*Official* Alienware M18xR1/R2 CPU Overclocking Thread - Learn How and Share Tips Here

I tried a 2920XM (OEM) and could only stay at 3.7Ghz max during the 8 threads run. Here, it's a 300mhz higher under the same conditions. At least, I'm able to finish the 32Mb run @4ghz 8threads.

 

Yeah, a few 2920XM's I tried would drop to 3.6-3.7ghz on an 8 thread run of anything taxing. They would also fail to peak to hit 4.5ghz on dual and single benches. This last one I have here is the only one so far to hit and hold 4.5ghz single and dual along with 3.9-4.3 (~4ghz avg) on 8 thread benches.

I was just wondering if you can get the 2960xm to hold 4.5ghz on single and dual threaded runs along with the ~4ghz avg on 8 threads.

If you could, give them a run along with Cinebench R10 64-bit Single and multithread bench run and scores.

 

Can't hold 4.5 on single at those multi's If I set em all 45-45-45-45 then yes, but multithreaded drop to stock.

 

So if you set your 2960xm to 45-45-45-45, level 3, max 'er out with stock A03 BIOS you can:

4.5ghz single core yes?
4.5ghz dual core yes?
4.0ghz four core yes?
3.9-4.3ghz eight core yes?

 

Cinebench R10 - 64 bit:

Single - 6660
Multi - 24352

Cinebench 11.5 - 64bit:

CPU - 7.44

 

Only single core hits 4.5.

all the rest fall back to stock (34-34-36)

 

it sounds like the 2960xm has a bit better oc ponential over the 2920xm in general.
would you guys agree?

 

Not from what I'm seeing so far. It seems to come down to the chip like 920xm/940xm days all over again. I think I'm going to hold onto this last 2920xm OEM, as the stock A03 Level 3 max 45x4 performance is blasting my previous 4 2920xm OEMs tested and so far beating 2960's. It did 7.2k+ / 25k+ in Cinebench R10 64-bit without touching Bclk. I'll take a look at that next.

If more 2960xm owners start submitting some baseline benchies with the STOCK Dell A03 BIOS (and then we'll throw in Johnksss "TO THE WALL!!!!" benches on his 2960xm in a few weeks with the modded BIOS) with a level 3 45-45-45-45 max OC, we'll get more data.

 

hummm, there is no way he should be behind a 2920xm while running a stock bios...it should always be slightly ahead....going to have to chalk that up to his os in general.


and i too agree..we need more people to test.

 

True, Aiki, weren't you having some sub optimal scores with your ES 2.4ghz 2920xm too?

Stock A03, but running with OC settings. I wanted him to set it to max Stock BIOS A03 OC settings 45's across the board, no Bclk adjustments to see what scores he was getting. The 5 2920XM's I tested under same conditions yielded Cinebench R10 64-bit scores ranging from 6.2k to 7.3k and Multi of ~23k -> 25.5k so there are variances in the chips as always when OC'ing. One even BSODed with the same settings as the others when pushed to a 45x multi and I had to dial it back.

Before you bring it to its knees , I'd like to see what scores you get in a stock BIOS Level 3 OC max no Bclk setting for comparisons.

The primary objective is to be able to hold a 45x multi on single and dual threaded and 40x+ on 4 threaded and 39x on 8 thread which is my baseline.

 

Ok so after setting all the multis to 45 I'm getting 45 on 1-4 threads, and 4.0ghz on 6-8 threads(prime95, wprime).
Cinebench r10 64bit - single 7429, multi 25907. Cinebench r11.5 64 bit - CPU 7.64.

Also, at those clocks the temps are under 85 in games. Room 24,5 elevated back.

Oh, and I had both OEM an es 2920 tested under the same conditions. Both showed much higher temps at lower clocks.

 

unlocked menu or still stock?

 

All stock.

 

Yeah, those sound right on target with my 2920xm and great temps. I clocked a 7321 and 25624 so they're basically equal. All on stock A03.

