| View previous topic :: View next topic |
| Author |
Message |
1clue
Advocate
Joined: 05 Feb 2006
Posts: 2569
|
 Posted: Thu Jul 29, 2010 10:40 pm Post subject: |
 |
|
Interesting.
Also maybe relevant is that I have abandoned the traditional approach toward RAM and virtual memory. I have 12G RAM, and 4x6G swap, one on each permanent drive.
Then I have tmpfs on places that make sense. I noticed last time I put Gentoo on that the OS was using unused RAM as disk cache, and with 6G I was almost never hitting swap. So I kind of reversed the roles a bit and tried to make it so with 12G I never swap but I can use tmpfs and disk cache to speed things up. I haven't disabled any swapping, but from what I read I should be able to get significant performance increases that way. It seems to work. |
|
| Back to top |
|
 |
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Fri Jul 30, 2010 2:36 am Post subject: |
|
|
OK, after fixing overcommit and raising -j, we have:
| Code: | Thu Jul 29 18:36:12 2010 >>> app-office/openoffice-3.2.1
merge time: 27 minutes and 19 seconds. |
Compared to 28mins 30sec, that's an astonishingly stunning improvement of 4% with respect to -j6...
When I compiled openoffice last in Jan 2007, it took 230 mins. |
|
| Back to top |
|
|
1clue
Advocate
Joined: 05 Feb 2006
Posts: 2569
|
| Posted: Fri Jul 30, 2010 4:50 am Post subject: |
|
|
Just as a point of reference, my system gets this for recompiling ooo.
real 44m24.235s
user 214m36.098s
sys 29m56.007s
This is at max 2.67 ghz, specs in my signature and absolutely no over-clocking of anything.
I didn't have to download it again, but I was also watching videos off youtube while this was going on. The video is a bit choppy but generally not bad. I attribute that to nspluginwrapper which is now necessary since Adobe blocked all the 64-bit versions. |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Fri Jul 30, 2010 5:02 am Post subject: |
|
|
| Yeah, I had chrome and thunderbird running but nothing CPU intensive like you. |
|
| Back to top |
|
|
1clue
Advocate
Joined: 05 Feb 2006
Posts: 2569
|
| Posted: Fri Jul 30, 2010 5:20 am Post subject: |
|
|
I really don't think video is all that intensive on the CPU with my setup, or for that matter for anything even sort of modern. I got a decent dual-head card with two monitors, but really I was using that as an indicator of full load on the CPU for emerge.
Even though the network was nowhere near saturated, the video was definitely choppy compared to after the build finished. The way I understand it (which could be very wrong) the card handles all the heavy lifting for video, the sound card handles the audio and the CPU doesn't do much but handle a couple interrupts that take almost nothing, but they need to be pretty close for timing. The fact that the video was choppy means (to me anyway) that all the cores were fully engaged in building ooo.
When I have some time I might come back and compare your specs to mine, and find out how much degradation there is for over-clocking.
My RAM is 7-7-7-7 for the low block and 7-7-7-8 for the high block. |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Fri Jul 30, 2010 5:28 am Post subject: |
|
|
| 1clue wrote: | I really don't think video is all that intensive on the CPU with my setup, or for that matter for anything even sort of modern. I got a decent dual-head card with two monitors, but really I was using that as an indicator of full load on the CPU for emerge.
Even though the network was nowhere near saturated, the video was definitely choppy compared to after the build finished. The way I understand it (which could be very wrong) the card handles all the heavy lifting for video, the sound card handles the audio and the CPU doesn't do much but handle a couple interrupts that take almost nothing, but they need to be pretty close for timing. The fact that the video was choppy means (to me anyway) that all the cores were fully engaged in building ooo.
When I have some time I might come back and compare your specs to mine, and find out how much degradation there is for over-clocking.
My RAM is 7-7-7-7 for the low block and 7-7-7-8 for the high block. |
I am currently running with very very relaxed timings on 10-10-10-30 just to test suspend-to-ram issues on my evga board. My RAM supports 8-8-8-24. I think I should be able to do better with 8-8-8-24, although, when I tested other compiles between 10-10-10-30 and 8-8-8-24, I saw less than 2% loss with relaxed timings. |
|
| Back to top |
|
|
1clue
Advocate
Joined: 05 Feb 2006
Posts: 2569
|
| Posted: Fri Jul 30, 2010 5:45 am Post subject: |
|
|
I'm surprised you went with ram that slow, but also surprised that it didn't make that much difference to your results.
I obviously have some more to learn on this. I basically went with the best quality I could afford for the whole system, going for real (official) numbers from reputable vendors rather than over-clocking. In my understanding you get diminishing returns from that, where the devices (RAM for example) take longer to settle the faster you push it, so you can wind up slower than your numbers would dictate.
