The Notebook Review forums were hosted by TechTarget, who shut down them down on January 31, 2022. This static read-only archive was pulled by NBR forum users between January 20 and January 31, 2022, in an effort to make sure that the valuable technical information that had been posted on the forums is preserved. For current discussions, many NBR forum users moved over to NotebookTalk.net after the shutdown.
Problems? See this thread at archive.org.
← Previous pageNext page →

    How to Overclock the Alienware 18 and Haswell CPU (or actually have it run full stock Turbo Speed)

    Discussion in 'Alienware 18 and M18x' started by Mr. Fox, Oct 15, 2013.

  1. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Maybe the person was having problem with it overheating. He's running 41x on 1 core load, and 40x for 2, 3, and 4 cores load.

    It's definitely not an Alienware, well at least not AW 18. Seems like the core current limit is locked on that machine, but then Clevo/Sager (might be a MSI too) doesn't screw it up, unlike Dell.

    I am more surprised at how it can reach 970 with that setting. I can't run 4.1GHz reliably without overheating in Cinebench.
     
  2. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    It might not be a reliable measurement. It's impossible to know one way or the other, but it is questionable at such a low clock speed with a less powerful CPU. If replicating the settings doesn't work for other 4900MQ owners to achieve similar scores, something probably went wrong with the benchmark calculation.
     
  3. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    I think the 4900MQ is exactly the same as the 4930MX, so at about the same clock speed the performance should be equal. It's just that the 4900MQ has locked multiplier, vs the unlocked multiplier of the 4930MX.

    The 2 4900MQ that scores 966 and 971 in XTU, are both Clevo P370SM. I found it by clicking the XTU hardware. It shows the specs.

    And Mr. Onyoto does indeed have 4930MX and 4900MQ. His XTU scores are right up there. He was like the first person to mention on T|I that swapping the CPU would reset the BIOS. I think that method forcefully induces a BIOS error, causing it to reset. So it's a sure fire each time. My method doesn't work the first time everytime, so one has to repeat it until it resets.

    I will try out the method of pushing BCLK. Either way, I am rather close to the limit of my CPU now. I need a bigger/stronger fan. The 11 CFM fan should come this week, so hopefully I can push it to 4.3GHz. Those Clevo owners need 1.15V to keep it at 4.1GHz, I need about 1.12V, but one of the core keeps hitting 100C and forcing a thermal shutdown.

    Maybe it's time for another reseat and repaste...
     
  4. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    Besides locked multipliers, it probably has locked TDP as well. Either one makes it not the same as the Extreme CPU. Both of those "features" make it an unfit product as far as being a performance enthusiast is concerned... more or less worthless. This would be just fine for click-n-go gaming and ordinary computing.
     
  5. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    The thing that makes me sad is that looking at those people with 4770K hitting 1.5k in XTU benchmark. Now I tell myself, shoulda gone for desktop.

    Well now I can't since I just upgraded the warranty to 3 yrs. Even if I do build one it won't be using Haswell CPU, instead it'll use the Ivybridge-E LGA 2011 4930K (nice name) or the 4820K.

    The 4900MQ comes from neutered 4930MX. So it's no surprise that it's performing so closely. Sadly on the 4930MX it's the thermals that's a limit. The integrated voltage regulator probably contributes to another 10W heat dissipation.
     
  6. Rafix

    Rafix Cave Canem

    Reputations:
    871
    Messages:
    839
    Likes Received:
    217
    Trophy Points:
    56
    There you go. I just ran my benchmark and destroyed you all with unmatchable results. Argument closed.

    [​IMG]
     
    Mr. Fox and kh90123 like this.
  7. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    The legendary 80486 CPU, probably the first x86 CPU that became popular. Mr. Rafix must be quite near to 30 yrs old, or perhaps more. :D

    Mr Rafix are you 30 yrs old as of 2013? Or did you inherit the computer from someone?
     
  8. Rafix

    Rafix Cave Canem

    Reputations:
    871
    Messages:
    839
    Likes Received:
    217
    Trophy Points:
    56
    lol Mr. Rafix is 40, and not only was he around when the 486 came out with its raw computational power, but he also had the pleasure to play with his father's IBM 286 having a 16Mhz CPU. I wish that back then I had the age and the knowledge to overclock it to 25Mhz :D

    Oh well, I have been patient, and now I can finally overclock my CPU as much as I want... oh wait, now that I think of it I cannot, Haswell doesn't let me to.
     
    kh90123 likes this.
  9. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Oh wait I did my math wrong. 2013 - 1973 = 40. I need to go repent now.

