• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

i7 920.. The Sweet Spot (Optimal OC)

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

deathman20

High Speed Premium Senior
Joined
Aug 5, 2002
Well over the past few days there was a post over at XS talking about CPU speed, games, and GPU performance. Trying to find that sweet spot in the CPU and system for optimal OC's. So decided to give it a little trial run. What I typically run at is 3.8Ghz with HT on. Mind you that HT doesn't seem to help me to much currently. Not in my 3D Rendering state so its just chewing extra power at the moment.

Still sure extra CPU power is more useful. Heck I could really see it with my old E8400 and dual 4850's, but times change. But now im in a more conservative state. So for majority of my time, going ot run it at different speed and settings. Still trying to get the most of my system but at less heat output (nice for the summer months) yet still retain that snappy feel.

Here is my system specs that I've tried.

My specs for 3.8Ghz is (Default Clocks)
Code:
Frequency/Voltage Control	
Dummy O.C	Disabled
Extreme Cooling	Disabled
CPU Clock Ratio	20x
CPU Host Frequency(Mhz)	190
MCH Strap	Auto
CPU Uncore Frequency(Mhz)	16
Spread Spectrum	Disabled
PCIE Frequency(Mhz)	100
	
Memory Feature	
Memory Control Setting	Enabled
Memory Frequency	2:8
Channel Intreleave Setting	6 way
Rank Interleave Setting	4 way
Memory Low Gap	Auto
**Parameters**	
tCL Setting	8
tRCD Setting	8
tRP Setting	8
tRAS Setting	21
tRFC Setting	74
Command Rate	1T
tRRD Setting	Auto
tFAW Setting	Auto
tRTP Setting	Auto
tWR Setting	Auto
Round  Trip Latency	Auto
tddWrTRd Setting	Auto
tdrWrTRd Setting	Auto
tsrWrTRd Setting	Auto
tddRdTWr Setting	Auto
tdrRdTWr Setting	Auto
tsrRdTWr Setting	Auto
tddRdTRd Setting	Auto
tdrRdTRd Setting	Auto
tsrRdTRd Setting	Auto
tddWrTWr Setting	Auto
tdrWrTWr Setting	Auto
tsrWrTWr Setting	Auto
	
Voltage Control	
EVGA Vdroop Control	W/O Vdroop
CPU Vcore	1.225v
CPU VTT Voltage	+100mV
CPU PLL Vcore	1.8v
Dimm Voltage	1.60v
DimmDQ Vref	+0mV
QPI PLL Vcore	1.15v
IOH Vcore	1.15v
IOH/ICH I/O Voltage	1.55v
ICH Vcore	1.1v
PWM Frequency	800KHz
	
CPU Feature	
Intel Speadstep	Disabled
Turbo Mode Function	Disabled
CxE Function	Enabled
Execute Disable Bit	Enabled
VirtualizationTechnology	Enabled
**Logical Processor Setting**	
Intel HT Technology	Enabled
Active Processor Cores	All
**QPI Controller Setting**	
QPI Control Setting	Enabled
QPI Link Fast Mode	Enabled
QPI Frequency Selection	Auto

Just this weekend I tried something for fun and went to 3.2Ghz (Trial Settings, Stable)
Code:
Frequency/Voltage Control	
Dummy O.C	Disabled
Extreme Cooling	Disabled
CPU Clock Ratio	16x
CPU Host Frequency(Mhz)	200
MCH Strap	Auto
CPU Uncore Frequency(Mhz)	16
Spread Spectrum	Disabled
PCIE Frequency(Mhz)	100
	
Memory Feature	
Memory Control Setting	Enabled
Memory Frequency	2:8
Channel Intreleave Setting	6 way
Rank Interleave Setting	4 way
Memory Low Gap	Auto
**Parameters**	
tCL Setting	8
tRCD Setting	8
tRP Setting	8
tRAS Setting	21
tRFC Setting	74
Command Rate	1T
tRRD Setting	Auto
tFAW Setting	Auto
tRTP Setting	Auto
tWR Setting	Auto
Round  Trip Latency	Auto
tddWrTRd Setting	Auto
tdrWrTRd Setting	Auto
tsrWrTRd Setting	Auto
tddRdTWr Setting	Auto
tdrRdTWr Setting	Auto
tsrRdTWr Setting	Auto
tddRdTRd Setting	Auto
tdrRdTRd Setting	Auto
tsrRdTRd Setting	Auto
tddWrTWr Setting	Auto
tdrWrTWr Setting	Auto
tsrWrTWr Setting	Auto
	
