• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Haswell overclocking discussion.

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
ehume said:
So I learned that to test this beast I must disable HT.
Click to expand...
Why do you disable HT versus just launching 4 threads instead of 8?

You'll get better GFlops, and higher heat output (the stress test part of the test), if you leave HT enabled so that all the background tasks in windows get handled by the virtual cores without intruding on the work being done by the linpack threads on the "real" cores.
 
Idontcare said:
Why do you disable HT versus just launching 4 threads instead of 8?

You'll get better GFlops, and higher heat output (the stress test part of the test), if you leave HT enabled so that all the background tasks in windows get handled by the virtual cores without intruding on the work being done by the linpack threads on the "real" cores.
Click to expand...

I was sternly lectured in another thread that Linpack is limited to one thread per core. There are links that report the results of testing, where multi-threading usually slowed processes down.

I started that thread because when I ran Linpack 11.0.5.009 with its batch file, it ran at 100 Gflops (default cpu speed). LinX with Linpack 11.0.5.009 ran one rep at 160Gflops, then stopped on error.

Watching in real time with TMonitor I could see Linpack stop using half the threads during the middle of a run. When I disabled hyperthreading, the Gflops on the default speed rose to 189 Gflops.

When I ran normal LinX (presumably pre-Linpack 10; no AVX) I was able to OC the chip to 4.7 GHz. When I ran LinX with Linpack 10 (has AVX), I had to back off the OC, down to 4.4-4.5 GHz. With Linpack 11 (AVX2), I'm down to 4.3 GHz.

Since the HT issue is a Linpack problem, I can turn HT back on for normal use.
 
So have you sorted the issue?

Here is my chip at 4.6GHz:

PHP: 
Intel(R) Optimized LINPACK Benchmark data
Current date/time: Mon Jul 08 07:15:02 2013
CPU frequency:    4.597 GHz
Number of CPUs: 1
Number of cores: 4
Number of threads: 4
Parameters are set to:
Number of tests: 3
Number of equations to solve (problem size) : 15000 20000 25000
Leading dimension of array                  : 15000 20000 25000
Number of trials to run                     : 10    7     5    
Data alignment value (in Kbytes)            : 4     4     4    
Maximum memory requested that can be used=705536800, at the size=25000
=================== Timing linear equation system solver ===================
Size   LDA    Align. Time(s)    GFlops   Residual     Residual(norm) Check
15000  15000  4      11.752     191.5010 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9905 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9879 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.703     192.2953 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.687     192.5559 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.716     192.0799 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.717     192.0665 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.812     190.5160 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.763     191.3084 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.719     192.0351 1.979849e-010 3.118294e-002   pass
20000  20000  4      24.100     221.3305 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.993     222.3244 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.976     222.4797 3.564028e-010 3.154945e-002   pass
20000  20000  4      24.165     220.7346 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.982     222.4211 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.982     222.4258 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.971     222.5265 3.564028e-010 3.154945e-002   pass
25000  25000  4      47.860     217.6767 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.688     218.4581 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.681     218.4915 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.732     218.2592 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.684     218.4779 6.220813e-010 3.537553e-002   pass
Performance Summary (GFlops)
Size   LDA    Align.  Average  Maximal
15000  15000  4       191.8337 192.5559
20000  20000  4       222.0347 222.5265
25000  25000  4       218.2727 218.4915
Residual checks PASSED
End of tests
Mon 07/08/2013 
07:29 AM
 
Very impressive, Balla. :thumbsup:

1/5th of a terraflop. Progress is amazing...

BallaTheFeared said:
So have you sorted the issue?