    Back then it seemed like CPU couldn't really be overclocked at all, until the recent past few years. I guess we could call it the golden age. Now due to rising thermal density the headroom is shrinking again.

    You can still overclock the 4930MX. Well, I can get 939 with only 4 GHz, I am sure one can tweak something out of it to get higher score at 4.1 GHz.
    XTU overclocking records @ HWBOT

    Here's my take on Cinebench 11.5. I couldn't really go any higher, hitting the hard thermal ceiling. See if you can go pass me, go try it, it'll be fun, they say :D

    Cinebench_record_3.jpg
     
  10. Rafix

    Rafix Cave Canem

    Reputations:
    871
    Messages:
    839
    Likes Received:
    217
    Trophy Points:
    56
    eh eh eh :)

    I was joking. I can actually overclock my Haswell moderately; I have reached a stable 4.2Ghz, which I use for my gaming. For some reason, lately I'm not seeing crazy temps in games that don't heavily depend on CPU as before. For example, in Batman CPU temps stays in the range of 75C~. As I mentioned in others posts, I don't really chase numbers. I don't have the time and patience to fine tune the system in an attempt to gain reach a better score--not my thing. But, you will never pass my 486 record, Sir! :D
     
  11. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    Im not, surprised in the slightest. Its throttling that is causing us to not get as high. when I cooled my house and ran the winning mark for 4930mx my temps never went past 81 degrees but it still throttles, and it was not the voltage either, there was plenty of room. It was only 4.3GHz after all.

    The highest scorers are oing to be the ones who throttle less, and are probably around 4ghz - 4.2ghz I got 968 at 4ghz 4.1 same settings \I get like 965
     
  12. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Well I don't even have a 486 CPU. My first two desktop have AMD CPUs. The thing with this Haswell CPU is, higher clock speed doesn't necessarily mean better performance. You need to give it juice where it needs it, but only moderately. And it also needs to be undervolted slightly. I only need 1.12V at 4.1GHz, so 4.2GHz prob won't need 1.2V. The core voltage adjustment have subtle but noticeable effects.

    I noticed that sometimes when it throttles, the clock speed didn't drop. What exactly does it throttle, the core voltage? Or the processor/core current?

    Yeah well I said it some time ago, the key to getting the highest scores in this benchmark is to prevent throttling. Any throttling and the score will take a dunk. That's why I have been keeping it at 4GHz, since my ambient temp right now is probably 10C higher than you guys'. It's so hot it makes you sleepy during the day.
     
    Rafix likes this.
  13. Rafix

    Rafix Cave Canem

    Reputations:
    871
    Messages:
    839
    Likes Received:
    217
    Trophy Points:
    56
    eh eh eh :)

    I was joking. I can actually overclock my Haswell moderately; I have reached a stable 4.2Ghz, which I use for my gaming. For some reason, lately I'm not seeing crazy temps in games that don't heavily depend on CPU as before. For example, in Batman CPU temps stays in the range of 75C~. As I mentioned in others posts, I don't really chase numbers. I don't have the time and patience to fine tune the system in an attempt to gain reach a better score--not my thing. But, you will never pass my 486 record, Sir! :D
     
  14. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    so, little to no effort at all, I can hit 950..

    but I know for a fact, XTU overwritten a few things in my bios, voltage wise, also, none of my toher profiles work anymore, they are not corrupted, they just don't fit my system for some odd reason, again guessing that XTU does a few hidden things at times, its not just switching or moving things visually, it actually writes something.

    Anyways I decided you can all try this 4.1Ghz gaming safe profile, I hope it works, if it does not work for everyone, then maybe some settings and not the same.

    dropbox:XTU version: https://www.dropbox.com/s/bo4jjxzlmaj0l09/XTU-Setup.exe
    profile: https://www.dropbox.com/s/51o2h7fkclufnl1/hopefully safe.xtu

    If this works for you guys, I imagine, someone here will easily beat my score in the next few days.