Voltage Control	
EVGA Vdroop Control	W/O Vdroop
CPU Vcore	1.175v
CPU VTT Voltage	+100mV
CPU PLL Vcore	1.8v
Dimm Voltage	1.60v
DimmDQ Vref	+0mV
QPI PLL Vcore	1.15v
IOH Vcore	1.15v
IOH/ICH I/O Voltage	1.55v
ICH Vcore	1.1v
PWM Frequency	800KHz
	
CPU Feature	
Intel Speadstep	Disabled
Turbo Mode Function	Disabled
CxE Function	Enabled
Execute Disable Bit	Enabled
VirtualizationTechnology	Enabled
**Logical Processor Setting**	
Intel HT Technology	Enabled
Active Processor Cores	All
**QPI Controller Setting**	
QPI Control Setting	Enabled
QPI Link Fast Mode	Enabled
QPI Frequency Selection	Auto

After all said and done I dropped roughly 10C, the system is stable and voltage might be able to go lower... and out of all of it, from one game I did test it dropped 5FPS average (75FPS to 70FPS or a 7% drop for a 18% decrease in CPU power). Now the min FPS was affected the most on the test (sorry don't have it with me right now) but actual game play so far was solid still no noticeable difference. The game was a benchmark run on X3:TC which in itself is considerably demanding in both GPU and CPU, though its not heavily multi threaded. So if not mistaken its relaying heavily on 1-2 CPU cores.

Added note to above instead of hovering in the mid 60's for gaming now its in the mid to lower 50's with ambient around 23-25C.

Ok so its good here at this speed. I still have HT turned on. For a few things I do semi regularly (compiling videos to wmv) it would be nice to have still a slightly faster processor speed which it did fall % wise as much as the CPU came down in speed. So now thinking I might bump the multi to 17x for for 3.4Ghz and turn off HT. This will give me a mild bump (roughly 6%), still having the same settings for ram (maxing my rams default speed out) and might be able to still drop the power levels even more. Course to note even at 1.175V at 3.2Ghz its already below my VID for my CPU which is nice.

Now I don't typically have a lot of time to test things out hence why limited number of tests (just did X3:TC). But just giving people a little more to consider if they want to tweak there OC for bang for the performance instead of running rigs to the edge all the time. Course its the OC forums and heck would of kept it going strong but this just intrigued me from power/heat standpoint of why not try it out.

More results will come...
 
Did a 3.4Ghz run as well now with no HT
Code:
Frequency/Voltage Control	
Dummy O.C	Disabled
Extreme Cooling	Disabled
CPU Clock Ratio	17x
CPU Host Frequency(Mhz)	200
MCH Strap	Auto
CPU Uncore Frequency(Mhz)	16
Spread Spectrum	Disabled
PCIE Frequency(Mhz)	102
	
Memory Feature	
Memory Control Setting	Enabled
Memory Frequency	2:8
Channel Intreleave Setting	6 way
Rank Interleave Setting	4 way
Memory Low Gap	Auto
**Parameters**	
tCL Setting	8
tRCD Setting	8
tRP Setting	8
tRAS Setting	21
tRFC Setting	74
Command Rate	1T
tRRD Setting	Auto
tFAW Setting	Auto
tRTP Setting	Auto
tWR Setting	Auto
Round  Trip Latency	Auto
tddWrTRd Setting	Auto
tdrWrTRd Setting	Auto
tsrWrTRd Setting	Auto
tddRdTWr Setting	Auto
tdrRdTWr Setting	Auto
tsrRdTWr Setting	Auto
tddRdTRd Setting	Auto
tdrRdTRd Setting	Auto
tsrRdTRd Setting	Auto
tddWrTWr Setting	Auto
tdrWrTWr Setting	Auto
tsrWrTWr Setting	Auto
	
