Jump to content

2920XM Overclocking Guide - Detailed Results


mw86

Recommended Posts

*WORK IN PROGRESS MATERIAL POSTED IS READY BUT MORE TO COME* An Extreme Mobile Processor costs $1000 retail and at least $400 for a good engineering sample. There are plenty owners of the 2920/60XM chips and owners of M18x using unlocked bios provides optimal control over the unlocked processors power parameters including voltage (Turbo Flex VID), amperage (primary plane/TDC) and wattage (Short/Long Watt Limit / TDP). I had a system previous this one (i feel as though new system as motherboard, gpus and cpu are different) that i will refer to as 2920XM OEM sample 1. Later I will add test results from this first sample. The following is based on what i am going to refer to as 2920XM OEM sample 2 which by my test handles about 100mhz or more less and required a bit higher volts to get there. The following results are of sample 2 under the noted test conditions.

Room temp was 70F using thermostat. The test system was Alienware M18x R1 with dual 6970M, i7 2920XM oem, oem ram 1600mhz modules @ 1333mhz, oem dual 750gb sata 2 HDs in Raid 0. Fan control was used and so was a notebook cooler and a small lift of the notebooks rear on rear of notebook cooler. These results represent a system given optimal airflow not enhanced cooling in anyway as in air intake is fresh room temp air and exhaust is blown away from system enough as to not recycle hot air back into system. This cooling cannot contribute to any sort of enhanced effect skewing results. The purpose is to reveal given no power limitations at all what overclocks are stable and what settings would one need to accomplish it. Given the method i chose the data should provide adequate understanding of how speed goes up and so does heat so in a dynamic chip as such a high speed overclock is only handled best in one to two cores and decreases as cores are added otherwise temperature becomes a limitation making such an overclock unviable... so stick to a given range and you shouldn't see temps higher in other apps since Intel Burn Test produces more heat and cpu load than most real world use apps. So use this chart as willed one can derive everyday overclocks or use reference for those of you who benchmark. I personally have and see more use not in benchmarking but real world use. To get similar results fan control is not required and very optional as system fan control will still start cooling when the cpu gets too hot, I just like to use defensive or pre-emptive cooling (preventative). For those of you overclocking its all about bang for buck yes? So pick one that best suits your system usage and get that buck out of your XM cpu. With these and the unlocked bios it takes the guessing game out of this or depending on a single overclock someone mentioned to cover all your needs... no its best to know an entire range. Brian K. you just acquired your 2920XM try gaming with the main overclock i not below using 0 flex... I am certain its useful. Before i made this i have a notebook of jumbled notes i made.

This is my first attempt at this and to provide users with a useful guide. I will add more as I have time. I hope you all like it I tend to devalue my work. For about 3 months I gamed on 4.4ghz and used 44/44/44/44 at 0 flex using 97watts long limit and 2400 primary plane so 300watts the 97 watts keeps 0 flex voltage enough as under heavy load the 97 watts causes the cpu to drop in a more acceptable range but now I mapped the following I will use the last overclock at the bottom using 0 flex. The most coolest and highest 8thread capable overclock is the one using the least voltage. If one can use enhanced cooling like an air-conditioned climate controlled room perhaps one can use higher 8 thread loads but for average room temp the following should paint a great visualization of recommended ranges.