    I hope it does not do any irreversible damage, but use at your own discretion, and if your toying around with OCing, get used to removing your CPU a lot, if your planning on using XTU. (To unbrick it, when it writes faulty settings to hidden registers and voltages along the caches and cores.)

    tip: if some settings seem the same, possibility wise, it seems XTU likes to switch them a lot, so I would not use a value that can not be shared by another part of the cpu... like I/0 offset and cache offset - don't make these different. CPU voltage - its in mV's 10mV= +0.010v to the cpu, just add that manually, i.e if you use 1.21 you will want to make it 1.22v got it ?

    ask away.

    My next thing Im going to try is videobenching, but after a recent re-pasting, and adding 10 copper shims, I need to wait a few days for it to settle 100%. I did as a test, add the copper shims to the first/main gpu, which is usually 10 degrees warmer then the other when idle and such. Well now its 11 degree's less when Idle, so I'm abit worried, if that's normal, if its fine, expect some neat scores from that too, in the future.

    [​IMG]
     
  15. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    Actually, temps don't matter, from my experience on these cpu's. seriously, it will shut down if it needs too. But, it will first downclock so fast theres no way to notice, I've noticed the best thing is a moderately controlled OC but with enough power to sustain it, so that the jumps up and down are not very numerous, but rather, allows the CPU to actually hit 80 dgerees asap and then rise over 10 seconds to 9192 and turn the fans on. Its really voltage that causes it to shutoff, 99% the time though, not temps. Im positive about this, from my experience, its the vDroop, it might be easier to work with if the integrated vr's are disabled.
     
    Rafix likes this.
  16. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    This is probably one of the reasons why the dual AC adapter mod provides no benefit. The sucker shuts itself off before the system power draw has an opportunity to exhaust the resources of just a single AC adapter. The behavior is the same with single or dual AC adapters and it doesn't usually draw enough power to trip the breaker on just one AC adapter. There is also what appears to be a system power regulator/breaker because you can force the system to shut itself off without overclocking the CPU at all.

    If you push the voltage on the GPUs more than a little the system shuts off in the same manner it does with CPU overclocking. Attempting the overclock the CPU and GPUs at the same time is difficult because the system simply turns itself off if you try to achieve anything truly amazing with Futuremark benchmarks. You can set the CPU down to 3.1GHz, adjust GPU voltage to 1.1V and it makes it about 2 seconds into 3DMark11 test #1 before the system turns off. Somewhere around 1.05V is as high as you can go before it turns itself off. This is one of the reasons why the 18 cannot effectively compete with the performance of an M18xR1/R2. This behavior does not occur with the M18xR1/R2.

    If you push the CPU to the ragged edge, where it is on the verge of shutting down, you will see clock speeds drop insanely low, like 8-10MHz, and this occurs in the absence of overheating. If you push the CPU and GPU overclocking together the CPU will reach low clock states like 400-800MHz that totally ruin benchmarks results. I see this trying to bench using air conditioning... temps are fine, benchmark is rocking along like a banshee, then BAM, the FPS drops like a rock and the CPU is running 800MHz.

    And, as you say, it has nothing at all to do with temperatures. This happens in the absence of discernible overheating. Although, overheating is still a serious problem with moderate overclocking because of the fans not functioning correctly.
     
  17. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    I believe it throttles, because under no load, it can sustain full clocks and voltage (without using much current) but the vdroop hits when it gets a load, and then needs to throttle because to sustain the load at the speed it was at, it need MUCH more power. The higher voltage for me, eliminated a tiny bit of the spikes, and much higher wattage/power allowed it to sustain the load at the clock speed.. but theres a limit, that I suspect is somewhere near 135-136watts. If we go by efficiency, most things done right are 90-95% efficient, so probably 145-150 watts would give a proper fail proof of 136watts, meaning, that while the current limiter would cut off past 136watts, the wattage we're using is clean and devoid of spikes/artifacts/noise, whatever you want to call it.

    anyone understand what I mean to say, have observed?

    Its like on adesktop system, if you want to pull 1200 watts total, you don't use a 1200 watt powersupply, you use a 1500, because pulling 1200 form a 1200watts is probably drawing, 1350 from the 1200, but only 1280-1290 form the 1500 because of efficiency. Effeciencies are rated how they react at 50% load on desktop psu's and then its a hyperbole on a graph from there (exponentially degrading performance as you reach the apex/peak/highest power/max power etc).