 

Technically, those aren't right on target. See the attached pic.
Now, even if you can match the 2960xm that only makes 1 out of 7 chips with that potential. You said it yourself, it took you 5 CPU's to get one good OC'ing chip. I checked 2 2920xm chips so far and neither could hold those clocks. So even if got lucky with your CPU it's still slightly behind my 2960XM (provided we are on the same page with the testing conditions). My conclusion is obvious, - the 2960XM will most likely have a 200mhz edge in any scenario till we hit 49x4 + bus OC, but that's something I'd rather observe from a distance and let true benchers like John and Scook do their best.

BTW, did you set the Flex 0 or 25?

 

What flex did you use Aikimox for 4.5 4Thread to 4.0ghZ 8thread prime95 annd wprime? I want to compare using the limitations of the stock bios for you with no fan control used at room temp of 75+ and back raised like you mentioned. So I'll set 45/45/45/45, the flex you end up telling me and a baseclock of 100mhz or 99.8 set via bios? When set in bios my baseclock is always a bit under what I set in bios... 100.13 in bios gives me 100mhz in windows etc. So 99.8 in bios us less than 99.8 in windows. These chips I think are equal when set the same. The fact that 2920xm had 5 bad chips is not a way to compare I've only had my one that came with the system and had not had bad temps for the clocks I'm setting on stock paste or on ICD24 which I am using now. The preset "base speed" of the 2960xm is the only difference they are the same silicon and build. I posted 4.8 and 4.9 8thread wprime 32m and even with a lighter load than that test 8thread at 5ghz on stock baseclock. you dont have to submit your chip to that kind of testing but I know John will and I hope to see what he gets so as to match it with my 2920xm... I was planning to upgrade to the 2960xm and sell my 2920xm... But I haven't seen anything by the 2960XM that is any faster than the 2920xm. The only comparison that fairly compares the two is when overclocked and so far 2960Xm chips haven't had any submissions of benches showing higher overclocks, but there have been many high benchmarks made so far on the 2920xm since so afr the 2920xm owners that are willing to push their chips have submitted scores etc... So we really need any user at all to push their 2960xm to equal or better speeds than the 2920xm. Bios limits aren't the processor limits. We certainly can compare speeds at same settings and get a fair comparison at that level... But if we are all talking about which chip is more capable it is that they are both equal until benched higher to prove it. aikimox whatever flex you tell me I'll compare for you... can't judge all 2920 by a couple duds... How many 2960xm will be dud overclockers... Who knows not enough have them yet.

 

Yeah, I could use better memory too. I haven't bitten the bullet for 1866. But yes, what I am saying is all previous 2920xm's I tested were subpar to the one I'm using/testing now. And the scores are equal, considering I held the same frequencies as your 2960xm under the same conditions which was the idea goal I was testing.

The previous chips just wouldn't hit the same frequencies regardless of heat or settings. They just hit the aforementioned walls and sat there up to and including one that would outright BSOD.

That's not to say any and all 2960xm's will hit the same peaks your 2960xm and my 2920xm is hitting under stock conditions. Residual is right, and it is in line with my original hypothesis:

In an overclock scenario, using stock A03 Overclock options available in the M18x, the 2920xm = 2960xm.

7426 vs 7321 = 1.43% gain
26039 vs 25624 1.6% gain

More data is needed to fully prove or disprove this scenario, but so far, they are basically equal and a 920xm=940xm scenario again and it comes down to the quality of the chip/batch/run.

Any real gains/differences between the chips will show when the big boy overclockers get ahold of them and go to the extremes (ba bump bump) with the modded BIOS and cooling.

But for everyday use with the stock Alienware overclocking bios options, there is no tangible gains using a 2960xm over the 2920xm dependent on the chip from either.

