Multiplier manipulation XM processors =)

Hope I am not digging up too old of a thread but I am getting good info from this and wanted to ask a question.

Are most people just running the 32M test and if all multi's stay at max set (24 or 25 or whatever) then calling it good?
I find I can keep 26 at a reasonable TDP/TDC but if I run the 1024M test it starts to throttle after running 25%-50% through (depending on multi and tdp/tdc)

So, I guess I am just curious what tests people are doing before calling it good.

oh, and this is in my M17x-R2 (just found this thread here so there you go)

D.

 

Well, wPrime test will give you a 100% load across all cores (I assume you set it to 8 threads) which actually won't happen in a game, at least not over such a long time as during the 1024M test.

It's just not possible for the cooling system to keep the processor temp low enough during a long time under such a load...

When you get the 32M on 24x multi it should work fine for normal gaming and similar imo.

Btw, what are your max temps during the 32M test and what are your settings? Oh, and the time (8 threads)?

 

You might need to look into some mods to reduce your temps like a good thermal paste job and some high end paste. Might need to tighten down the cpu heatsink once you have the system on and running prime since a slight crooked heatsink will result into uneven temps. You can tighten each screw while monitoring temp and see what is the best combination. Throttling only happens when you have high temps and you can use throttle stop to prevent the throttling but ultimately it is running too hot and causing the chip to downclock it self to protect it. Also lower your TDP/TDC since that will result in cooler temps since you are pushing less power. You need to find the bare minimum TDP/TDC to make the multipliers stick under full load and extra TDP/TDC just goes to HEAT.

 

anybody looking to get rid of their 920/940xm?

 

Don't think so. Starting price on ebay is about 410$ at the moment.
There are only "want to buy" threads for xm cpus at the NBR marketplace atm, as far as I know.

 

yea noticed the prices are slowly going lower though... so i guess ill wait abit till they get down to 350$-ish ... around xmas they were up to 460$ on ebay!

Also, is there a difference in Overclocking headroom between a 920 and a 940? or is the 940 really a rebranded 920 with higher stock multipliers?

 

Hey guys been looking for a answer, what are the default multipliers that throttle stop show for the 920 and the 940 for every core combination. Right now I have a 720qm 13/13/18/21
Thanks!

 

The prices seem to have crashed on ebay I'm seeing 920xm popping up for £260 thats not bad at all considering most were over £500 late last year.

 

yea i'd wait a little more... as more sandybridge extreme cpus ship, the demand for 920/940xm will decrease, lowering prices a little more

 

Standard setting of my 920xm ES was:

tdp/tdc : 62/62
turbo limits: 24 / 23 / 17 / 17.

Someone with a 920xm told me recently that his standard tdp/tdc values were 55/49... but the standard multis are definitely the same.

 

Sorry for the late reply.
My temps max at 93C and all 4 cores are in the same temp range.
So it should not be throttling due to temperature.
Plus, this was done with my external laptop stand/cooler off since I wanted to see how the system itself would handle the heat so temps will drop a bit when I turn on the fans of the external laptop cooler

I know that running all 4 on turbo for a 1024M run is a bit extreme (will not have this happen in everyday computing despite how many raw images I am stitching together) and I can easily get a stable 27+ multi on a 32M run; I am just trying to figure out how "strict" I should be on what I call my max STABLE OC.

I had a bit of a scare as I can't do anything over a 2%OC in the bios (3% posts but I get errors in p95).
At first I thought the chip might be bad but then I realized it was my 8gb ram not liking going faster than 1333 (despite upping voltage) so I need to do any OC'ing with throttlestop. and multi manipulation

TIA

D.

 

Re multipliers:

940XM (OEM) - 25/24/18/18 with 62W/62A TDP/TDC

920XM (OEM) - 24/23/17/17 with 55W/47A TDP/TDC

820QM (ES) - 23/21/15/15 @ 45W/37A TDP/TDC

 

so Rev, does that mean once the multipliers/TDP/TDC are altered on the 920/940xm with ThrottleStop, that They overclock to the exact same amount (give or take)???

Is there any benefit for going with a 940 over a 920xm on ebay?? I'm trying to decide which one to get

 

I would say the 940xm is NOT worth the price difference (double the amount of a 920xm).

Using throttlestop you could easily set everything the same as a 940xm and never know the difference (except the extra change in your wallet/purse/shoe box)

As far as I am aware, the 940xm and 920xm are the exact same chip, just set different for the price premium.

D.

