*OFFICIAL* Alienware M18x Benchmark Thread - Part 2

. And at the end of the day...It was just a better binned chip....
this is okay i dont care i know this.

well all chips are not equal... yes. not all chips reach speeds as high as others. I think if a chip has the same potential... that potential is not fully used just plugging anything in... you know how to overclock you have more systems under your belt then anyone i know... so there really is a methodology to all this and those here that are still learning look to people like you who have the knowledge and experience to help them reach for the same thing you have accomplished.

Pertaining to the above I am getting at yes your pic shows it is sitting in windows idle at 5.1 max turbo... running at 5050mhz... put a load on it it still will drop... raise your amps and watts higher than you have too much and you may crash since even heavy loads chip would not drop and perhaps would not stay stable on the desktop or in everyday apps. you stated at beggining of the post with all the scores if we are testing well thats another thing... well in my case that's all it is. We can run 5ghz all day as long as enough settings are correct if we set higher amps and watts it may require more vid and eventually not be enough to stay stable. so binning aside as in we already know you have a good binned chip... I just would like to see more behavior of you cpu in other scenarios than the official benches and sitting idle on desktop i'll include a few of mine as example. I already posted that one of 4.7ghz in Intel burn showing 205 watt peak thats with the igp's added watts over 8threads. Probably wont see anything else out there making that kind of load or atleast running much more than that at that load. this particular thing i mention was because I knew 4.4ghz 0flex can run at just that in ibt 8thread if amps and watts are high enough... so whats the next level? well you can probably go over the 4.7ghz i shown but just sitting on my desk thats all i could work to without hitting thermal shutdown temp ranges. If i am only one here that cares its okay i don't mind . if you want to join in for curiosity cool. even if it all comes down to if a chips capable or not... if you didn't have a plan plugging in just anything wont get that person in your example very far. like mine seems to max at 5090mhz... there's a chance though it can go higher and i had yet to boot it right... Shaden's ES might go higher than 4.5ghz... just because it seems to be the max it may be more sensitive than others perhaps temps are still an issue.

Spoiler

here we see Intel Processor Identification Utility 5ghz 100% 8threads with 85 watts


here we see Prime95 5ghz 1threads large fft with 60 watts


here we see a dual core bench Extreme CPU Benchmark by AgentGod (like Intel Burn Test) with hwinfo graphing the 4 cores speeds though a dual core bench all 4 cores were running in the 5ghz~


here we see Unclewebbs Load Tester single core showing max 99flex which was not required just was reference

if you wonder why all this... it may not be necessary and may make it seem pointless when someone gets lucky and just plugs in a world record (cough John cough lol)... but for others who are not so lucky we have each others findings to go by... you got a new chip and it got you where you are the rest of us still put the effort in anyway... if Shaden was to just give up and say oh my chip can only do 4.5ghz I think it would be a mistake... these things are fickle sometimes and each goal with them has its own best settings to use... then again... some chips never go beyond what was first tested... Shaden can get all this down and get a better binned chip and would be able to accomplish a lot because of taking the time to learn all of this... perhaps you thought I cared about this because of benching I should have been clearer... by benching i can see the systems handling so after all scores are acquired and all is said in done i don't care much about having the top bench i will be using all my findings to set up the cpu on a permanent overclock that's all. i might go out and replace my 2920xm for a 2960xm but it will help knowing what it should do and what it could do.

Edit: John if 2960xm yours can do that Intel processor identification utility at 5.1 -5.26 ghz at 100% load on all 8 threads than it may very well be that the 2960xm's in general are better binned... please show me us if it does sometime... since that would mean it handles light 4core 5.1-5.26ghz... and drops to 4933 under a load as much as wprime 8thread. that is useful knowledge... hey if it does that you can probably run 52/52/51/50 without a prob...

 

Fox, To verify, (you have the same clock speeds); What is the difference between Graphic test 1 and Graphic test 3? Is it .87 volt vs the .92 volt?

 

@residualvoltage
You really sure you want to go with that post huh?

 

No, all of the tests were at the same voltage (0.92v), just different clock speeds. The test 1 and test 2 are the benchmark runs within 3DMark11... the first test and the third test. My point was identifying which clock speeds produced the best results and remained stable. Those in green were the "sweet spot" so to speak, at least in my own system. YMMV, but those may be good starting points to try with.

 

It was strange with the 10 fps difference. Now even stranger if they where at the same (Flashed .92v). How come a 10 fps gain? Also where is test 2?