Voltage Control	
EVGA Vdroop Control	W/O Vdroop
CPU Vcore	1.175v
CPU VTT Voltage	+100mV
CPU PLL Vcore	1.8v
Dimm Voltage	1.60v
DimmDQ Vref	+0mV
QPI PLL Vcore	1.15v
IOH Vcore	1.15v
IOH/ICH I/O Voltage	1.55v
ICH Vcore	1.1v
PWM Frequency	800KHz
	
CPU Feature	
Intel Speadstep	Enabled
Turbo Mode Function	Disabled
CxE Function	Enabled
Execute Disable Bit	Enabled
VirtualizationTechnology	Enabled
**Logical Processor Setting**	
Intel HT Technology	Disabled
Active Processor Cores	All
**QPI Controller Setting**	
QPI Control Setting	Enabled
QPI Link Fast Mode	Enabled
QPI Frequency Selection	4.6Ghz

Redid a test in X3:TC

All I can say.. wow that game really dislikes HT, big time. Guessing some other things might agree with that as well.

Anyways my results from 3 Tests.

Code:
 The Settings
Game Version: 1.2.0.0

Operating System:	Windows Vista™ Ultimate (6.0, Build 6002) Service Pack 2 (6002.lh_sp2rtm.090410-1830)
DirectX Version:	DirectX 10
Graphics Card Info:	NVIDIA GeForce GTX 285 [3819 MB]    Driver: 8.15.0011.8585
CPU Info:		Intel(R) Core(TM) i7 CPU         920  @ 2.67GHz (8 CPUs), ~2.7GHz
Usable System Memory:	6134 MB


Video Settings During Benchmark:
	Screen Resolution:	1920 x 1200 * [32], fullscreen
	Using Shader Profile:	3.0
	Antialias Mode:		2x
	Anisotropic Mode:	8x

Graphics & Shader Settings:
	Texture Quality:	high
	Shader Quality:		high
	High Quality Bumpmaps:	enabled
	Environment Mapping:	enabled
	Glow Filter:		enabled
	Ship Color Variations:	enabled
	More Dynamic Lights:	enabled
	Used Vertexsize:	40 bytes

The results
Code:
3.8Ghz w/ HT
Framerates
--------------------------------------------------------------------------------
  Scene "Trade"	 54.3  average fps 19.0 minimum fps 125.0 maximum fps
  Scene "Fight"	 100.6 average fps 44.0 minimum fps 187.0 maximum fps
  Scene "Build"	 99.4  average fps 65.0 minimum fps 131.0 maximum fps
  Scene "Think"	 49.2  average fps 35.0 minimum fps 84.0 maximum fps

Overall average framerate: 75.9 fps

3.2Ghz w/ HT
Framerates
--------------------------------------------------------------------------------
  Scene "Trade"	 48.3 average fps	 16.0 minimum fps	 114.0 maximum fps
  Scene "Fight"	 92.8 average fps	 37.0 minimum fps	 181.0 maximum fps
  Scene "Build"	 96.0 average fps	 62.0 minimum fps	 127.0 maximum fps
  Scene "Think"	 46.4 average fps	 29.0 minimum fps	 80.0 maximum fps

Overall average framerate: 70.9 fps

3.4Ghz w/o HT
Framerates
--------------------------------------------------------------------------------
  Scene "Trade"	 51.1  average fps 17.0 minimum fps 119.0 maximum fps
  Scene "Fight"	 107.5 average fps 40.0 minimum fps 240.0 maximum fps
  Scene "Build"	 127.9 average fps 84.0 minimum fps 174.0 maximum fps
  Scene "Think"	 59.0  average fps 34.0 minimum fps 100.0 maximum fps

Overall average framerate: 86.4 fps

Comparison sake between them all.
Code:
3.8Ghz (HT) vs 3.2Ghz (HT)
Scene "Trade"	12%	average	19%	min fps	10%	max fps
Scene "Fight"	8%	average	19%	min fps	3%	max fps
Scene "Build"	4%	average	5%	min fps	3%	max fps
Scene "Think"	6%	average	21%	min fps	5%	max fps