Here is my chip at 4.6GHz:

PHP: 
Intel(R) Optimized LINPACK Benchmark data
Current date/time: Mon Jul 08 07:15:02 2013
CPU frequency:    4.597 GHz
Number of CPUs: 1
Number of cores: 4
Number of threads: 4
Parameters are set to:
Number of tests: 3
Number of equations to solve (problem size) : 15000 20000 25000
Leading dimension of array                  : 15000 20000 25000
Number of trials to run                     : 10    7     5    
Data alignment value (in Kbytes)            : 4     4     4    
Maximum memory requested that can be used=705536800, at the size=25000
=================== Timing linear equation system solver ===================
Size   LDA    Align. Time(s)    GFlops   Residual     Residual(norm) Check
15000  15000  4      11.752     191.5010 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9905 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9879 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.703     192.2953 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.687     192.5559 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.716     192.0799 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.717     192.0665 1.979849e-010 3.118294e-002   pass
Click to expand...
 
ehume said:
I was sternly lectured in another thread that Linpack is limited to one thread per core. There are links that report the results of testing, where multi-threading usually slowed processes down.

I started that thread because when I ran Linpack 11.0.5.009 with its batch file, it ran at 100 Gflops (default cpu speed). LinX with Linpack 11.0.5.009 ran one rep at 160Gflops, then stopped on error.

Watching in real time with TMonitor I could see Linpack stop using half the threads during the middle of a run. When I disabled hyperthreading, the Gflops on the default speed rose to 189 Gflops.

When I ran normal LinX (presumably pre-Linpack 10; no AVX) I was able to OC the chip to 4.7 GHz. When I ran LinX with Linpack 10 (has AVX), I had to back off the OC, down to 4.4-4.5 GHz. With Linpack 11 (AVX2), I'm down to 4.3 GHz.

Since the HT issue is a Linpack problem, I can turn HT back on for normal use.
Click to expand...
Running the batch file on a quad core CPU with HT activated, 2 physical cores and 2 HT-cores will get threads assigned to them. I've modified the batch file so it assignes the threads to the physical cores. Just change the contents of "runme_xeon64.bat" to the following:
Code: 
@echo off
SETLOCAL

rem Setting up affinity for better threading performance
set KMP_AFFINITY=nowarnings,compact,granularity=fine
rem Setting path to OpenMP library
set PATH=..\..\..\redist\intel64\compiler;%PATH%

echo Running linpack_xeon64.exe. Output can be found in win_xeon64.txt.
start /b /affinity 55 linpack_xeon64.exe lininput_xeon64 > win_xeon64.txt

echo.
echo When this window closes the calculation is done.

ENDLOCAL
The important piece was adding "start /b /affinity 55", the rest is mostly cosmetical.
 
Last edited:
it may be more of a Motherboard related issue than the CPU itself. i don't suppose the CPUs will change that much over time while M/B will have their bioses tweaked and so on.
 
I think some of the board vendors have it working, mine doesn't work though :|

Not that I have a lot to complain about though, platform feels mature outside of that no major quarks that I've experienced yet.
 
BallaTheFeared said:
So have you sorted the issue?

Here is my chip at 4.6GHz:

PHP: 
Intel(R) Optimized LINPACK Benchmark data
Current date/time: Mon Jul 08 07:15:02 2013
CPU frequency:    4.597 GHz
Number of CPUs: 1
Number of cores: 4
Number of threads: 4
Parameters are set to:
Number of tests: 3
Number of equations to solve (problem size) : 15000 20000 25000
Leading dimension of array                  : 15000 20000 25000
Number of trials to run                     : 10    7     5    
Data alignment value (in Kbytes)            : 4     4     4    
Maximum memory requested that can be used=705536800, at the size=25000
=================== Timing linear equation system solver ===================
Size   LDA    Align. Time(s)    GFlops   Residual     Residual(norm) Check
15000  15000  4      11.752     191.5010 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9905 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.722     191.9879 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.703     192.2953 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.687     192.5559 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.716     192.0799 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.717     192.0665 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.812     190.5160 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.763     191.3084 1.979849e-010 3.118294e-002   pass
15000  15000  4      11.719     192.0351 1.979849e-010 3.118294e-002   pass
20000  20000  4      24.100     221.3305 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.993     222.3244 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.976     222.4797 3.564028e-010 3.154945e-002   pass
20000  20000  4      24.165     220.7346 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.982     222.4211 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.982     222.4258 3.564028e-010 3.154945e-002   pass
20000  20000  4      23.971     222.5265 3.564028e-010 3.154945e-002   pass
25000  25000  4      47.860     217.6767 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.688     218.4581 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.681     218.4915 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.732     218.2592 6.220813e-010 3.537553e-002   pass
25000  25000  4      47.684     218.4779 6.220813e-010 3.537553e-002   pass
Performance Summary (GFlops)
Size   LDA    Align.  Average  Maximal
15000  15000  4       191.8337 192.5559
20000  20000  4       222.0347 222.5265
25000  25000  4       218.2727 218.4915
Residual checks PASSED
End of tests
Mon 07/08/2013 
07:29 AM
Click to expand...