    All this is supported by the fact, that lower clocks are matching if not even beating the higher clocks, and would also prove a current limiting factor somewhere, but only someone with an understanding of how to read those mumbo jumbo dataspec sheets could tell for sure. I've not used much of my electrical engineering experience in like a decade so, I don't even bother, being too rusty and behind times to find out for certain.

    Yeah , I don't believe its heat, go ahead and find whatever pushes the heat to higher 90's and keeps it there, it will function properly and throttle because of heat if it needs too, but we're seeing it throttle because of something else. The higher currents, and voltages should be able to sustain the higher clocks without as many drops, but, not at too high a clock, I don't think gaming past 4-4.1 is advised, if your pushing a lot of power along the rails for video too... whatever clock can be sustained near perfectly is going to result in best performances, because you could be at 2x the clock but if it throttle half the time to 30% the clock, then your at a loss.. and creating a large vdrop that's gonna crash the system and maybe even damage components from crashing currents.
     
  18. Rafix

    Rafix Cave Canem

    Reputations:
    871
    Messages:
    839
    Likes Received:
    217
    Trophy Points:
    56
    I understand, and thanks for sharing all this info and the result of your long hours of experiments with this machine. As I mentioned in your other thread, I am currently able to use a stable 4.2Ghz without any evident throttling or performance drop in game. I monitor core frequencies and I don't see any dives, not even when playing Crysis 3 maxed out and temps reach mid 80s C.

    See, for me is different. If I go lower than 1.2V when I set 4.2Ghz, I get blue screen or auto-reboots. Each system and chips are different; that's why an XTU profile may work perfectly in one machine and make another completely unstable.
     
  19. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    he probably only needs 1.12v because he is never actually under load. we can maintain our full clock at 0.97v and 0% load too.... but will need higher for any load. 1.12v, Im guessing is like nothing past the power of ...running 3.9Ghz.. or less who knows. I do not believe it could maintain 99% load , not at 4Ghz.

    Its like me running at 8Ghz... sure, its fine, but 0% load... and only 1.22v
     
  20. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    No, you're wrong there about my voltage woodzstack. I run Cinbench and XTU benchmark, back to back, and Cinebench pushes the CPU far more than XTU would. XTU benchmark is a cyclic load, and it gives time for the CPU to cool down, and heat up again. Cinebench doesn't, it's a heavy load from the start till the end. Although, from my observation the power spikes are higher with XTU during the cycles.

    Perhaps you can try to break my Cinebench record here, should be easy peasy for you: kh90123`s Cinebench R11.5 score: 8.74 points with a Core i7 4930MX

    Or you can try beating my 3Dmark11 here :D:
    http://www.3dmark.com/3dm11/7415002

    As I have said, I run 1.1V for 4GHz, and no more than 1.12V for 4.1GHz at 100% load. The cooler you can keep the CPU, the more efficient it's, and the more potential for undervolting is there.

    And many things have been said. I will add my 2 cents here.

    One can unbrick the system without removing the CPU, by just removing the RAM, I have documented the steps in another thread. Swapping CPU would probably reset the BIOS on the first try, but removing RAM might take a few tries.

    One should leave IO Offset and Cache Offset alone. I never really see the point of messing with it, the CPU is running hot enough extra voltages only mean extra heat. And just a warning, setting a negative cache offset will most likely brick your system. It bricked mine a few times, until I found out what I did.

    I can give the GPU 1.075V, but since I run my CPU at a lower clock presumably there's a little bit more power left to the GPU. Beyond that system is unstable.
     
  21. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,441
    Messages:
    58,202
    Likes Received:
    17,918
    Trophy Points:
    931
    The voltage required with intel chips changes very little with the cooling setup of a laptop, it does apply with liquid nitrogen and to some extent water but mostly it will come down to chip quality.
     
  22. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    yeah exactly, Ive wanted to make sure people don't try using these. Its risky, and unless there is counterweigh added to a negative voltage on these, it will crash/brick.

    hey Meaker, I didn't know you flew a stabber...(hopefully fleet issue), The higher voltage, saves me from potential voltage drops, to a slight degree. Now if you don't have spikes dropping in your benches, like XTU does - hey, then you have MUCH more room to go. My cpu seems to run stable at 4.3Ghz at 0.96v until a spike... so maybe I could run my benches using cinebench or something, if its free software.