3.4Ghz (no HT) vs 3.2Ghz (HT)
Scene "Trade"	6%	average	6%	min fps	4%	max fps
Scene "Fight"	16%	average	8%	min fps	33%	max fps
Scene "Build"	33%	average	35%	min fps	37%	max fps
Scene "Think"	27%	average	17%	min fps	25%	max fps

3.8Ghz (HT) vs 3.4Ghz (no HT)
Scene "Trade"	6%	average	12%	min fps	5%	max fps
Scene "Fight"	-6%	average	10%	min fps	-22%	max fps
Scene "Build"	-22%	average	-23%	min fps	-25%	max fps
Scene "Think"	-17%	average	3%	min fps	-16%	max fps

Amazing.... 3.4Ghz nearly beats out the 3.8Ghz with HT.

Now yes this is gaming tests in the CPU section, but here is the thread that started my intrest into looking at this task. http://www.xtremesystems.org/forums/showthread.php?t=226969

Oh yeah and at 3.4Ghz I dropped another few C off the top temp with HT disabled, which now sits at 61,56,61,56. Man this thing will barely breaking into the 50's under load! Who says you can't have high performance at cool temps on air. So from my 3.8Ghz w/ HT to 3.4Ghz w/o HT I shaved off roughly 15C in temps so far. 3.4Ghz still needs to be verified that it is stable at said speed.

Has anyone else done similar tests to this in the past... really looking at the numbers between tests, clock speeds and the performance differences? I know at least with X3:TC more CPU the better of course, but not every game is like this and it will take me a long time to get all those tests since I won't be running tests for a few days (PC going to compile some things).

From the looks of it so far... 3.4Ghz is looking nice, and maybe push it to 3.6Ghz if im not going to touch voltage to much. Seems like if I bump to 1.1875 it jumps to 1.31V on the board, but its back to normal by 1.2V bios so not sure. Course im going for optimal clock... with minimal heat increase/voltage increase/ yet still delivers in majority of my programs that I use. If need be I got my fail safe for more power with my 3.8Ghz clock w/ HT which is really only good for 2 things I work on.

Commits guys, some 60 people have looked at this yet no response?
 
Most games are still not utilizing quads and are heavily GPU limited. Though a sli/crossfire setup should benifit from quads/higher clocks because the CPU starts limiting the GPUs.

AH ha, I found the link. LONG review but lots of good info. http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/19

A few quick conclusions:
FPS: i7 > C2Q > C2D with sli/crossfire (new cards).
FPS not increasing much after 3.2ghz-3.5ghz on i7, even with sli/crossfire.

I've been reading some of the HT on/off debates and there aren't a lot of #'s to back up the claims. I'm going to test it out soon, as I like the idea of dropping 7c if the extra threads aren't being utilized. I'll turn it back on in a second if I think it will be needed.

HT vs HT disabled. It actually looks like they got slightly better fps with HT disabled. Wow. Now I really have to try it. Skip down past the synthetics.
http://www.insidehw.com/Reviews/CPU/Gaming-Performance-Core-i7-vs-Core-2-Quad/Page-3.html

I've got a 2nd 275 in the mail, I'll test and see if HT helps sli/crossfire. I also want to see the difference between PCIe 2.0 8x vs 16x but thats off topic.

I'd like to think that in a year or two when the next round of system killer games are released that a 4ghz i7 is better utilized.

//hides from an angry mob of C2Q fans :).
 
Most games are still not utilizing quads and are heavily GPU limited. Though a sli/crossfire setup should benifit from quads/higher clocks because the CPU starts limiting the GPUs.

Very true, SLI benefits from the higher clocks and no doubt I have no issues with that, definatly not arguing that fact. More aimed at the single GPU (1 core) setups that most people typically have.

AH ha, I found the link. LONG review but lots of good info. http://www.guru3d.com/article/core-i7-multigpu-sli-crossfire-game-performance-review/19

A few quick conclusions:
FPS: i7 > C2Q > C2D with sli/crossfire (new cards).
FPS not increasing much after 3.2ghz-3.5ghz on i7, even with sli/crossfire.