 

To the best of my knowledge the two extreme processors are for all practical purposes the same once the magic of ThrottleStop is applied. It may be that the 940XM will accommodate a 29x setting for a single core, whereas the 920XM no longer will (at least in my experience, following recent BIOS revisions). We need to check with Inap on that; he has cranked up the 940 as high as it will go, using extra power and extreme supplemental cooling. Remember, you start running into power/heat walls around 25x-26x across the board.

 

ok...thanks Rev... in any case... what's the highest OC attainable without any extra cooling? Also... Has anyone tried using a combo of Setfsb and TS to OC? which proves to be more stable? I would think its TS no? since you can feed it more juice by adjusting the TDP/TDC settings huh

 

there pretty much the same, only the 1 multi difference, which carries over to oc too.
920 stable @ 26 all across
940 stable @ 27 all across

so you pay alot more for that 1 multi.

 

intel does a lot of testing before releasing individual chips. The use the binning process where the test each chip to see how stable they are and how high they can overclock. Take the sandy bridge desktop chips 2500k and 2600k...they are both the same thing except hyper threading and cache. They test both and the one that is able to hold a higher overclock and stable they will bin that as the 2600k and enable hyper threading. On the 2500k they will disable the hyperthreading and sell it for $100 less. Essentially they are the same chips but intel is cherry/hand picking higher end chips and unlocking more features. The 2500k can overclock just as high but there is essentially a 1 multiplier difference along with hyperthreading. Intel previously use to sell unlock cards so you can unlock the multiplier or hyper threading. It is much cheaper to make 1 chip and internally lock features or speeds...than to make 27 millions combinations of chips.

THe 940xm is a higher binned chip than the 920xm and has 1 extra multiplier. Inap and me use to get into battles with my 920xm ES and his 940xm and he was able to keep 1 multiplier ahead of me. I got that chip up slightly higher clock than his for a bench session but his was stable for longer. The 940xm he got compared to his own 920xm...the 920xm was able to go higher. Also his 940xm needed less power Tdp/tdc to hold the multipliers compared to mine

Is it worth the premium to go to 940xm versus 920xm...well if you like to brag about your e-peen size than yes ...in reality both are just as good and either one can beat the chip. Both will perform at stock 24x and 25x and you can safely bump each one up 1 multiplier to 25x and 26x. Making your 920xm into 940xm and 940xm into 960xm .

Buy which ever one is cheaper and if you have to choose

920xm OEM versus 940xm ES = choose 920xm OEM

 

24x across all cores should be possible, some people pushed it up to 26x, but I think this really depends on the ambient temperature as well as the time you have a 100% load, because it will get hot anyways after a while under full load. It probably even depends on each chip.. keep in mind that the 920xm is guaranteed to work at 24x multi on one core (as long the TDP/TDC values are ok), the 940xm up to 25x on single core turbo boost.

...everything else is "bonus" and hence it's totall possible that some chips are more likely to run at a higher multi across all cores than others, but as far as I know, you shouldn't have difficulties to get 24x across 4 cores in a M15x.

However, it's always a good idea to have a good airflow below the system.


I wouldn't recommend using TS and setFSB, though I have to say I never actually gave it a try.
The 3.2 GHz or even a bit more across all cores which you can get with TS is already a great overclock imo, and I think there's no need to push the CPU even more, reasons are stability, heat dissipation as well as power consumption.

SetFSB can interfere with turbo boost when used on a i7 CPU, I don't know whether TS will prevent this...
Anyway, TS is a powerful tool for the extreme processors and it's not necessary to use setFSB as long as we have TS imo.

 

yea but like i was thinking... 24 across the board with the FSB bumped to 150mhz .. that's 3.6ghz on each core...i'll play around with it when i get it i guess... anyways ok i'm convinced... ill get a 920xm... and i think i almost have a deal on a oem one

 

I don't recall ever hearing of a 920/940xm in a motherboard with the unlocked PLL necessary for SetFSB, and it would have been enough of a novelty to get a lot of attention. But ultimately it probably doesn't matter whether you use 24@150Mhz or [email protected] to get to 3.6Ghz, except for collateral effects of the higher BCLK on memory timings, etc.

On binning, what Faiz23 says is true generally, but binning was never used to separate 920's and 940's. Intel just relabeled the processor and bumped the multipliers one notch; the 920 disappeared and the 940 appeared simultaneously from a production standpoint. Any sales overlap between the two was a result of existing 920 inventory.