    1.2 v makes me never hit 100 degrees 1.32 keeps me throttling and non-stop 99-100 degrees. 1.22v though keeps me so very close to 99-100 that it doesn not throttle(as severely) but comes close.
     
  23. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    So I went and check the "hopefully safe" XTU profile you gave.

    Do you really need 1.2V at 4.1GHz? Seems a bit too high.

    Also, why do you lower the long duration power to 75W after 128s? 75W it's on the low end for a 4.1GHz. I have seen it spikes up to about 83W in XTU benchmark. Cinebench would be a constant close to 75W, slowly creeping up as the CPU gets hotter.

    I am not sure for 4.1GHz or higher, but for 4GHz 115A for Processor Current Limit (ICCMAX) is about as much as you will need for any load.
     
  24. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    hey, if the setting works, the way it does for me, for anyone else. they would end up 2nd or 3rd place on the rankings. But, how did the profile work, is the question. Whatever you8 are not sure about, why not go and try and then come back and speak your observation.

    Turbo boost max and whatever, don't really work (OR perhaps not the way we suspect they do). I think I left them once at 75 watts, when I hit 960+ running 4.3Ghz... but, the core current limit and such, does work.

    But by all means, feel free to move that slider that's at 75w all the way to 4096 or whatever is max. Shouldn't make a difference in the world, at least did not on my pc, as the bios also has those values set to 32,000 or something..
     
  25. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    I have not seen anyone else who ran your profile yet. I am not saying anything is wrong with it, but it's not optimized for everyone. Run the same benchmark at a tropical country and your system will overheat and crash. Sometimes it's good to practice some discretion before you lash out on people.

    Turbo boost works exactly the way it is, if you adjust it in BIOS. There's a MSR lock which is on by default. XTU doesn't override the boost values in BIOS even if you disable them. That's not intended behavior. It's a problem with XTU. If one sets the short power limit to 75W in BIOS, you will notice diminished performance in XTU benchmark and Cinebench. The CPU easily consumes more than 75W, you don't notice it because you don't run benchmark that last longer than 128s.

    How turbo boost works is simple. It will allow a maximum power that's specified by the Turbo Boost Short Power Limit (Short Duration PWR Limit Option in BIOS) in Watts, for a duration that's specified by Turbo Boost Power Time Window (Long Duration Time Window in BIOS), in seconds. After that duration of time expires, it will lower the maximum power to the CPU to Turbo Boost Power Max (Long Duration PWR Limit in BIOS).
     
  26. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    I used his profile and did not experience any problems applying it. It didn't change my CPU performance much from the profile I had already posted... very similar, in fact.

    Those long and short duration turbo sliders are like caps or limiters. It does not force a certain amount of power to the CPU, but raises the limit on what is available. Setting them too low can hinder performance. The same appears to be true with Core Current Limit and Processor Current Limit. As best I can tell, the only settings that actually push the values are the voltage adjustments. The amps/watts should be largely dynamic, based on demand. So, yes... you can slide them all the way to the right. The time window limit being lower can help minimize throttling, but may increase temperatures. Hover over the value in XTU and you will see a description. It is the number of seconds the CPU cores must stay BELOW the turbo power maximum. An unlocked desktop i7 CPU default time window is something like 1 or 2 seconds.
     
  27. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    The default for our AW 18 should be 28s, or at least that's what it said in the BIOS.

    His settings doesn't work for me. Given that my ambient temp is at least 10C higher than what you guys get, I guess it's inevitable. I will have to find my own settings.
     
  28. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    I think you are correct. And, a 10°C difference in ambient temps is huge. You are starting from a cold boot with your system normalized to a higher room temperature as a distinct disadvantage.
     
  29. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    whatever we do in XTU does something to values we can not see, Anywhere. I can run xtu, after flashing my bios, to stay clean, and then go back and reset things to what I use in bios again, and only use that, but things are different in performance.
    Its so bloody friggin annoying.
    My videocards seem to not want to go past 20-25% powerusage sometimes, occasionally 50%... I think they seem to follow the cpu's throttling. if I am throttling when I start a game, the videocards might stay throttles to same ratio.

    This is annoying.