Could you post uncore en mem frequency of this overclock? Iam getting similar results with 4.4 running 1:1 with 1866 ram.

I added 44000 to the test to max out my ram:

Code: 
Intel(R) Optimized LINPACK Benchmark data

Current date/time: Sat Jul 13 19:41:34 2013

CPU frequency:    4.398 GHz
Number of CPUs: -1
Number of cores: 1
Number of threads: 4

Parameters are set to:

Number of tests: 13
Number of equations to solve (problem size) : 1000  2000  3000  4000  5000  10000 15000 20000 25000 30000 35000 40000 44000
Leading dimension of array                  : 1000  2000  3000  4000  5000  10000 15000 20000 25000 30000 35000 40000 44000
Number of trials to run                     : 4     4     4     4     4     2     2     2     2     1     1     1     1    
Data alignment value (in Kbytes)            : 4     4     4     4     4     4     4     4     4     4     4     4     4    

Maximum memory requested that can be used=2603982208, at the size=44000

=================== Timing linear equation system solver ===================

Size   LDA    Align. Time(s)    GFlops   Residual     Residual(norm) Check
1000   1000   4      0.022      30.6667  1.194739e-012 4.074366e-002   pass
1000   1000   4      0.005      129.1783 1.194739e-012 4.074366e-002   pass
1000   1000   4      0.005      128.9536 1.194739e-012 4.074366e-002   pass
1000   1000   4      0.005      138.9880 1.194739e-012 4.074366e-002   pass
2000   2000   4      0.039      136.5982 4.536926e-012 3.946570e-002   pass
2000   2000   4      0.038      141.8822 4.536926e-012 3.946570e-002   pass
2000   2000   4      0.038      141.7703 4.536926e-012 3.946570e-002   pass
2000   2000   4      0.038      140.8703 4.536926e-012 3.946570e-002   pass
3000   3000   4      0.117      154.3954 8.334888e-012 3.209566e-002   pass
3000   3000   4      0.119      151.3027 8.334888e-012 3.209566e-002   pass
3000   3000   4      0.114      157.9014 8.334888e-012 3.209566e-002   pass
3000   3000   4      0.114      157.8114 8.334888e-012 3.209566e-002   pass
4000   4000   4      0.254      167.8629 1.519912e-011 3.312792e-002   pass
4000   4000   4      0.252      169.5147 1.519912e-011 3.312792e-002   pass
4000   4000   4      0.251      169.7998 1.519912e-011 3.312792e-002   pass
4000   4000   4      0.252      169.4740 1.519912e-011 3.312792e-002   pass
5000   5000   4      0.472      176.4859 2.471656e-011 3.446525e-002   pass
5000   5000   4      0.470      177.5435 2.471656e-011 3.446525e-002   pass
5000   5000   4      0.471      177.0446 2.471656e-011 3.446525e-002   pass
5000   5000   4      0.470      177.3364 2.471656e-011 3.446525e-002   pass
10000  10000  4      3.313      201.2790 9.436774e-011 3.327502e-002   pass
10000  10000  4      3.283      203.1198 9.436774e-011 3.327502e-002   pass
15000  15000  4      10.584     212.6284 2.169435e-010 3.416896e-002   pass
15000  15000  4      10.585     212.6036 2.169435e-010 3.416896e-002   pass
20000  20000  4      24.108     221.2601 3.504283e-010 3.102058e-002   pass
20000  20000  4      24.235     220.1044 3.504283e-010 3.102058e-002   pass
25000  25000  4      47.111     221.1354 5.194889e-010 2.954147e-002   pass
25000  25000  4      47.251     220.4802 5.194889e-010 2.954147e-002   pass
30000  30000  4      80.207     224.4427 7.133177e-010 2.811906e-002   pass
35000  35000  4      125.535    227.7118 1.085449e-009 3.150893e-002   pass
40000  40000  4      187.353    227.7507 1.338275e-009 2.976370e-002   pass
44000  44000  4      251.677    225.6589 1.851860e-009 3.409328e-002   pass