 

m15x with TS

CPU-Z Validator 3.1


m17x with TS/Setfsb combo (same chip)

CPU-Z Validator 3.1

Setfsb rarely works on a m15x and pretty much always works on a m17x

No issues with using the combo ts/setfsb just keep in mind you are increasing bus speed which results in ram and timings being overclocked and can lead to unstable overclock. Now you have to worry about your cpu crapping out or ram crapping out and resulting to BSOD. I was not able to go to 30x on m17x using setfsb but 30x on m15x with stock fsb was fine.

 

I'll see when i get my 920xm ... maybe i can work my magic... and i like the idea of running at 150 instead of 133 precisely for my ram to work a lil quicker...

Thanks everyone Rev, Faiz, Inap... nice seeing you guys still around... barely know anyone around here anymore haha... jokers banned? r3 came out with 1 gpu? Whats going on :O ???.. times are changing

 

yeah have fun bsoding all day, hahahaha
on my 2nd M15x with 920 i had setFSB working with it but only manage about 145 with 26x multi.

 

(((STEREO))), you bring a certain familiar spirit back to the forum. Welcome, my friend. What happened with Joker? Moderators are frowning on any criticism of AW's new baby.

Nice rig you're driving these days.

 

Just got my 920XM running it at (home made) 960XM
The Logic:
920XM = 17/17/23/24
940XM = 18/18/24/25
960XM = 19/19/25/26

 

You should push it a bit more... 24 across all ccores that's the way to go
...like the official Alienware overclock before Bios A08.

 

Yeah I would but that's my normal processing speeds such as watching movies and browsing and photoshop. Will defo push higher.

 

Yeah, this will be more than enough for watching movies and browsing

 

congrats on the 920 upgrade. you got the makings of a beast right now.

 

What's the most you could bump up the multiplier on a 940XM without having to alter the TDP/TDC?

Alternatively, what's a TDP/TDC that would probably work for putting it at 25 or 26 across all?

Thanks.

ALIENWARE M15x | Core i7-940XM (ES) | GTX 460M | 8GB DDR3 1333 | 1080p | Blu-Ray | 256GB SSD

 

It's hard to tell as every chip is a bit different. It's even possible that some 940xm won't even handle a 25 or 26 multi across all cores very well... the turbo boost is designed to work for 25 multi on one core.

However, usually the 940xm are pretty tough, indeed, and they may handle this OC pretty well, but the TDP/TDC can vary from chip to chip as I said.

My 920xm for example works fine with a 24 across all cores wit a 87/77 setting, but I think I could set it even lower to get it to work, I havent played with the settings in a while.

Imo the best thing is to try different TS settings and test it with wPrime or something similar. With throttlestop you'll see instantly whether you have to raise the values or not.

 

Thanks. I'll try to feel it out.

Is there any rule of thumb ratio for TDP/TDC?

ALIENWARE M15x | Core i7-940XM (ES) | GTX 460M | 8GB DDR3 1333 | 1080p | Blu-Ray | 256GB SSD

 

No, some people even reported that adjusting the TDC didn't change anything for them, no matter how low or how high they set it.

Just go ahead and try some settings.


By the way, when you go to the user control panel you can edit your signature and don't need to post it every time

 

This can provide some guidance:

​

 

Wow, that chart is perfect (as a guideline, obviously).

Thanks to you both, this has been a great help.

 

Hey guys... FINALLY got my 920xm CPU in today... And slapped it in... Seems I got the tightening just right on the screws cuz temps are pretty damn even (I was worried it would be hard to get them that way after the headaches I had with my 5870s)

Anyways... So I didn get that much time to play around with it... But from my initial impressions... Doesn't look too good.

I'm on bios A10, 5% CPU OC + 150mv bios settings...

So I decided to start low... And first aim to get 24x across the board... Well, it seems that even with TDP/TDC set to 95/85 .... When I run wprime 32m ... Throttle stop shows the multi's only maxing out at around 20.88... WHAT THE HELLL... did I get a dud? A garbage edition xm ? LOL ...

Funny thing is that now that I came on the forums... my 32m score is ~ 8.4secs... just about the same as the chart that rev posted up with 24x across the board...

shouldnt TS (2.99) show 24.00 under the FID column if it's really going at that multi? Cuz 20.88 is pretty pathetic... But the half decent 8sec score would suggest otherwise? I dunno ??? I wish I was home to test it out more... Only had about 20 mins with it

Didn't even try to go higher to 25 or 26 or 27... Cuz I figured heck I can't even go past 21 it seems... Anyone have an idea what's going on here?