 

did u try 3dmark11 before ?

take a look at this
it's not the same test/map so ..

 

Devil.... LOL I see GPU test 1-2-3-4 GOT IT!!! And yes Devil, I know all about CPU, GPU, etc.

I just pay attention to the overall score

I am a GURU .....

 

I'm not a guru by any means, but the testing I did was to identify the best (highest) speeds to operate the cards in my system without heat or stability issues. I chose tests 1 and 3 to help with that. Tests 2 and 4 were skipped for the sake of expediency.

 

Apparently you haven’t figured out that you can not run anything above 5000 ghz on full load 8 threads for very long. And no one is going to be running no 5 ghz anything in here off the desktop for everyday use. And that intel turbo boost will not show higher than 5.0 on anything more than 2 cores. The M18X is setup that way and at this point and time. No way to get around it. Also. most people would be very content on a realistic over clock of 45x on all 8 threads. Which is very do able. But trying to sell people on extreme settings for every day use is just not realistic. Sorry, but it's not. Now i have personally tested at least 6 different xm cpus from different people. And just as stated before. It's the chip itself. With sandy Bridge a chip can do a certain multiplier right out the door with out you doing anything spectacular. And those simple settings are:
45x on all cores
10 flex
1000 power plane
80W
80W
56 Turbo time

Then use xtu to up the multipliers to the next stage then reboot and see what happens. Or you can bclk your way up to see if the next multiplier will work.
At this point is when the work starts. When you see how high you can get into window. then you have something to shoot for for stability, if that's what the user is after.

That is a standard that everyone can use when they flash the unlocked bios. Now for those not flashing. 37x@8 Threads Give or take.

Now im all for testing things because that's what i do, but as far as this being really useful. it's not. Just because our settings work for us... Does not mean it will work for anyone else. And that is a fact. Prime example. Scook9 can only do a 45x on his OEM 2920xm. And we tried everything. So if you drop your goal record a little..more people can get on baord, but trying to sell people on 50x...Not gonna happen.

Now I son't want you taking this post in the wrong context or anything. Because that's not my intentions.....but a little more realistic over clocks will go allot farther than 50x. Im just sayin'

 

okay my bad! sry
maybe you was trying to referencing to a benchmark with the difference of 10fps higher than Mr.fox's benchmarks.

it's just ... if you look at it in a different way it feels like you was trying to ask for it ^^.

 

Only because test 1 is more demanding on the system resources than test 3.

 

She should be able to run one of the settings I have listed on the other page, but for nvidia...you really need a super high over clock to compete with amd. and it's still not good enough.(3dmark11)



Side note:
Prior Post.
5.1 ghz and 5.2 ghz on 2 cores

 

I'm not 100% positive about this, but I have a hunch that is because you can totally disable more eye candy stuff (such as tess) with AMD than you can with NVIDIA. I think you still end up with NVIDIA eye candy slowing down the benchmark a little bit when all settings are high performance because you are turning it down rather than shutting it off.

You can really see that in Heaven and 3DMark11 with AMD cards. Certain elements (such as the ground and some vegetation - LOL - and cannons on the ship) completely disappear with AMD cards when everything is shut off. There is a whole lot less for the AMD cards to have to render because it is skipping over some stuff that the NVIDIA cards go ahead and render.

 

Yep. It was the same with LOD, but since everyone could do it.(amd & nvidia) it was allowed in the end. I did a run with tess at about 30 to 40 percent off and got a 9.1k gpu score on nvidia cards. So you can only imagine what it would do with it completely off.(3dmark11)

Spoiler

URL=http://imageshack.us/photo/my-images/716/capture12l.jpg/] [/URL]

 

I just got some time from the holidays to start playing with the CPU. Now I have the unlocked A05 BIOS and I have the Pri Plane set to 0 for automatic. My Memory is 1866 but I have it set for auto as well. The CPU doesn't even like 100.58 let alone anything higher. I have tried with my memory set to 1333 etc but no dice. I feel dumb but when I had my AMD system I was able to get the BUS speed to 105 easily in the stock BIOS. Sorry if I sound like a total noob.

 

Well getting some perspective can always help... i got no one to bounce things off of... I appreciate the constructive criticisms.