The tests were done by stock baseclock that is 99.8 mhz in the bios which shows up as 99.7 in Windows. Thefollowing is presented and tested by using multipliers as the variable and for majority of tests other variables are the voltage and some watts are different but was needed to be higher as under the type of load in the program Intel Burn test i was using most cpus under stock settings are dropping to 3.2-3.4 or maybe 3.6-3.8ghz with a basic overclock. Using more amps is required not vice versa as one has to be higher the other doesnt need to be higher to yield better performance at same speed ranges and to prevent the mhz from dropping under load due to power restrictions... these cpus actively dynamically raise speed and volts many times a second and when a heavy work load like per say encoding, rendering, some gaming (the newer games are starting more two recomend quad core cpus) the mhz tend to drop even if temps are in check and thats because amps and watts arent high enough or voltage is too high. Given 97 amps and start raising watts shortly after 10-20 watts its ineffective at actually boosting the mhz max speed that it will maintain at those loads on 8 threads. Now do the opposite and you get a better performance yield... do both and your all set but I suggest raising AMPS / Primary Plane high and only ever changing watts so 2400 primary plane or 300 amps max has shown me it can push any ghz under any load other factors would be only limiting factors (temps). So 300 amps and 95-170 watts should be good to maintain any of the ranges below no matter how much your heating up. Note cpu heat raising requires more amps and watts for same load and speed interestingly enough... so like 97 amp 99 watt might maintain a speed under average temps but as temps raise from normal usage that amount alone can require more amps or watts if the watt and or amp limit retricts it anyway the speed drops... this is not throttling as that usually pertains to temps max limit imposed by maker. Another form of the cpu speed dropping not from thermal throttling is a pattern i noticed that despite all limiting factors not limiting things lol that if flex was too high it didnt matter how high i had amps or watts the voltage or flex being too high eventually imposes its own strange drop in mhz starting in 8thread working back to single thread the higher you go and huge amounts of heat if you attempt to near that new imposed max speed at that voltage so thats why the below chart shows where the thresholds are under such heavy workloads/ temps and even voltage ranges.

There are other methods of overclock i have explored and can be used well too. using same high amps like 300 set a slightly higher frequency set than voltage will handle and use a little lower watts and under light loads you can use higher frequencies but as load raises youve saved it from bsod124 needing voltage because the cpu starts naturally dropping speed because of the watt limit you've set. example is setting 45X on multis and using lower flex than needed to keep cool and just enough watts to maintain 4.5ghz in a couple cores but when it comes down to it and voltage requirements start raising the watts keep it in check. So have at it decide what ranges you like as noted you dont need watts higher than 95-170 for the most part but amps should be way above 776primary plane or 97 amps which is the stock range. As i have time i have lower workload apps ill map out too which will take less time.... wprime and vantage are some post suggestions if you want me to try some for you. lesson benchers wprime 32m gets higher scores at higher mhz 8thread in both cases and only runs for a few seconds so heat is less an issue and higher scores yield from the higher mhz so do runs like that and if going for 1024m it about the highest 8thread it will do but temps will be an issue so you need to use as low a voltage possible again in both cases since unlike below we arent going for crash proofing but fastest coolest overclock to get a higher bench score... dont confuse below for that as its more for showing what can be used ofr a all day everyday overclock but has its uses for you guys benching and or trying different speeds and keep crashing not knowing whats the issue. not all cpus are alike but overall your cpus should handle the same ranges given the same settings... and if the app has lower load then it should no doubt run full speed and even cooler than i have posted. please not this is stock factory paste and i noted i have 2 cpu cores that are 10c hotter than other two from factory under and even 8thread load accross cores... so actually that helps make these results more protective of you guys because most wont see as high max temps on the cores as i did...

5 tests of Intel Burn Test (system is on stock OEM paste) 2000MB ram load at 5tests unless otherwise noted

Results are interpreted as follows if a temp is noted per core that range was tested and passed and that was the resulting temp max via Throttle Stop with more data option (8 reads a second), if a - is noted the test was not performed possibly uneeded, if BSOD124 is noted that meant the test was performed and crashed due to requiring more voltage and finally if Thermal Shutdown is written the system failed to maintain thermal stability and thermal protection kicked in turning off the system. each test of 5 took any-wheres from 6 minutes to 15min depending on threads used