Indeed good, and thats basically what I've been seeing in "most" cases from others. Mind you I haven't gotten that far myself and doubt I'll get to test any of that til later this week at least more clock speeds. I know at least the game im running/testing, CPU helps it out but its the question of how much does it really assist it and is it worth it for that fact.

[/quote]I've been reading some of the HT on/off debates and there aren't a lot of #'s to back up the claims. I'm going to test it out soon, as I like the idea of dropping 7c if the extra threads aren't being utilized. I'll turn it back on in a second if I think it will be needed.

HT vs HT disabled. It actually looks like they got slightly better fps with HT disabled. Wow. Now I really have to try it. Skip down past the synthetics.
http://www.insidehw.com/Reviews/CPU/Gaming-Performance-Core-i7-vs-Core-2-Quad/Page-3.html[/quote]

Yeah the HT on vs off Debate is strong. Me I'd rather leave it on... I mean I have it why not type of sence. Course with just the single test I did last night majority of the time 3.8Ghz w/ HT and 3.4 w/o HT is pretty close, and yet in some cases the 3.4 annihilates the 3.8Ghz clocks and a little. Then looking at the 3.8 vs the 3.2 clocks and average they are close, min is right on for the performance drop in the one test, and high well its pretty steady.

Flipping that HT switch on it with the increase in clocks really threw a wrench into the test numbers but figured it was good to post it at least.

I've got a 2nd 275 in the mail, I'll test and see if HT helps sli/crossfire. I also want to see the difference between PCIe 2.0 8x vs 16x but thats off topic.

I'd like to think that in a year or two when the next round of system killer games are released that a 4ghz i7 is better utilized.

//hides from an angry mob of C2Q fans :).

8x vs 16x is at very best minimal. I know there was a discussion on this recently in a thread that they are pretty neck and neck. Check out the ATI section and there is a sticking by ******** in there. He did a nice comparison between them.

Course as time goes on hopefully the CPU will become the main bottleneck again and can crank the speeds up and up preferably with HT ;)

Oh and educated guess on my WMV encoding... at 3.6Ghz w/o HT will probably tie or beat my 3.8Ghz w/ HT in encoding time. Just from what I saw quickly with my 3.4Ghz w/o HT clock that it was getting pretty close to it. Downside of course it uses nearly 90-95% of the CPU power all the time its doing it, still snappy though.

PS of yeah if you want to post some of your findings in here be my guest. I'm all for it.
 
Yeah the HT on vs off Debate is strong. Me I'd rather leave it on... I mean I have it why not type of sence. Course with just the single test I did last night majority of the time 3.8Ghz w/ HT and 3.4 w/o HT is pretty close, and yet in some cases the 3.4 annihilates the 3.8Ghz clocks and a little. Then looking at the 3.8 vs the 3.2 clocks and average they are close, min is right on for the performance drop in the one test, and high well its pretty steady.

I'd rather leave it on too. But if 3.4ghz HT off can beat 3.8 HT on, then turning it off seems like a good idea. Especially because temps drop and you could probably lower the vcore. You should do some 3.8 HT off comparisons. I wonder if I can hit 4.4ghz HT off :).

I'll post my findings too.
 
Yeah I'll have to try 3.8Ghz with HT off. I know I have before... and temps where in the upper 60's if not mistaken for 3.8Ghz.
 
8x vs 16x is at very best minimal. I know there was a discussion on this recently in a thread that they are pretty neck and neck. Check out the ATI section and there is a sticking by ******** in there. He did a nice comparison between them.

Haha, censored. Did you forget the r in Brolloks name? Very interesting, only a couple FPS difference. Though my 275's are on par with 4890s, and his tests were on 4870's. I think I'll still do some benching of my own. Maybe throw up a thread.

//I might pick up that game you are benching with, not sure how I missed the last 5 :).
 
I've been running w/o HT @ 4ghz for a week due to temps. I have a single 4870, and have not noticed any drop in fps while gaming. I dropped a couple hundred pts in '06, but I've seen that much variance just running it multiple times.
 
I'm not sure why Intel re-introduced the HT feature but I'd like it off most of the time. As far as I know, not many apps/games would benefit from HT.
 
Well had more time tonight than expected....