Also, do you guys set the 5% OC in the bios and increase the voltage to 150mv? Or is it better to leave those settings stock? Any help would be greatly appreciated. Thanks guys

 

wow nice congrats on finally pulling the trigger. hmm only way is to test it with a higher tdp, haven't tested it on an M17x before so i'm not sure about the bios settings.

 

I think I'm just gonna go all stock in the bios n see what happens... Gonna have to be late night tho I'm still not home

 

try this, force turbo multi with throttlestop (set multi to turbo and set TRL to 24 across the board)

keep the tdp/tdc at stock and run a benchmark.

now for me, if i increase the tdp/tdc just incrementally, i will see my multiplier go up in increments as a result. if not, i don't think it's working properly.

also, make sure your windows profile is set to performance.

 

Alright so basically, i was right. The M17x Overclocking bios settings were the problem. As soon as i set it all stock in the bios, things started to make sense (and take off .. lol)

Its weird because you'd think that the setting to increase the cpu core voltage by either 75mv or 150mv would increase stability (as is the case with gpu overclocking almost ALL the time) ... but in this case, it would actually do the complete opposite... Haven't really tested whether the setting in the bios for increasing the 133mhz base clock by 5% affects things... but i have a feeling it's going to also screw things up... but anyways.. this is what i managed going all stock in the bios OC'ing...

TDP/TDC : Multi on all 4 cores

88/62 : 24x
90/62 : 25x
95:62 : 26x .......

Makes much more sense now ...Those are the absolute lowest TDP/TDC settings for getting it to run constantly at those multi's! I also found that TDC does absolutely nothing for increasing stability... so i left it at stock.

So far 27x has me bsoding pretty damn quick... but im wondering if theres a TDP setting that would allow it across the board.... anyone know the max TDP settings the cpu / motherboard can handle? anyone see settings past 100W? like 105 & 110? I know thats pretty crazy, but i mean just for a 32m run? debating whether i should try it or not.. hehehe...

Also got to try what multi's i can achieve by settings diffrent ones to each core (ex: 28/24/17/17 or something)... but that will have to wait for tomorrow... its 6:42 am here ... i'm off to bed now

 

on my 27x runs i have to set tpd to 105. but i think your 920 just can't handle the 27x all across try 27-27-26-26

 

27x on 4th core was FAIL. Best i could do

Stable:

Core 1: 28x-26x
Core 2: 28x-26x
Core 3: 27x-26x
Core 4: 26x-24x

 

WOOOOT!! 105 did the trick!!!! 27x across the boards baby .. i just didn't know if 105 was too high and didnt wanan fry anything! .. im wondering how much wattage these cpu's can take? Geeze...

 

what do you guys use for everyday usage? considering the cpu never reaches 100%load.. isn;t it safe to leave it on like 26x across the board since it'll never have to use it anyways? or shoul di go for like 28, 28, 26,26 if it can handle it

 

i run 26x all day and only bump it up to bench. normally 105 does the trick but i went as high as 115 with no problems but it didn't really affect anything past the 105.

 

Man, those R2's can handle the power. Congrats, (((Stereo))). What kind of temps are you seeing with 27x flat?

 

temps? well i only benched with those multi's on the wprime 32m test... heats up real quick to 77-80*C in the ~7secs it takes to complete the test... but im sure it would sky rocket if i did the 1024 test... but i'm not even going to try that lol

 

CPU vCore increase would give you more stability if you were overclocking via SetFSB, but the thing with adding voltage is that it heats up the CPU quick. When running benches, I've run my CPU at 125W in throttle stop before. I've done that for wPrime1024 before. I posted my 920XM scores on HWbot.org. I think thats about what I had with wPrime32M when I ran a 6.599sec run. Don't remember. But as far as everyday use, I don't use throttle stop just stock. But when benching, it's all about keeping the CPU temp down as low as you can to prevent throttling.

You are right though, increasing voltage will cause the CPU to throttle. The cause is the fact that the BIOS is set to instruct the OS to throttle the CPU at 95*C, not the 100*C as others have stated.

 

The official Intel thermal throttling temperature for the XM CPUs is 100C but it's possible for laptop manufacturers to set the bios so it starts throttling at a temperature lower than that. It's probably not a good idea but maybe I can encourage the bios to let you guys get a few more degrees out of your CPU before this bios induced throttling starts. It might be the difference between a good run and a new world record.