolers. With corsair H60 I can even get 40C on load (56C on 100% load of 4.6 GHz CPU) and about room temps on idle. Even small heatsink with a tiny silent fan will outperform large passive one. Confirmed that with my chipset cooling...

[PAQer]

Intel Core i7-3770K (Ivy Bridge)

It's much like a Core i7-2700K developed in 22nm tech process. 22nm means it's much cooler, power efficient and must have great OC potential!

Bad news.
http://www.overclockers.ru/hardnews/...o_95_vatt.html

[Piotr Tarsa]

Ridiculous. They've touted their "3D" tri-gate 22nm process as groundbreaking, when it comes to power efficiency and now they behave like it's just the other way about. Either we're hitting the power efficiency wall or Intel is just lazy.

[Nova]

Here's an early benchmark,though it has nothing about OC.Ivy Bridge is rumored to launch on April 23,guess you have to wait till then for OC benchmarks and heat dissipation.
http://www.anandtech.com/show/5626/i...-core-i7-3770k

[encode]

Bad news indeed... Anyway, within a week we will see the real results, not some rumors. It reminds me rumors about new super fast RAMBUS memory in Radeon HD 7900 series. Finally on retail boxes we read just DDR5, as previously!
If 3D transistors and Ivy Bridge will fail, as in latest rumors...

I hope that 95W came from powerful IGP onboard!

Anyway, AMD planned to release some stunning stuff as well. The Trinity APUs - high clocked Bulldozer with Radeon HD 7000 series onboard is not a joke. I think even if Sandy Bridge had a better IGP, to play, say Battlefield 2, I can live without discrete graphics card at all. I realized that I'm not an extreme gamer, but for casual gaming I need something faster that current i7 and even Llano APUs deliver.

Will wait and see what to pack inside my SUGO SG08...

One thing I can tell for sure. Companies cannot deliver new stunning technologies every year! For CPUs and such we must wait at least 5 years to say we have something new and it's much better than previous hardware. Generations passing thru.

[encode]

Ivy Bridge sells with 95W TDP, but uses a maximum of 77W

[Black_Fox]

Yes, official TDP is the same, real will be lower. The 3D tech is better, but there are rumors circling that starting from ~4.6 GHz it is hard to cool the new Ivy Bridge even with Corsair H100.
http://forums.aria.co.uk/showthread....w*-56K-WARNING
http://i1095.photobucket.com/albums/...ak/orhy1gM.png

[encode]

I hope it's just rumors... Or else I get Z77 board and i7-2700K. I'm currently have 4.6 GHz with ITX MoBo, 4.7 is possible, but unstable. And this is just with H60...

[encode]

Looks like ASUS P8Z77-I DELUXE is the best! These blue heatsinks are questionable - I'd prefer ROG styled black and red one, but specs are truly awesome!
ASRock Z77E-ITX looks better, but has weaker specs...

[encode]

SO many overclockers refer to Super PI 1M test. Decided to test my system.

4.3 GHz - 8.689 sec
4.4 GHz - 8.518 sec
4.5 GHz - 8.315 sec
4.6 GHz - 8.143 sec
4.7 GHz - 7.945 sec

[Matt Mahoney]

Why does superpi.exe need admin access?

1M = 32.339 sec.
Processor: Intel Pentium Dual CPU T3200 2000 MHz
Core: Merom M0
Motherboard: Gateway M7301-U
Memory: 3072 MB DDR2
Clock: 333 MHz
Timings: 5-5-5-15

[encode]

Why does superpi.exe need admin access?

My guesses, because it scans the hardware. But you can ask the author: http://www.superpi.net

[encode]

Continue upgrading.

Now I have Corsair H80 water cooler - really cool thing - much better than H60! Thick radiator, fan controller on the top of a pump...

Decided to not go with Ivy Bridge and upgraded my CPU to Intel Core i7-2700K! Well, at the same speed it has the same performance as 2600K. But, they behave differently at overclock - my 2600K @ 4.5 GHz with Prime95 instantly throttles down to 4.0 GHz and then to 3.4 GHz, shortly. @ 4.6 GHz it hangs the system. But at modest load I even able to run super PI at 4.7 GHz!
2700K doesn't throttle! It just hangs the system at insufficient power - Yes my current motherboard has weak power design! So, 4.6 GHz is the max overclock currently. @ 4.7 it freezes/hangs each time. But @ 4.5 GHz 2700K is is really stable with Prime95! Sweet!

My 2600K was made in Costa Rica, and 2700K was made in Malaysia.

Waiting for G.Skill ARES 2133 MHz CL9 2x4GB memory - delivery delayed...

And of course I'm looking forward for ASUS P8Z77-I DELUXE - to bring more power for overclock to 2700K!

[encode]

Upgraded RAM.

So, how memory speed affects compression time?