I definitely have trouble saying what I mean. The overclocking then only comes in two flavors over the top and optimally cool... in that case it goes with the basic formula you note above... and the top speed 8 thread encoding wise is going to be the highest 8thread turbo multi that when run under heavy load does not hit max temps... so it will require the lowest or 1 flex away from the lowest stable based on high load scenarios. This brings me to testing i did do if one can run IBT for long term 8thread said setup should be sufficient for encoding. So technically 4.8-4.9 is a bit high for such a load long term... sure you can have a higher flex and set a long watt limit to accommodate for heat... but we are best using the max IBT could handle since it could then encode at those speeds for long durations without trouble. So in short term testing the reference of 0flex is based on IBT 8thread with high amps but 97 watts long limit. As long as its 0 flex that always stayed cool enough for me to complete about two runs. So i propose a logical overclock 8thread is 4.4ghz-4.6~ for encoding. For people that can take extra heat and want higher speed in all ranges except the 8thread a 4.6-4.933 will be the most for a higher everyday overclock but will need the wattage and amp limit lower than the low flex overclock above to combat the added heat... and that way under the same 8 thread load it would drop to a stable speed but alas it will be a lower speed in 8 thread and if its IBT like above for the flex need for 4.8-4.933 then the cpu wont run at 4.4-4.6~ it will be in the 3.7-4.1 ghz range at that flex required or if you raise the amps and watts higher it can maintain higher speeds but that's not likely as those temps would render the overclock useless like you said it does no good if it cant calculate anything for its efforts.
Seems the first is the best and i ran the 4.4ghz for about 2 months as my everyday clocks it was well worth it especially comparing temps i would be seeing if i used too high an overclock.

 

Yes, I agree. Mai should be able to run up to (at the very least) 750/1500/1600 speeds at the Dell under-volted settings and should be rock solid. She might even do 790/1580/1600 or more without an issue. For my system, and I think I remember seeing Xeneize comment the same thing, around 800MHz and higher, there are stability issues unless the voltage is raised to 0.92v NVIDIA reference spec.

OT: What wallpaper are you using, John? The part that I can see looks very interesting, but I don't recognize it as Alienware wallpaper.

 

Now this makes much more sense! You and I can push the boundaries, but we still have to have the normal reachable stuff for everyone else..


So some more testing. No matter what 5.0 ghz drops to anywhere between 4.8 to 4.9333 ghz on 8 threads, but will hold if you drop HT. Which Allot of people sometimes do(desktop world). and the numbers are still pretty close.


Brother Fox: random form the net

 

I have ram that is sensitive to overclock could just be mine but i need 1.55v minimum and once i start to overclock that baseclock mine needs higher. Also XMP2 always works for me where as 1333 timings for xmp1 seemed tighter and less stable with the baseclock overclocked... so then auto on the timings should be best but for me it wasn't the case... I can't go over 105.13 via bios. Mine is Kingston HyperX 9-9-9-24 but its the basic 1.65v rated version not the low volt version. for long term stability baseclock is always iffy... but if just benching then try to see if its one of the above... if you tried all that then ignore my suggestions.

 

Gracias, bro. Cool looking, indeed.

I could set higher BCLK with AMD GPUs in my system also. I think it may have something to do with the GTX 580M cards possible requiring more current than 6790M/6990M. I could set BCLK to 106 with AMD GPUs and 104.6 is about my limit now before the system stopped booting. I also found increasing memory voltage to 1.55v or 1.60v is necessary when moving my BCLK up. Leaving it set to 1.5v or auto doesn't yield good results for me. This is with a non-XM CPU with locked multipliers, but my experience in relation to the above sounds very similar.

 

I think i got a 105 to 106 out of my amd setup as well. And 105 with this setup.

 

well sometimes it takes a friend to ground ya lol cool. yes and the numbers are probably very nice with the 4 core no HT. so yes all intensive purposes those are the ranges of 8thread.
4990 on 8threads wont drop @50flex high amp and watt under less load than TSbench, wprime etc etc... but serves no purpose because as soon as a few seconds go by it hits thermal shut off for me thats why pic is at 71C max... it only sustains for the entire test of that app. It drops yes to the 4.8-4.933 if its a heavier load and if not 99c is reached if IPIU ran longer. So for real use we have to keep the amp and watt lower than that i this case and therefore who cares about 4933-4990 8thread its such a minuscule difference. I think you can get it John but as has been determined it is not for real use but perhaps that you can say the range is 4.8-4.990 8thread max... otherwise it is as is 4.933 8t max as seen by wprime 8thread load...