[ATTACH=CONFIG]3394
[/ATTACH] 1 Before a Dual Core run showing IBT and TS configured [ATTACH=CONFIG]3395[/ATTACH] 2 During the beggining of a Dual Core run [ATTACH=CONFIG]3396[/ATTACH] 3 During the beggining of a Dual Core run SHOWING AFFINITY [ATTACH=CONFIG]3397[/ATTACH] 4 During the beggining of a Dual Core run SHOWING AFFINITY SET TO DUAL CORE [ATTACH=CONFIG]3398[/ATTACH] 5 During the beggining of a Dual Core run SHOWING AFFINITY SET TO DUAL CORE AND IT RUNNING

temps and stats on 44/44/43/43 300 amp/300watt

[ATTACH=CONFIG]3399[/ATTACH] 44 44 43 43 tdp 300 tdc 300 flex 0 wei cpu 7.8 [ATTACH=CONFIG]3400[/ATTACH] 44 44 43 43 tdp 300 tdc 300 flex 0 wei cpu 7.8 20tests IBT 2 4 6 8 threads [ATTACH=CONFIG]3401[/ATTACH] 44 44 43 43 tdp 300 tdc 300 flex 0 wei cpu 7.8 wprime 1024m 23m [ATTACH=CONFIG]3402[/ATTACH] 44 44 43 43 tdp 300 tdc 300 flex 0 wei cpu 7.8 super pi 32m [ATTACH=CONFIG]3403[/ATTACH] 44 44 43 43 tdp 300 tdc 300 flex 0 wei cpu 7.8 super pi all scores