Well it's time for me to hit the hay.
Good luck and have fun. |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Fri Jul 30, 2010 6:04 am Post subject: |
|
|
| 1clue wrote: | I'm surprised you went with ram that slow, but also surprised that it didn't make that much difference to your results.
I obviously have some more to learn on this. I basically went with the best quality I could afford for the whole system, going for real (official) numbers from reputable vendors rather than over-clocking. In my understanding you get diminishing returns from that, where the devices (RAM for example) take longer to settle the faster you push it, so you can wind up slower than your numbers would dictate.
Well it's time for me to hit the hay.
Good luck and have fun. |
8-8-8-24-1T is very fast! Yes, there are faster modules at CAS 7 but 12GB is tough for CAS7.
All benchmarks pointed to the fact that RAM timings don't gain that much with DDR3. As I said, less than 2% gain in my experience from CAS 10 to 8. That's technically 40 seconds saved on OOO build.... Or may be not if you are purely CPU crunching bound.
My overclocking was slow long process. Learnt a lot during first 1 month. Took it all the way to 4.4Ghz with HT ON and 4.5Ghz with HT OFF. I don't run those clocks because of suspend-to-ram issues on Evga boards when overclocking with vcore adjustments. Suspend-to-ram works for me only if I use Auto vcore. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Fri Jul 30, 2010 11:21 pm Post subject: |
|
|
| Quote: | | Suspend-to-ram works for me only if I use Auto vcore. |
Does it ever not return from suspend? I have this happen every once and a while and I am only running my i7 920 at 3.0Ghz (bclk 150).
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Fri Jul 30, 2010 11:32 pm Post subject: |
|
|
| drescherjm wrote: | | Quote: | | Suspend-to-ram works for me only if I use Auto vcore. |
Does it ever not return from suspend? I have this happen every once and a while and I am only running my i7 920 at 3.0Ghz (bclk 150). |
If I use manual vcore, the 3rd attempt (i.e. 3rd cycle of suspend-resume) to resume will fail. The system just reboots normally. I think this problem is specific to Evga boards.
If I use Auto vcore, then I can suspend and resume any number of times (with my current hardware). Only caveat in there is that some of the temperatures rise every 6 cycles. But that just means that every 6 cycles, I have to suspend the system 2 or 3 times to get the temps down to normal levels. |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Fri Jul 30, 2010 11:40 pm Post subject: |
|
|
I am using auto vcore on my evga 3sli. Suspend works most of the time but I have had 2 to 4 times out of 20 to 30 times I have tried where it did not come back. The fans, hard drives and HD activity light come on (not but no video, no keyboard, no network) so I had to hit the reset button to get out of that.
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Sat Jul 31, 2010 12:23 am Post subject: |
|
|
| drescherjm wrote: | | I am using auto vcore on my evga 3sli. Suspend works most of the time but I have had 2 to 4 times out of 20 to 30 times I have tried where it did not come back. The fans, hard drives and HD activity light come on (not but no video, no keyboard, no network) so I had to hit the reset button to get out of that. |
If video doesn't post, most likely a graphics card issue. What kernel and what driver version are u using? |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
| Posted: Sat Jul 31, 2010 12:52 am Post subject: |
|
|
I was thinking it could be a video issue. Darn it just did not wake up again.
I am now using zen-sources-2.6.34_p1
| Code: | jmd1 ~ # uname -a
Linux jmd1 2.6.34-zen1-sleep #8 ZEN SMP PREEMPT Thu Jul 8 23:34:54 EDT 2010 x86_64 Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz GenuineIntel GNU/Linux
jmd1 ~ # equery l nvidia-drivers
* Searching for nvidiadrivers ...
[IP-] [ ] x11-drivers/nvidia-drivers-195.36.24:0
|
The video card is a fanless nvidia 8400GS card.