im sure you have ran into the app. when launched for about 2-3 secs the cpu is under an 8thread load on a 8thread system of like linx or something but it it the lowest 8thread load i have ever seen on an app to date... apparently this version just came out this month.
http://downloadcenter.intel.com/Detail_Desc.aspx?ProductID=1881&DwnldID=7838&lang=eng&iid=dc_rss

So 2 core 2 thread 5.269 - 5000mhz max thus far. I am still amazed by 53x registering but i see it worked only in the same way i used 51x... the lowest baseclock available of 99.4 is great for those validations but even helping make that super pi run stable.
i wish i wrote the amps/watts but i do know although overkill 300amp and 300 watt with the 50 flex was sufficient to get that result and without a doubt doesn't need that but i forget what i had used. So do you think you'll attempt 53x at 99.5,.6,.7 etc?

as for baseclock perhaps nvidia pci lanes are more complicated than the amd systems? Plus nvidia has a lot tied in compared to AMD single feature of directe compute.

 

The key. It's not based off of the multiplier. it's based off the speed. 53x was just to see if it was possible. But this chip is considered a 52x (5200 mhz or so) chip max and bench able at 49x (4900mhz or so).

Also that .4 drop in blck gave me 4 mhz vs 52x with higher bclk. This tells me what the max the chip is. And I will not be doing dry ice on this one.

 

John, Is it possible that folding warped my heatsink. It was folding fine for a week and a half straight. Than I took the system apart to adjust my cable connecting the express card slot board. Didn't touch the heatsink or or anything. Put the system back together and all of a sudden Im overheating. It would top out at 100% load around 83*C and now it would skyrocket to 95 with a slight load. Adjusting the heatsink did nothing. Tried a 920xm instead of the 940xm and same thing. Now Im folding on the 540m and getting ready to order a heatsink. I can't explain the sudden heat dissipation issues. Even the 540m is hitting 70* at 100% load. Not dangerous, but high for a 540m.

Any ideas? New heatsink is my only guess. After that maybe a mobo? Hopefully the parts or bracket that the heat sink screws into isn't wonky.

 

Did you try a bios reset, then work back up to your normal over clock settings? Shaden and I had the same thing happen or at least i would like to think so. We did the same things you did and still high temps. Finally did a bios reset and ran c1e and the system dropped back in line.

 

That's the thing, Its all stock. Ill try again when I get back home though. Ill clear the cmos. I will kiss you if that works.

 

Very weird indeed, but we feel your pain.

 

DR650 that's frustrating in the least... perhaps the heat did in fact cause those issues. I hope you get to the bottom of this soon.

John I sent you PM. I apologize for the way i spoke earlier it was definitely uncalled for.

 

No worries, we are good Brother RV.

 

@Res, Thanks yea it is kinda frustrating to not be able to solve something. It eats at me

Well tomorrow I order a heatsink. Tired resetting the bios and some other things, like cleaning extra well, and nothing Hopefully a new heatsink fixes the issue.

 

does it looked warped?

 

Not at all. I'm thinking maybe it's bent along the pipes. So when it's screwed down, it isn't sitting flat. I know my last heatsink was like that and a new one helped. But this one was working fine one min, and then I opened the system. didn't even remove the frame above the heatsink, and then when i put it back together I have the heat issue. I ruled out the CPU as the cuplrit as my 540M runs hotter than it should. At only 25w, its going above 70°C and I tried a 920XM. Same thing.

 

This happened to me when I had an M17x-R2 and I did the retention mod. I fixed it by bending the heatsink with my hands until the temperatures leveled out.

 

So yes could be bending on fin end causing heatsink to not sit right, could be the tension was much greater and suddenly its lower and would warrant enhancing the tension on the heatsink. only other thing is perhaps the nights on end of folding at home caused damage to to cooling surface of your heatsink did you visually inspect the under surface that touches the chip before repasting and putting in that non-xm? In this case if it were the issue would lapping help? On my 2920 i know a few times i ran testing unnecessarily hot and noticed rough black spot or so on that surface upon repaste... Seems a heatsinks cooling properties can be fleeting sometimes?

 

Thanks guys, I may fiddle with it tonight. I just ordered another heatsink, some thermal pads, and some AS5 so once that arrives, hopefully I'll be good again. In the meantime I may try and tweak the heatsink to a better fit. I figure with a heatsink on the way I have nothing to lose

Just a little observation, Why does Mr. Fox not have a Pifast score? Easy 5 pts buddy!