    I really have no clue whats up with them. Im guessing.

    edit:

    look at this, when I maxed out power in bios, you know 32,xxx whatever.. seems it added 200 watts to that and made thos evalues for my graphics cards.

    [​IMG]

    I think, these values affect a lot more then we were lead onto. I think the rails/planes are shared.

    maybe someone else can give me a tip, if they know.

    see where it says max power % ?!?@
     
  30. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Maybe you need to flash the vBIOS and overclock them too. :D

    Anyway, I found out new stuff about this CPU. For my CPU running at 4GHz, with adaptive voltage mode enabled, the lowest I can underclock is 1.167V. It doesn't want to go lower than that. Since I know it's stable with only 1.1V, I have to either apply a negative dynamic core voltage offset, or use static voltage. I prefer the latter.

    And just for reference, for my CPU,
    4GHz @ 1.101V,
    4.1GHz @ 1.115V,
    4.2GHz @ 1.155V (not sure if more voltage is needed, my CPU is overheating at this voltage)

    As you can see, every increment of 100MHz, requires exponentially more voltage.
     
  31. Meaker@Sager

    Meaker@Sager Company Representative

    Reputations:
    9,441
    Messages:
    58,202
    Likes Received:
    17,918
    Trophy Points:
    931
    Yep which is why the power requirements start to really take off past 4ghz with haswell.
     
  32. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Actually when the CPU gets a little bit warm, it will need 5-10mV more than what I stated. And that has noticeable effects on load temps.
     
  33. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    *plays "Mission Impossible" theme song* Your mission, should you choose to accept, is to achieve extreme performance from a Haswell processor... this CPU will self-destruct in 5 seconds... good luck, Mr. Phelps.
     
  34. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Just to update my findings. Discovered some new info.

    Mr. Fox asked me a question, since my Processor Current Limit wouldn't stick in XTU:
    http://forum.notebookreview.com/alienware-m18x/736651-email-susan-thomas-4.html#post9440414

    And I went back to test with my stock Win 7 which is still in the stock HDD, and my findings are here:
    http://forum.notebookreview.com/alienware-m18x/736651-email-susan-thomas-5.html#post9441013

    In short the value (Processor Current Limit) sticks, but doesn't in Win 8.1.

    Then it made me suspicious that it could be related to fast startup/fast boot, and so here's my findings:
    http://forum.notebookreview.com/alienware-m18x/736651-email-susan-thomas-5.html#post9441152

    In short, don't disable hibernation (powercfg /h off, powercfg /h on), as it will disable fast boot/fast startup, which will lead to XTU's Processor Current Limit not sticking.

    EDIT: Ok fast startup isn't very good when you're benching, as if it crashes then you want the system to cold boot properly. My suggestion for Win 8/8.1 users would be to disable it when benching/overclocking, and only enable it once you have found a stable setting.
     
    Mr. Fox likes this.
  35. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    AFAIK, all haswells have the same issues we are having, but, the extrme core ones, still out perform, clock for clock, volt for volt, right ? I mean, the 4930mx and 4900mq are the only 8m cache ones, and should technically be not only higher binned, but that means more efficient too, right ?
     
    Mr. Fox likes this.
  36. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    Yes, that is correct. At lower clocks speeds that are within its reach the 4900MQ is similar, but it still lacks potential and has limitations compared to the 4930MX. The 4700, 4800, 4702 and other Haswell processors I have tinkered with are unimpressive to put it kindly. The 4930MX can actually do some remarkable things by itself, but when operating in conjunction with powerful graphics cards quickly becomes unimpressive in a laptop. I think that maybe its severe power demands when overclocked creates a situation of "robbing Peter to pay Paul" and there's not enough juice left for everything else to work properly with a single 330W AC adapter. On top of that, challenging operating temperatures and power throttling (even without high temps) doesn't help matters. These same defects also show up in varying degrees with Haswell desktop processors. It can be called "efficiency" as much as they want, but all that really matters at the end of the day is the achievement of excellent results and consistency. It is a challenge to achieve either of those from an overclocked Haswell CPU.
     