 

My settings:
Stock A03 BIOS
45-45-45-45
Long Duration PWR 80
Time Window 56
Short Duration PWR 99
Flex 0
Stock bus/baseclock.
Room 77f
Elevated back
Fans AUTO (using only one monitoring program - Throttlestop, everything stock)
Max temps in Wprime1024: 86C
Max temps in Prime95, 10min run: 87C
Thermal Material: ICD24
Stock 2-piped HS.

Stable clocks (Wprime1024/Prime95):
1core/2threads - 4.5ghz
2core/4threads - 4.5Ghz
3core/6threads - 4Ghz
4core/8threads - 4Ghz then after 20sec 3.9Ghz.

 

dont get me wrong...but these test aren't really proving anything...out the box..anything set at a higher speed will be faster. simple math. but it's not till you over clock the two that you see the separation in who is bos. just like the 920/940. for all sakes and purposes..all they did was up it two multipliers. so for the 2960xm to be better..it needs to match or beat a 2920 xm at the stage that we have already set. and so far...it hasn't come remotely close yet.
so that means:
wprime155
superpi
pifast
cpuz
or cinebench in a full over clocked state.


this other type of testing wont prove anything but that they are just that. 200 mhz apart. and in some testing...the numbers will be slightly bigger than others.

side note:
need to see actual scores for wprime.

examples:
5k cpuz
167 sec wprime 1024
5.1 sec wprime 32
vantage cpu 31.5k
3dmark06 cpu 8600 windows 7
super pi 7.5 secs
3dmark11 10k physics

so until those numbers are surpassed...the 2920xm is top dog.

 

That's exactly my point, they should stay 200mhz apart till you hit the limit and then it's a lottery. One thing though (which is irrelevant for you as a bencher after seeing your cooling system, lol) is the temps at equal OC'.

Here's a couple of Wprime scores:

 

That scenario is only applicable at stock clocks. Once you start overclocking, they effectively become equal in regards to access to same overclocking settings. The quality of the individual chip is what will set one chip apart from the other.

The question is do the 2960xm's coming off the line have more overclocking potential than 2920xm's? It took me 5 chips to find an equal chip with a variance of scores that were up to 15% slower. You tested two 2920xm'sand they were worse too, so will most 2960xm's come right out of the gate achieving your scores versus having to find the occasional 2920xm that will match it? Like I said, right now the results are looking fairly equal, but we need more 2960xm data.

 

True, we need more data, and will be getting it soon as there's a bunch of 2960 orders getting shipped to our clients, a few of those are active forum members

 

it just like when i was binning 2600k chips. i went through 12 chips just to keep one that was a 5600 mhz chip. does not mater what they all do at stock since any one can achieve the same results...it's when it goes past it's range that you find out what the chip is made of. even if your chip bsod doesn't mean it's junk...only that it needs more or less voltage to operate. they do not all operate the same in this aspect. some may look good with low voltage, but wont over clock for a damn thing. some take much more voltage and are over clocking bandits...and vice versa. that's why the chip has to be pushed to find out if it's worth anything. on this scale...you can contend with temps and who was lucky enough to go through the test. your not putting a hella strain on it. it has to complete the whole test at the lowest multiplier set. if it drops off that. the over clock isn't good or your temps are too high for the over clock and you must lower it till you can complete the whole test at the same multiplier. hell, my es is listed as the second fasted chip here. and it's the first gen chip...what does that say for oem? and it's only 2400 mhz.


but i do see that you're trying to see what they do at the same settings and all...but this becomes luck of the draw in the long run.... just my opinion is all though

 

Exactly give the same playing field and give them unlimited field to run will show which chip is top dog or if they are equal. When the overclocks are in low range it will only come to cooling etc and not which chip ran faster.. so as soon as a 2960 user steps up and tries 5ghz or gets over 4638mhz 1024m 8t wprime then we are ready to compare results. Just one brave soul to join me and John in the twilight zone? Step up and show us what a 2960 can do we have all been waiting ever since its release. Me and John seem to be bleeding edge guys lol joking now we need a third with a 2960 or John to recover his.