I performed this test with:

Kingston 1333 MHz CL9 2x4 GB
Corsair Dominator GT 1866 MHz CL9 2x4 GB
G.Skill ARES 2133 MHz CL9 2x4 GB

Intel Core i7-2700K @ 4.6 GHz water cooled with Corsair H80
240 GB SSD Corsair Force GT

Kingston 1333 MHz CL9 2x4 GB

C:\>timer lzma e enwik9 enwik9.z -d27

Timer 3.01 Copyright (c) 2002-2003 Igor Pavlov 2003-07-10

LZMA 9.20 : Igor Pavlov : Public domain : 2010-11-18

Kernel Time = 1.185 = 00:00:01.185 = 0%
User Time = 802.453 = 00:13:22.453 = 99%
Process Time = 803.639 = 00:13:23.639 = 99%
Global Time = 804.716 = 00:13:24.716 = 100%


Corsair Dominator GT 1866 MHz CL9 2x4 GB

C:\>timer lzma e enwik9 enwik9.z -d27

Timer 3.01 Copyright (c) 2002-2003 Igor Pavlov 2003-07-10

LZMA 9.20 : Igor Pavlov : Public domain : 2010-11-18

Kernel Time = 0.748 = 00:00:00.748 = 0%
User Time = 708.759 = 00:11:48.759 = 99%
Process Time = 709.508 = 00:11:49.508 = 99%
Global Time = 710.834 = 00:11:50.834 = 100%


G.Skill ARES 2133 MHz CL9 2x4 GB

C:\>timer lzma e enwik9 enwik9.z -d27

Timer 3.01 Copyright (c) 2002-2003 Igor Pavlov 2003-07-10

LZMA 9.20 : Igor Pavlov : Public domain : 2010-11-18

Kernel Time = 0.686 = 00:00:00.686 = 0%
User Time = 686.248 = 00:11:26.248 = 99%
Process Time = 686.934 = 00:11:26.934 = 99%
Global Time = 688.354 = 00:11:28.354 = 100%

[Piotr Tarsa]

Compression is one of few things casuals do and which are sensitive to memory speed. Could you also test PAQ and PPMd? 7-Zip is covered by CPU reviews, but (rather) obviously they don't use PAQ for benchmarking.

[encode]

Done!

[encode]

After a couple of days testing, I consider that 2700K is better overclocker. Now I get 4.6 GHz stable, Prime95 testing passed - no throttling, 4.7 GHz is semi-stable - but I able to run many instances of BCM optimizers at such CPU clocks. Looks like thermal compound started working (Arctic Silver 5) or chip warmed up, or something. The results then I installed a brand new CPU was slightly different - 4.6 was not stable, 4.7 was nearly unreachable...

Still, no ASUS P8Z77-I DELUXE available in Russia...

[Bulat Ziganshin]

http://www.citilink.ru/catalog/parts...boards/669834/
http://www.regard.ru/catalog/tovar104973.htm
http://xpert.ru/products.php?getProduct=111212&ajax=1

[encode]

P8Z77-I is a miniITX motherboard. Tons of Z77 ATX/mATX available for a long time...

[encode]

The funniest thing - 3770K is available and it's cheaper than 2700K... Well, will wait for more results and for heat issue fix, if it's possible. Rumors saying we may expect 6-core LGA1155 CPU. That can be more interesting...

[encode]

2700K IGP performance hits P3152 3DMarks - slightly less than 3770K!
http://3dmark.com/3dmv/4064544
I guess 2700K is just better in many ways than that new 3770K...

[Piotr Tarsa]

From what I read recently, issues with Ivy Bridge comes from power density - Ivy Bridge die is very small, almost 2x smaller than Sandy Bridge thanks to process shrink, but at the same time power usage didn't fell twice. Because of that, it's very hard to dissipate the heat that comes from such a small die. If that's true then maintaining single-thread performance on transition to 14nm could be impossible, let alone increasing the performance.

[Bulat Ziganshin]

Piotr, it's the usual problem of first generation of chips, assembled using new technology norms. it was the same with first 32 nm devices, 45 nm devices and so on. just wait for the new steppings. i'm sure that haswell will be fine as sandy was. the only real issue may be increasing of GPU part, but i hope it doesn't count when we use external video card

[Piotr Tarsa]

I wouldn't be so sure. They (Intel) projected Pentium IV's to have 10 GHz clock speed, but as we all know, they failed because of power density issues. Now power density strikes back. They had to lower clock speed from Pentium IV to Core 2, but fortunately they substantially improved IPC. Now IPC is relatively high and x86 decoding is rather complex enough that increasing IPC further would be too costly to be profitable.

The next logical step would be to lower clocks somewhat and increase the number of cores. But cache coherence in x86 architecture requirements impose serious overhead with many (real?) cores, ie over about 20 cores.

I'm not saying that this will necessarily take place in next Tick or Tock, but there's somewhat high probability.