  37. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    I don't think its heat that is an issue like it used to be, so much as because efficiency being so much greater, we're getting more gains for every volt - however - you have to look at a larger picture. imagine it was 100% efficient. what goes in comes out. voltage goes in and I/o's come out, dispersing that voltage. the relation towards that has become so tight, theres no more playing around. to keep a logic high ( i.e +1) still requires a flip on the transistor, i.e flip flops.. the things we measure as IOPS each latching of a set of flops being one IOP .. the gap for playing around on these has become very tight. when you start to play with nand gates and ICs, you use like 1.5 volts and anything around 1.2 is high and anything under usually 0.3 and less is low... and whats inbetween doesn't happen. well, processors use much less now. the measurable difference in that is non-existant now.


    now back in the day, increasing 100mghz on our Pentium 2's was huge. but only required a small boost in voltage. but we were looking at what, 1,000MIPs or something ?? to become like 1100MIPS... 100mgz extra today is giving us, more or less 10,000MIPS more. essentially, the voltage HAS NO CHOICE but to almost skyrocket. it would seem logically that it should only go up 10%, but that's not true, we're not adding 10% more IC to the cpu on die, we're pushing an exponentially degrading algorithm past normal limits and bounds.

    to walk up a flight if stairs, at a normally fast but doable speed, requires normal not- so stressful strain and little to no energy. But try running up those steps, and then go faster, and faster.. the amount of energy to do so, rises exponentially to a point, no matter what effort we put in, we're not gaining ground at all. limitation of our physical body, even if we spend 10x more energy, we're only going to maybe double our ability, with possible catastrophe.

    This is how I sum up cpu's today. silly analogies I know.

    Also, anyone who thinks the previous generation is better, needs to look at the facts. Hawsell's ARE faster and more efficient. The ability to play with them, and crunch the higher clocks, is not as rewarding, but the gians per clock are higher then on previous generations. I can hit 120,000MIPS with my 4930mx. I'd like to see any other mobile processor do that. This is around the ability of the desktop's previous generation when overclocked. my 3820qm from my R2 net me about 75,000 MIPS, when overclocked to its full 3.8Ghz on all 4 cores.. Im almost double that with only 4.3Ghz on this cpu.
    Ive looked around, and some people were able to get 110,000MIPS with the 3rd generation extreme's but around 4.9 Ghz. Hands down, the Hawsell wins.
     
  38. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    Numbers don't lie. people lie. ! Cake is not a lie, the game's developers were a lie!
     
    Mr. Fox likes this.
  39. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    Well, here is what my numbers show with a ThrottleStop CPU benchmark as well as 3DMark11 Physics Test. Both laptops in the same room, less that 3 feet apart, same ambient temps, same core clock speeds, both running Windows 7, both using the same AC adapter. Cooling with nothing but internal fans. Having lower core temperatures does make a significant difference in performance.




















    Alienware 18 - 4930MX @ 4.3GHz - 3DMark11 Physics Score: 9136 | TS 32M Test: 8.244 sec Alienware M18xR2 - 3920XM @ 4.3GHz - 3DMark11 Physics Score: 11008 | TS 32M Test: 7.023 sec
    [​IMG] [​IMG]
    And, just for good measure... wPrime 32M test 4930MX @ 4.3GHz - 6.792 sec ...and wPrime 32M test 3920XM @ 4.3GHz - 5.735 sec
    [​IMG] [​IMG]
    Second chance re-run of same tests using XTU profile in opening post for 4930XM, plus CPU fan forced to 100% for lower temps - wPrime 32M 6.049 sec | TS 32M: 7.501 | 3DMark11 Physics: 10299 - better results, but 3920XM is still more powerful than 4930MX at same clock speed Screen shot from opening post showing optimized XTU settings for 4.3GHz

    Caution: these 4930MX optimized settings may not work the same for your CPU
    [​IMG] [​IMG]
    click image to enlarge in a new browser window​
    Winner: Ivy Bridge 3920XM

    CPU performance also influences GPU/3D performance and gaming results - see these examples...
    M18xR2-P13951.jpg AW18-P12516.jpg P13078-4930MX.jpg
     
    Last edited by a moderator: May 6, 2015
    TBoneSan and kh90123 like this.
  40. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Oh about fast startup.

    If you shut down the computer the value of Processor Current Limit persists in XTU, since the devices don't get reinitialized. On the other hand, if you restart the computer, the value will go back to default value of 55A. With fast boot and fast startup, a restart is like a true cold boot.
     