 

Good question, DR... just laziness on my part. I had downloaded an old version and could not get it to work correctly. I looked at John's second screen shot and saw he was using hexus pifast so I found that this morning. So... here goes, for 5 easy points.

looks like 21.69... let me try again setting affinity and priority

Spoiler

meh... affinity on 4 cores w/o HT and priority set to real time only shaved it to 20.92

Spoiler

The second run is without the batch file, just using my own syntax at the cmd prompt. (Otherwise, I could not set affinity or priority because the program would launch and finish on its own.) Is that kosher, or must the batch file be used? If not, then I guess 21.69 is the "official" score.

Edit: just noticed the 3.66GHz reading in CPU-Z on the second run. It's not very often that the 2760QM hits the 35x multiplier long enough to see it.

 

Nice job Mr. Fox. Glad you got you pifast done.

DDR650 gonna attempt a lapping? Maybe a drunken lapping hahaha.

Lol I wrote DDr650 hahaha

 

lol. I may, but I took the easy route and just bought a heatsink. If that doesn't work I won't know what to do quite yet.

 

Lol I mean the heatsink you have now. Lol do people only lap CPU surface or some do it to heatsink surface too?

 

Lapping the heatsink would make sense. It would just be more difficult to do. And CPU laping is more effective in the desktop world. I have a few desktop CPUs and a mobo. If I ever get the other components I'll lap my desktop chips and bench them with a dice cup. I just never have time for it But yea lapping both surfaces would be best to get the most contact when done right.

 

Cool so if you had time before new heatsink gets there you could lap it and see if it helped the temp issues. Could be the heat from folding at home lowered the surfaces heat wicking properties... I know a soldering iron tip when heated and not taken care of properly loses its ability to properly transfer heat from the tip to components. When learning to use one I actually melted a tip before I knew about proper care.
I guess for mobile CPU it could be a big risk since the CPU die surface is most likely thinner. I know I wouldn't lap a mobile CPU myself but but the heatsink sounds like a good idea. I think for most people CPU lapping doesn't make sense. But for desktop CPUs I can see its uses for some people.

 

I am going to be selling my 2920es

I am thinking of moving it next week...

 

why you seling it?

 

Getting a upgrade Shaden? If so did you find a deal or buying direct from dell so warranty applies?

 

Yes I am upgrading further, lol crazy perhaps, but I am ... I have had it running as high as 4.5GHz in benches, and I might be hitting 4.7Ghz tonight ... we will see. Just playing with flex/bclk atm .. I am close though

Have not decided if I am going 3rd party or Dell .. but either way ... I am selling it.

 

Going for that upgrade are we. Got your feet a bit wet and now looking for more. I can relate to that!

 

Hrm, new PB ... Still not quite hittin my 4.7GHz ... seems to just flake out ... or drop out of multi's ... First CPU test it will hold @ 45x ... second it drops to 43x ... annoying ... if I increase flex, both drop out further, if I decrease, both drop out further

BUT ... here is the PB for today.

Spoiler

 

2920xm es John thrashes... tasting the sweet xm goodness was not enough... Jumps on 2960xm OEM... Result world record GOLD. Shaden receives 2920xm es.... Shaden dominates it.... Then the itch the craving for power... Shaden upgrades to OEM result to be determined. The fate of the magic ES? Will the 2920xm es claim another Extremist? Only time will tell if the 2920XM ES will claim another. ;-)

Edit: Wow wonderfully done Brother Shaden. Yes flex can be helpful or counterproductive. I can suggest this in every test I ever did the only reason I used a higher flex is if I got a specific bsod like 00000124 or 0000101..etc. So I say drop that flex a touch... It technically shouldn't drop do to lack of flex... the voltage keeps it stable so if your system didn't A freeze or B bsod with like 00000124 or 0000101 don't worry you can use lower. Give it another run if you have time that's quite a nice one your running very optimal overall.

 

It's the temps. And your just about where i got to running vantage.

ROTFL!

Yes, so far that has to be one of the best performing ones. None of the others have hit 30k vantage yet.

 

Hehehe I cracked myself up. Yes I'd say heat is all too, drop temps by 2-3 more c and shouldn't be dropping at all. I found after 70-80 c it started to drop speed despite the thermal throttle temp set... but if that temp is set too low its even worse. Being on the ES with over 30k vantage is a gem for sure. ;-) that's one determined chip. Lol I almost made it sound like the chips alive lol... Or is it? :-D