[/CENTER]
TDP/Watts TDC/Amp/Primary Plane Voltage/Flex Speed Range Tested Ghz Multipliers 1 Core 2 thread Temps 2 Core 4 Thread Degrees Celcius 3 core 6 Thread 4 Core 8 Thread
99 2400 Primary Plane / 300 Amp 25 4.5 45/45/45/45 68 100 - -
105 300 25 4.0 40/40/40/40 - - 99 -
99 300 25 3.7 37/37/37/37 - - - 99
99 300 25 4.6 46/45/40/37 69 - - -
99 300 25 4.6 46/45/45/45 73 - - -
99 300 25 4.7 47/45/45/45 71 - - -
99 300 25 4.8 48/45/45/45 BSOD124 - - -
99 300 27 4.8 48/45/45/45 71 - - -
105 300 28 4.8 48/45/45/45 BSOD124 - - -
105 300 29 4.8 48/45/45/45 BSOD124 72c - - -
110 300 30 4.8 48/45/45/45 BSOD124 73C - - -
110 300 31 4.8 48/45/45/45 76 - - -
300 300 31 4.8 48/45/45/45 BSOD124 - - -
300 300 99 4.8 48/45/45/45 91 - - -
300 300 99 4.9 49/45/45/45 92 - - -
300 300 99 5.0 50/45/45/45 BSOD124 (CPU FREQUENCY CEILING) - - -
300 300 33 4.9 49/45/45/45 BSOD124 - - -
300 300 44 4.9 49/45/45/45 81 - - -
300 300 44 4.9-3.7 49/45/40/37 81 86 95 100
300 300 31 4.8 48/45/45/45 BSOD124 - - -
300 300 33 4.8 48/45/45/45 75 - - -
300 300 32 4.8 48/45/45/45 BSOD124 75C - - -
300 300 33 4.5 48/45/45/45 - 89 100 -
300 300 33 4.0 48/45/40/37 - - 92 -
300 300 33 3.7 48/45/40/37 - - - 97
300 300 33 4.6 48/46/41/37 - 86 97 -
300 300 33 4.7 48/47/41/37 - 87 - -
300 300 33 4.7 48/47/45/45 - 91 - -
300 300 33 4.8 48/48/45/45 - BSOD124 - -
300 300 34 4.8 48/48/45/45 - BSOD124 - -
300 300 35 4.8 48/48/45/45 - BSOD124 - -
300 300 36 4.8 48/48/45/45 - 95 - -
300 300 36 3.7-4.1 48/48/41/37 - - 98 95
300 300 33 3.8-4.2 48/47/42/38 - - 98 97
300 300 36 3.8 48/48/41/38 - - - 98
300 300 33 3.9-4.8 48/47/43/39 BSOD124 86 100 99
300 300 44 3.7-4.9 49/48/40/37 80 91 91 96
300 300 34 3.7-4.8 48/47/43/39 BSOD124 - - -
300 300 35 4.8 48/47/43/39 79 88 100 THERMAL SHUTDOWN
300 300 35 3.8 48/47/43/38 - - - 98
300 300 20 4.7 47/45/45/45 BSOD124 - - -
300 300 44 4.9 49/48/41/38 84 93 97 99
300 300 44 4.9 49/49/41/38 83 BSOD124 - -
300 300 50 4.9 49/49/41/38 - BSOD124 - -
300 300 24 4.7 47/45/45/45 70 - - -
300 300 25 4.7 47/47/45/45 - BSOD124 - -
300 300 27 4.7 47/47/45/45 74 85 - -
300 300 27 4.7 47/47/45/40 - - - THERMAL SHUTDOWN
300 300 27 4.7 47/47/45/39 - - 100 99
300 300 24 4.6 47/46/45/45 BSOD124 82 - -
300 300 25 4.7 47/46/45/45 70 85 - -
300 300 25 4.5-4.0 47/46/45/40 - - THERMAL SHUTDOWN THERMAL SHUTDOWN
300 300 24 4.5-4.0 47/45/45/40 - - 100 THERMAL SHUTDOWN
300 300 25 4.4-3.9 47/46/44/39 - - 100 99
300 300 24 4.5-3.9 47/45/45/39 - - - -
300 300 20 4.6 46/46/45/45 - 87 - -
300 300 19 4.6-4.5 46/46/45/45 71 87 THERMAL SHUTDOWN THERMAL SHUTDOWN
300 300 18 4.6 46/46/45/45 - 84 - -
300 300 16 4.6 46/46/45/40 - 84 - -
300 300 14 4.6 46/46/45/40 - BSOD124 - -
300 300 15 4.6 46/46/45/40 - BSOD124 - -
300 300 16 4.0 46/46/45/40 - - - THERMAL SHUTDOWN
300 300 16 4.6-3.9 46/46/45/39 - BSOD124 100 99
300 300 17 4.6-3.9 46/46/45/39 BSOD124 84 100 99
300 300 14 4.6 46/45/45/40 BSOD124 - - -
300 300 18 4.6-3.9 46/46/45/39 70 84 100 99
300 300 15 4.6 46/45/45/40 BSOD124 - - -
300 300 16 4.6 46/45/45/40 BSOD124 - - -
300 300 17 4.6 46/45/45/40 BSOD124 - - -
300 300 14 4.5-4.0 45/45/45/40 67 82 100 THERMAL SHUTDOWN
300 300 12 4.5-4.0 45/45/45/40 63 76 96 100
300 300 10 4.5-4.0 45/45/45/40 63 75 96 97
300 300 9 4.5-4.0 45/45/45/40 - BSOD124 94 96
300 300 10 4.5-4.1 45/45/45/41 - - 95 99
300 300 O 4.4 44/44/44/42 BSOD124 - - -
300 300 1 4.4 44/44/44/42 BSOD124 - - -
300 300 2 4.4-4.2 44/44/44/42 BSOD124 - - -
300 300 4 4.4-4.2 44/44/44/42 60 73 93 99
300 300 O 4.3 43/43/43/43 59 70 89 99
300 300 O 4.4-4.3 44/43/43/43 59 76 88 98
300 300 O 4.5-4.3 45/43/43/43 BSOD124 - - -
300 300 O 4.4-4.3 44/44/43/43 59 70 89 99
300 *THIS TEST UNLIKE ALL OF ABOVE WAS DONE 20 TESTS FOR EACH CORE INSTEAD OF 5 LIKE THE ABOVE TESTS PER CORE. 300 *NOTE THE TEMPS ARE BARELY HIGHER THAN THE 5 TESTS UNDER IBT MOST CPUS MUST DROP THEIR SPEED ONLY A TEST OR TWO IN CONFIGURED WITH 300 WATT MAX AND AMP IT WONT (THESE ARE NOT ACTUALLY DRAWING ANYTHING NEAR THAT THE VALUES ARENT TAKEN LITERAL BUT USED IN A FORMULA MADE BY iNTEL) O 4.4-4.3 44/44/43/43 60 73 91 99
Edited by mw86
  • Thumbs Up 10
Link to comment
Share on other sites