| Code: | 02:00.0 VGA compatible controller: nVidia Corporation G98 [GeForce 8400 GS] (rev a1) (prog-if 00 [VGA controller])
Subsystem: ASUSTeK Computer Inc. Device 82b2
Flags: bus master, fast devsel, latency 0, IRQ 16
Memory at de000000 (32-bit, non-prefetchable) [size=16M]
Memory at c0000000 (64-bit, prefetchable) [size=256M]
Memory at dc000000 (64-bit, non-prefetchable) [size=32M]
I/O ports at 9f00 [size=128]
[virtual] Expansion ROM at dffe0000 [disabled] [size=128K]
Capabilities: [60] Power Management version 3
Capabilities: [68] MSI: Enable- Count=1/1 Maskable- 64bit+
Capabilities: [78] Express Endpoint, MSI 00
Capabilities: [100] Virtual Channel
Capabilities: [128] Power Budgeting <?>
Capabilities: [600] Vendor Specific Information: ID=0001 Rev=1 Len=024 < |
_________________
John
My gentoo overlay
Instructons for overlay |
|
| Back to top |
|
|
devsk
Advocate
Joined: 24 Oct 2003
Posts: 3003
Location: Bay Area, CA
|
| Posted: Sat Jul 31, 2010 2:10 am Post subject: |
|
|
| I use the latest stable vanilla-sources and the latest ~amd64 nvidia driver. Seems to work fine with my fan-less 9600GT (I removed the shroud and the noisy fan, put a VGA fan next to it, and my temps dropped). |
|
| Back to top |
|
|
drescherjm
Advocate
Joined: 05 Jun 2004
Posts: 2790
Location: Pittsburgh, PA, USA
|
|
| Back to top |
|
|
MajikC
n00b
Joined: 19 Oct 2003
Posts: 34
Location: Scotland
|
| Posted: Fri Aug 13, 2010 4:08 am Post subject: |
|
|
A bit late adding this but a reason the i7 will not compile much faster with different memory speed will be due to the massive shared CPU cache. A full OOO compile may actually fit completely on the CPU's cache  so the initial move of the files from disk to CPU and the final executables (plus related files) back to disk will use memory... |
|
| Back to top |
|
|
Ormaaj
Guru
Joined: 28 Jan 2008
Posts: 319
|
| Posted: Fri Aug 13, 2010 4:16 pm Post subject: |
|
|
| God it's already getting close to sandy bridge this year. Two year tick-tock is way too short. Surely they can build busses and interfaces with enough overhead to handle two freaking years worth of new processors. They're probably missing out on sales since nobody wants to buy a whole new system just to upgrade their CPU. Maybe if AMD would get off their ass and provide some competition we wouldn't have to deal with so much planned-obsolescence bullshit. |
|
| Back to top |
|
|
DaggyStyle
Watchman
Joined: 22 Mar 2006
Posts: 5941
|
| Posted: Fri Aug 13, 2010 5:26 pm Post subject: |
|
|
| Ormaaj wrote: | | God it's already getting close to sandy bridge this year. Two year tick-tock is way too short. Surely they can build busses and interfaces with enough overhead to handle two freaking years worth of new processors. They're probably missing out on sales since nobody wants to buy a whole new system just to upgrade their CPU. Maybe if AMD would get off their ass and provide some competition we wouldn't have to deal with so much planned-obsolescence bullshit. |
depending for what, the new X6 are giving a good fight for the i7s.
nobody knows how will sandy bridge will perform, on the same page, no one knows how bulldozer will perform, we just need to wait and see.
but usually, price per performance currently, amd is the better choice most of the times.
_________________
Only two things are infinite, the universe and human stupidity and I'm not sure about the former - Albert Einstein |
|
| Back to top |
|
|
krinn
Watchman
Joined: 02 May 2003
Posts: 7470
|
| Posted: Fri Aug 13, 2010 5:34 pm Post subject: |
|
|
| even the jaguar has been update twice time in 5 years, so don't bet on computers for upgrade. |
|
| Back to top |
|
|
TwinGears
n00b
Joined: 30 Sep 2003
Posts: 33
Location: between the chair and the keyboard on and off
|
| Posted: Thu Aug 19, 2010 8:35 pm Post subject: Re: Intel Core i7 - anybody got one? :) |
|
|
| paulbiz wrote: | Anybody got a Core i7? Does it work in Gentoo with any available motherboard? It looks mouth-watering  |
Sure thing, using this CFLAGS without issues
CFLAGS="-march=core2 -mtune=generic -msse4.2 -msahf -O2 -pipe"
it's the slowest of the i7 series but it's been great running Gentoo since I got it....
_________________
Can the mind be still? You have to love a good old fashion classical question.
sip:TwinGears@ekiga.net |
|
| Back to top |
|
|
Ormaaj
Guru
Joined: 28 Jan 2008
Posts: 319
|
| Posted: Fri Aug 20, 2010 4:22 pm Post subject: Re: Intel Core i7 - anybody got one? :) |
|
|
| TwinGears wrote: | | paulbiz wrote: | | Anybody got a Core i7? Does it work in Gentoo with any available motherboard? It looks mouth-watering |
Sure thing, using this CFLAGS without issues
CFLAGS="-march=core2 -mtune=generic -msse4.2 -msahf -O2 -pipe"
it's the slowest of the i7 series but it's been great running Gentoo since I got it.... |
It looks like gcc-4.5.x is doing msse4(.x) with march=native now. No more need to specify it manually. Afaict there has never been a need to specify -msahf at least since I've been on i7 |
|
| Back to top |
|
|
|
Kernel & Hardware