Performance Summary (GFlops)

Size   LDA    Align.  Average  Maximal
1000   1000   4       106.9467 138.9880
2000   2000   4       140.2803 141.8822
3000   3000   4       155.3527 157.9014
4000   4000   4       169.1629 169.7998
5000   5000   4       177.1026 177.5435
10000  10000  4       202.1994 203.1198
15000  15000  4       212.6160 212.6284
20000  20000  4       220.6822 221.2601
25000  25000  4       220.8078 221.1354
30000  30000  4       224.4427 224.4427
35000  35000  4       227.7118 227.7118
40000  40000  4       227.7507 227.7507
44000  44000  4       225.6589 225.6589

Residual checks PASSED

End of tests
By following discussion in this thread i had to ditch my 4.5 and 4.6 aida/prime stable profiles (thanks guys:twisted🙂, because they couldnt pass latest linpack. Iam now 4.4 linpack stable and will try to shoot for more.
 
Last edited:
4.6/4.6 2400MHz on the ram.

Maybe there is something wrong with my chip/board?

What do you get in Cinebench?

9279802004_dd98e5a631_o.png


7.9 is at 4.6GHz

8.58 is at 5GHz cpu 5GHz cpuNB
 
Just realized you have 4670k, those results are of 4770k. Strangely, at these clocks not so huge difference in gflops but 1 point in cinebench:

141001_cinebench.PNG
 
Can someone do me a favor and post there entire batch file. It seems like I can't get the newest linx to run all cores and ht cores.
 
You can't. Linpack will run faster if you disable HT -- HT slows it down. The file you want to change is the data input file. I can't recall the name. It has no suffix.
 
I've built four rigs with 4770Ks over the last few weeks and all have had no problems reaching 4.5 -4.6ghz @1.28 -1.30v .Motherboards I've built with so far are Asus Maximus VI Hero, Gigabyte GA-Z87X-UD3H, and Gigabyte GA-Z87X-OC. All boards have been very basic to OC with just a change of the multi and vcore adjustment.
Compared to Ivy bridge the only negative is the well documented higher load temps which is very noticeable, almost dramatic difference at 100% load even with corsair H100s and CM Seidon 240s. We are in winter in New Zealand , the real test will be this summer and if I receive zero heat related emails from customers..

Just built a Haswell system today with a GA-Z87X-UD3H, 2 EVGA 4GB gtx 670 SCs , 32GB 2400 Kingston Beast etc in a Coolermaster Storm stryker , currently running the usual barrage of tests as I type this , offtopic but damn this case is brilliant and feature rich for the $ , cooling is top notch though not quite up to an Ft02 or raven rv03, heres a pick..

ekvp.jpg
 
Last edited:
NTMBK said:
3/5, needs more monitors
Click to expand...