  • 1 month later...

been running 4.3ghz core states disabled with 1024amp set and 122watt... when that ran fine for a month i upped it to 1024 watt also. Using 0flex on 43/43/43/43 and under only gaming i didnt see to far over 80c... recently i upped it to 44x so 44/44/44/44 core states disabled that i used 3flex for as I recorded in the above chart and still been gaming for awhile over various games and 88c i saw max so far... everything is very zippy. I'm thinking of setting 45x or 46x and setting Throttle Stops "Alarm function to switch to 43 or 44x if it reaches 10DTS so 90c :) maybe ill be more preventative and set 15 DTS or 85c :) suprisingly idle is around 40c-50c max idle... but performance is zippy no c-states. I may for benefit of lower temps enable cstates again and even set cpu power options to Balanced mode so cpu speed is tied to load and unless high load is detected turbo wll not kick in then. test games have been Mass Effect 3 Demo, Skyrim, BF3, The Witcher 2 and will also be trying Dirst 3 and Rage too. since for me 43x (0flex) and 44x (3flex) have ran so cool in most apps I use (i know real rendering or transcoding etc would produce more heat than most apps i use) so i'll try the 45/46x theory at 85c/90c and under and 43/44x when above... would be great if Asetek's prototype is released. Given their claims of their liquid cooling setup in my opinion and observation it can be done on stock cooling as they only ran benchmarks and at ranges possible on stock heatsinks... had they claimed 4.4ghz for longer than simply each of the short common 3d benchmarks they would have claim to a feat not possible on stock M18x cooling... looking forward to see actual comparison data but in mean time i hope those of you on M18x with XM realize that given the same 44x overclock your system can survive vantage and 3dmark just the same. I've been gaming and can say that in actual games the temps seem fairly in check at 43x as i had been using it at that for months at 0flex... those worried about overheating can use Throttle Stop to make your own safety speed drop.

Example:

-set two profiles up in Throttle Stop

-Set one to your everyday high overclock... example 44x @ 3 turbo flex in bios on main AC profile with cstates and anything else up to you. Go to TRL set all 4 active core multiplier options to the same value and make sure eist is checked and set multiplier is checked. press down on the arrows under to the right of set multiplier until you see some number below the number you set in TRL and now press back up until you see turbo. leave it like that... i do this because if you were using a multiplier higher before and it said turbo to start its still set to the old multi and once you press down you see the number one multi down from where you were and by pressing up and showing turbo it shows the number that is 1 less that what is set in trl and the next press up it shows turbo thats how you know its set to the number in TRL for 4active cores. I had to do this so its running the correct multiplier in Throttle Stop.

-set other profile the same way to a cpu speed all 4 cores can handle even under extreme load and estimated summer temps (overestimate to be safe) so this Multiplier might be 4ghz-3.5ghz 40x-35x (better than system throttling to 2.5-2.7ghz... unnecessary and causes games and apps sometimes to stutter from the sudden drop...)

-Go into Throttle Stop options and check "Alarm". Set use profile to the second profile number you made for downclocking. Set "DTS" to for a example ill say 80c so DTS would be 20 meaning or 20 degree units away from max temp before thermal shutdown (100c for i7 2900XM series). Since stock throttle temp of M18x is 85c you would then see further drop in mhz at 85c to base speed of xm series (2920xm 2.5ghz and 2960xm 2.7ghz). If on unlocked bios you could set Throttle Temp to via unlocked menus under "thermal management" to 90c (max menu limitation). Then maybe set DTS alarm to 85c or 15 is the value you would enter in Throttle stop options. If an advanced user you can use a bios with a modified value in the unlocked bios higher than max throttle temp value based on menu limitations so it can be set over 90c as long as you dont re-enter that thermal management menu and save bios as doing so will reset it to default 85c :| So one example on that is set 95C using bios modding software (hex editor is example... or get a modified version from a felow TI user) and then set Throttle stop alarm to 90c so 10 DTS and you will be on max overclock at under 90c and at max safe overclock while overheating at under 95c and at 95c + the system 2.7-2.5ghz throttle would kick in :)