3.8Ghz without HT
Code:
Framerates
--------------------------------------------------------------------------------
  Scene "Trade"	 57.9  average fps 20.0 minimum fps 134.0 maximum fps
  Scene "Fight"	 118.2 average fps 45.0 minimum fps 263.0 maximum fps
  Scene "Build"	 133.3 average fps 91.0 minimum fps 182.0 maximum fps
  Scene "Think"	 62.8  average fps 38.0 minimum fps 109.0 maximum fps

Now how does that compare? Well find out ;)

3.8Ghz w/ HT vs 3.8Ghz w/o HT
Code:
Scene "Trade"	-6%	average	-5%	min fps	-7%	max fps
Scene "Fight"	-15%	average	-2%	min fps	-29%	max fps
Scene "Build"	-25%	average	-29%	min fps	-28%	max fps
Scene "Think"	-22%	average	-8%	min fps	-23%	max fps

3.8Ghz w/ HT is getting slaughtered!

3.8Ghz w/o HT vs 3.4Ghz w/o HT
Code:
Scene "Trade"	13%	average	18%	min fps	13%	max fps
Scene "Fight"	10%	average	13%	min fps	10%	max fps
Scene "Build"	4%	average	8%	min fps	5%	max fps
Scene "Think"	6%	average	12%	min fps	9%	max fps

Not bad for only a 12% increase in CPU speed.
Had to bump my CPU volts up to be on the safe side to 1.2V bios. Still under 70C for load under LinX which would put it to roughly 60C load normal applications under full load.

As for benchmarking using this game nab the demo/benchmark program from www.egosoft.com .

As for no temp drop after turning off HT... yeah you should see a temp drop even with similar voltage, though of course turning down the voltage to make it stable will help drop that temp. Me personally I can see a huge drop in temps turning it off and even increasing speed at the same voltage. To each there own though.

Man makes me want to keep it at 3.8Ghz now ;) GOing to have to try actual game results since I know how it was running before. Wish I could get SC:FA benchmark to work :/
 
@deathman20: Wow, I can't believe HT has such a high overhead that you are actually loosing FPS with it enabled. And since HT isn't being properly utilized it provides no benefit for gamers. Thanks for the link I'll try that out tonight. FPS+temp+vcore considered, I'm probably going to turn it off. After I do some benchmarks for comparison w/ and w/o sli.

@gigabit: We have the exact same chip (as you probably remember). I had a lot of problems with temps until I re-did my cooling. I'm using AS5. You can read about everything else I did at http://www.ocforums.com/showthread.php?t=606178&page=2 .
 
pardon me, i am in the process of ordering my 920 and will soon be bothering you all about tips/tricks/techniques to ocing, but firstly what exactly is HT feature y'all keep referring to and also which program are you benchmarking with?
 
HT is hyperthreading, it makes the operating system think the CPU has 8 cores.

It's really more of a feature for people doing supercomputing that run 10s of thousands of threads at once with giant clusters of processors.

Nice review deathman :)
 
I run my setup @ 4g 24/7 with HT off since I use no programs that make use of more than two threads. The main thing I do is record, edit and transcode HD TV programs to XviD. XviD will use a little of a second core but nothing more than that, my P4 with HT was clock for clock real close to the same speed of my PentiumD and my i7 with 2 cores and Ht disabled was clock for clock as fast as it with all cores enabled.

My editing program's speed is mainly based on source and destination disk speed (for this purpose my 1.5tb seagates are the fastest even tho my re3 1tb WDs are close).

Running with Ht disabled really only lowers temps when running at full load like P95 or LinX, there is very little difference when running normally except that you can usually run with less v-core which will also lower temps a little.

With HT enabled I run with v-droop disabled and end up with a idle voltage of 1.136 and a load of 1.36 all the rest of my setting are the same wheather I have Ht enabled or disabled.
With Ht disabled I run with v-droop enabled and end up with a idle voltage of 1.12 and a load of 1.28.

Here are some shots of idle and load voltages and speeds that I use both with HT enabled and disabled. Ambient temps were 20~21° for all tests

P95W7In-ht.jpg

P95W7Ln-ht.jpg

P95W7I.gif

P95W7L.gif
 
Back