Ha! The two on the right are linked up to customers Storm stryker , the other 4 are linked to my rig in the background. I work a lot with Korean moniters so normaly have a few around..

What motivated me to bother switching from 3770K to 4770K was the prospect of having a mobo with all 6GBs ports, and an ever increasing Steam collection. Did consider just getting a 500GB SSD but it seemed more interesting to try a few different SSDs and compare them. So far a Samsung 840 and Vertex 3 Max Iops have joined the Sandisk extreme and Intel 520 plus a second 2TB mechanical for storage. Long story short if your looking for storage expansion then 1150 makes a lot of sence.

As for socket 1150 boards.. so far the Asus Maximus Hero has ticked the most boxes out of the boards Ive built with.. the UD3H and Z87X Overclock are rock solid as well and would recommend them to clients depending on budget and requirements however when it came to my own rig the Maximus Hero was the winner as it offers a lot of value for the money, a bonus is that the onboard sound on the hero was good enough to sell off my Asus xonar DX2 which reimbursed a nice chunk of the motherboard cost. Am keen to do a few builds with Asrock Z87s in the upcoming weeks and compare..
 
Last edited:
guskline said:
Piklar, what type of cpu cooler are you using for a 4770k OC'd to 4.5-4.6Ghz?
Click to expand...

The two 4770K systems I have here at the moment are both running 4.6ghz, the difference being the Stryker build requires 1.30v and the Raven build requires 1.28v to achieve a rock solid 4.6ghz .

The raven has a Corsair H70 with push/pull bitfenix spectre Pro fans mounted to the top exhaust like this -

mylw.jpg



The second PC has a Coolermaster Seidon 120XL with push/pull bitfenix spectres set up to exhaust out the back -

6kvh.jpg


The two previous builds before these were both in Corsair 500R cases had H100i and a Seidon 240 fitted to the top exhaust. The Seidon had a tiny edge in temps over the H100i by a couple of degrees, both 240 solutions brought load temps down by 4-8C compared to their 120mm offerings but these two current cases (especialy the raven) seem to be getting away with just 120mm rads on 4.6ghz 4770s because of their superior airflow..
 
Last edited:
Piklar, thank you for the info. Those are gorgeous builds.

BTW, what software do you use to check for stability?
 
I wish I had read this thread earlier. Someone mentioned that the OP should use Prime95 because the temps seem low. Running Aida64 with only stress FPU selected will do a better job than Prime95 and temps will also be higher. You may also need to bump up the vcore a notch or two.
 
guskline said:
Piklar, thank you for the info. Those are gorgeous builds.

BTW, what software do you use to check for stability?
Click to expand...

Thanks 🙂 Gotta luv a good build in winter!

Usually allow 48hours to stress test a decent project build, typical software used are Prime95 , Intelburntest , Unigine , 3Dmark , FurMark, CrystalDiskmark.
Monitoring software- Speedfan, GPUz, CPUz, Taskmanager.
Diagnostics software- Whocrashed , Eventviewer..
 
Last edited:
BallaTheFeared said:
4.6/4.6 2400MHz on the ram.

Maybe there is something wrong with my chip/board?

What do you get in Cinebench?

9279802004_dd98e5a631_o.png


7.9 is at 4.6GHz

8.58 is at 5GHz cpu 5GHz cpuNB
Click to expand...

My 1st time using Cinebench, 1st attempt is not any better than yours. And my CPU dropped from 4.6ghz to as low as 3.7ghz during the CPU test? and the Vcore increased from 2.8v to 2.96 which suggest the CPU needed the extra juice under load but the temps stayed the same, quite odd.. Took a screenie of it while in action..

6zjb.jpg



Screenshot while running OpenGL, notice CPU is back to its 24/7 clocks of 4.6ghz @2.80v

lbr3.jpg


Anyone know why the drop in CPU frequency while running the CPU test?
 
Last edited:
You must log in or register to reply here.

TRENDING THREADS

Back
Top