 

There are two different goals here:

1) Max stable and 24/7 OC' @ stock cooling.

2) Max possible OC' for submission

I'm only going after 1) at this point.

 

And given the right cooling they should both perform very similar with manual OC...

 

this is true...and since one is only clocked 200 mhz faster. under the exact same conditions..there is no way the 2920xm should catch a 2960xm. meaning
test are ran at chips respectable clocks.
2.5 vs 2.7 ghz or 3.5 vs 3.7 ghz.
if ran on the exact same machine....the 2960xm should naturally always finish on top. whether it be 1% or 20%.
and technically speaking..the 2960 should run hotter. given it's running faster. and these are the same chips with one being clocked higher is all.

as to over clocking...that remains to be seen.

 

True, I separate the two. #1 is my primary objective, but #2 is for fun and bragging rights. Here is a run on my 2920xm stock A03 Level 3 max Wprime 1.55, cinebench R10 and CPUZ, no Bclk adjustments at all, aka everyday use scenario. Like I said, I'm not seeing the 2960xm in normal stock A03 OC conditions beating this 2920xm.

 

Exactly and for the 24/7 overclock I bet we can make both 2920xm and the 2960xm do the same 24/7 under same cooling and settings. So they should be equal in example 1 and when we all agree on a 24/7 overclock we can test it on both and we should get similar scores... temps won't be good comparison because every single paste job handles different and every paste material handles different and other than having similar average ambient temps it is nearly impossible to make the same conditions between two systems around the world. So the results have to be taken with +/- 3% variation considered for conditions that will never be precisely the same.

 

interesting, whats your flex in the bios? level 3 oc puts it to 25.

Also from your runs all my Cine results were still higher as well the wprime32.

Edit: I'd also like to see your BIOS screenshot in order to set mine to the exact same settings.

 

My 1024 score is lower (203.585 vs. 207.559)
Your 32m is lower (6.458 vs. 6.73)

Cinebench R10 Single = 7395 vs 7426
Cinebench R10 Multi = 25776 vs 26039
+ you using better memory (Which Cinebench takes advantage of)

I know you know these chips are equal in our given criteria (A03 Stock, Level 3 normal day to day use).

Flex is what it defaults to, so I would say it is 25.

I'm going to start working on the Bclk later to see how high I can get that stable for day to day use.

 

My flex is 0. Ok will try your setting s later on. BTW, what's your superpi score at those settings?

 

Guys one thing remember when comparing... dual channel and single channel ram make some differences as well as amount of GB of ram and plus 1600mgz vs 1866... Even ram volts can make differences. I have 1333 ram rated at 1600mhz at 1.65v ram that's natively 1600mhz will do better and higher even better. Especially if a ram is rated high in MHz and rated at low volts like 1.5v... that would be the best scenario. So its still hard to compare Aikimox even if you set 1600mhz your ram will have an edge.

 

How much of a loss is there with the ES 2920XM compared to the retail 2920XM? I ask this because such stepping differences can affect many cache, variances in set associativity of the cache, prefetch, TBL, pipeline stall level sensitive simulations and similar tools. Might not be apparent in some common day to day tasks. But in some specialized applications the differences between an old ES and the retail chips can be significant.

 

All presets are editable with the + and - buttons of the keyboard. So for comparison you both can enter same values so flex, short limit, long limit, time limit, and the 4 multipliers are set the same. So both of you can take pic of bios set the same way... So nothing is different and even if Aiki your ram is faster programming it to same XMP 2 @ 1600mhz for both of you would be better comparison except Electrosoft has 8gb and you have 16gb... But for comparison in the least you should set 1600mhz not 1866 or its definitely in no way an even playing field.

Its a shame two members don't have identical ram, HD, and parts except only difference being 2920xm and 2960xm since that would give non skewed results.