You need to check "ac on battery off" and "battery monitoring" also you should check Default "AC profile" to the profile number for the max overclock you set first in Throttle stop options. Additionally if you can setup a Battery Profile and set a reasonable cpu multi (base clock i would highly recomend) and since Throttle Stop is so amazing it wont throttle as long as you check "Clock Modulation" and set it to 100% in the battery profile. If you chose to do so in Throttle stop options check under Default "Battery Profile" and set the number to the profile you made for battery speeds. now on main window of Throttle Stop hit "Turn ON" and you system will run at your max overclock as long as its under your alarm temp value you had set if its not it runs at your second profile made for long hauls processing or overheating moments and if it gets hotter still will throttle at the systems bios set throttle temp when reached and hopefully not but if temps kept rising your system would then shut off due to thermal protections in bios set to default of 100C.

*if using powersaver and or cstates at high frequencies and cpu drops and rises from 800mhz to extreme speeds can cause a voltage related bsod in rare cases if turbo overvoltage or turbo flex is too low as even cstates itself causes extreme overclocks to be instable due to the cpu no applying the higher needed voltage fast enough before the frequency went from 800mhz to the extreme value... not an issue with the above described settings.

*DC exit time in the Throttle stop options can help if system hangs upon plugging AC power in going from battery and your AC profile loads automatically increasing the delay from zero will allow more time to pass once plugged in before the AC profile you set is loaded. great feature Unclewebb :)

*You can enable "power saver" on main screen by checking it on a per profile basis and by setting the load percent in Throttle stop options called "power saver c0%" to whatever you want cpu to have as max load over all threads (physical and virtual cores) when under this value system will run at 800mhz on all cores which is SandyBridges minimum speed. when cpu load is over the C0% that is set (remember that even means the cpu load when its dropped to 800mhz times 4cores... so C0% may be inflated when cpu is sitting at 800mhz vs C0% observed running at normal speeds keep that in mind setting power saver c0% so cpu doesnt bounce between states as much us a higher value if its bouncing up and down from 800mhz rapidly unless desired), it will run at the normal programmed speed for the profile its currently running under in Throttle Stop... or dynamically if thats what the profile is set up as (core states enabled). Default is 35% I have heard 13% is a good hair trigger for it if you want minimal power saving but likely you will see cpu flipping from normal to 800mhz a lot set like that. So i suggest 25%-default 35% if you choose to use the feature (which can be combined iwth no core states so cpu runs max 4active speed unless under powersavers C0%) I show 25% in my example. Like i say above though "if using powersaver and or no cstates at high frequencies and cpu drops and rises from 800mhz to extreme speeds can cause a voltage related bsod in rare cases"... not common for these ranges and suggested power saver C0%.

post-34-14494992823683_thumb.png

*side note* in the testing in the OP i needed 4flex for 44x at those temps and on IBT on that chip sample this is my first 2960XM so it hasnt been as thoroughly tested but been fine at 3flex for 44x ( i try to use enough volts just in case jumps in load occur hence the IBT table above... more to come eventually)

post-34-14494992822172_thumb.png

post-34-14494992822489_thumb.png

post-34-14494992822781_thumb.png

post-34-14494992823083_thumb.png

post-34-14494992823365_thumb.png

Edited by mw86
  • Thumbs Up 3
Link to comment
Share on other sites

Guest
This topic is now closed to further replies.
×
×
  • Create New...

Important Information

By using this site, you agree to our Terms of Use. We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.