 

well at first it was seeming ES was limited but if you take my OEM and Johns ES we can see a great comparison and the results are they are nearly matched so if John would jump in and elaborate I think it will be clear that we both thrashed these chips and reached about the same plateau. John mentioned he noticed just a small difference I think so if there was one I'll let John elaborate on that right now. For 200-$300 vs $900 its a clear choice to save money as most people won't see a big difference.

 

True, I'm running 4GB of HyperX 1600mhz RAM. I think I noted in a previous post the ram discrepancy.

I'm going to lower the flex in increments and run some tests later to see if it will help raise the multi's/scores.

 

cool
reference: 4.6ghz ~ wprime 1024m ~ 14flex
4638mhz ~ wprime 1024m ~15flex (personal experience)
0flex = coolest chip but needs to be raised as higher quad speeds are used. Over 4.5ghz single will need a Voltage bump too but the Quad speeds require it faster.

 

(not directed at you...just answering from here... )


for 300 bucks, the chip is worth that and more! hands down. it can run with the big dawgs! and it can crawl with the pup's (4.5 to 7.5 watts).
what i found to be different.
max voltage 1.351V
max watts on anything but prime95 85 watts. (prime95 about 90 watts it would touch every now and then)
oem has a .150V increase over it along with pulling more watts.
but yet it's still faster than every single other mobile chip out.(other than 2920xm/2960xm oem's)

 

Looks like I'm not even close to the limit @ stock BIOS. Managed to get 6.3sec in Wprime32 (4.1Ghz through the whole run). The 1024 test is screwed at the end because of the cooling limitation. I have 25.5c in the room and all sensors get past 85C, so it throttles. But the beginning of the run was good (4.1-4.0ghz for 20%, then 3.9Ghz for the rest till the temps get out of whack.
Flex is still 0 BTW

 

Here's a little bump, still stock BIOS

 

That is a micro bump. They are equal and you know it.

 

But mine's got the "6" which makes it superior by default

 

Lol I actually like your answer its got the right spunk and drive lol. So I say it's official Aikimox...lol... your the Official 2960XM Spokesperson lol. Cheers.

 

Let me help Aikimox its this simple set your long watt limit just one or two watts lower so that when the end of the wprime run comes it slows down a little by then from heat but keeps it low enough for your goal. That's why we have a short long limit and its all adjustable hone it in on what will keep you speeds exactly where you want them don't go by what it appears to be letting you do and just set it... your the boss and that XM is your b-itch pardon my french lol

 

John and I have a small theory on that since ES is test version maybe the reading had a 1.3511 volt max on the chip because you were pulling clocks that require much more than 1.3511 so maybe it is a glitch where the reading is off from actual voltage which could be confirmed if someone attached leads like SVL7 did on a pinmod for a 920XM to read voltage from the chip itself over a voltmeter... John you returned yours but maybe Speedy or someone else with the ES could try it... in theory on my oem it appears to be 1.351V= 0flex and 25 flex = 1.49V so not sure but you still got to the same levels

 

((Kicks the dirt)) foiled again!

 

"paints a 6 over the 2" ha ha lol I could edit the hardware id in my registry so it would show a 6 instead lol joking. That's like painting an 11 on my amp and cranking it to it lol ha ha Spinal Tap lol

 

http://forum.notebookreview.com/ali...are-m18x-benchmark-thread-55.html#post7957593

here is more 2920 for comparison to 2960 when overclocked, to add to what John listed above already... plenty more to come when I have time... then about 2 weeks I'll be using AC and seeing true max... by then John will have his 2960XM...

 

Did some flex runs. So far Flex 10 provides the best results consistently.

As always stock A03 level 3.

Flex 10 let me break under 200 for 1024 run and sub 6.6 on 32m. In addition, Cinebench multi over 26k.

Edit: Image 1 is Flex 15. Image 2 is Flex 10. Image 3 is Bclk 103