  41. _deadbydawn_

    _deadbydawn_ Notebook Evangelist

    Reputations:
    49
    Messages:
    314
    Likes Received:
    278
    Trophy Points:
    76
    hi all

    i'm new to this forum, but have been reading since quite some time.

    i found this very interesting thread and i think maybe someone on here might have a clue about what i'm experiencing:

    my system is an alienware 18 with i7 4930mx, 2 x 780m 4gb, 16 gb ram.

    i followed the instructions given by @Mr. Fox and have the system running @4.3ghz. when i run 3d mark 11 with stock settings (cpu stays constantly on 2.4ghz on all cores), i get ~12000 points (is this too low?). temps on those runs for the cpu are around upper 70's for the cpu and close to 80 for the gpus. after applying the oc on the cpu and playing with it (multis from 39 up to 44, i got between 8000 to 10000 at the max. the temps were roughly the same as on stock.. if set the cpu to default settings and oc the gpus the graphicscore goes down, no matter what oc i use. the temps on the gpus (i.e. when set to +130 / +400) are still around 80 degrees.
    it just doesn't make sense to me, i'm used that at some point, the oc will revert and the performance goes down due to more heat, but i'm not used to this kind of behaviour. temps are toooo much different but the performance never even reaches the level of the stock settings... :confused:

    i'm really at a point where i don't have any more ideas. am i missing something (i.e. do i have to set something in throttlestop as well?)?

    maybe someone has an idea, as mine are running out :)

    thanks a lot

    cheers from basel, switzerland
    yves
     
    Mr. Fox likes this.
  42. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    Can you post a link to one of your 3DMark11 runs?
     
    _deadbydawn_ likes this.
  43. _deadbydawn_

    _deadbydawn_ Notebook Evangelist

    Reputations:
    49
    Messages:
    314
    Likes Received:
    278
    Trophy Points:
    76
    yeah i will do so this evening when i get home (when i wrote the post i realized that i forgot to run teamviewer on the aw18, because it's guarding the flat while i'm at work ;-) )



    EDIT: oh, what i forgot to mention: when the cpu is oc'd, the physics score goes down.
    can't wait to get home for the printscreens.
     
  44. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    So sad, all of us who got the 4930MX got ourselves a dud. :mad:
     
    _deadbydawn_ likes this.
  45. woodzstack

    woodzstack Alezka Computers , Official Clevo reseller.

    Reputations:
    1,201
    Messages:
    3,495
    Likes Received:
    2,593
    Trophy Points:
    231
    I was explaining the same issue before hand, that he is having with 3dmark. Its because your videocards won't go into performance mode, and your cpu throttles. However, stock Dell Drivers at stock speeds seem to work great right ?
     
  46. _deadbydawn_

    _deadbydawn_ Notebook Evangelist

    Reputations:
    49
    Messages:
    314
    Likes Received:
    278
    Trophy Points:
    76
    now after you mentioned it, i recall reading this before on here!
    i never even tried the stock drivers, i updated them right away. later on, when i get home, i will post the links to some bench runs and as well, uninstall the new drivers and reinstall the original ones which came with the system to try some more benching. when you were on stock drivers, did you gain some score oc'ing the cards? there's so much stuff concerning the aw18 which is just of this IT-world (but sadly not in a positive way), i don't even want to start about the fan tables...
     
  47. Mr. Fox

    Mr. Fox BGA Filth-Hating Elitist®

    Reputations:
    37,255
    Messages:
    39,358
    Likes Received:
    70,785
    Trophy Points:
    931
    Hey bud, I think he may have been making that comment just to be facetious and humorous.

    The stock Dell/Alienware video drivers are actually pretty decent, but they are not going to fix these performance issues.
     
  48. _deadbydawn_

    _deadbydawn_ Notebook Evangelist

    Reputations:
    49
    Messages:
    314
    Likes Received:
    278
    Trophy Points:
    76
    oh i see, but i think to remember someone saying that he's going to revert to the original drivers ^^

    well, first off i'll provide some links
     
  49. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    There seems to be page 21, but I can't go to page 21. It must be the forbidden page. Maybe the site's having problems again.

    Without overclocking the CPU and GPU, the scores should be about 12k in 3Dmark11, so you're not far off. There are currently some problems on the AW 18.

    First, 4930MX runs rather hot, and gets outperformed by Ivybridge 3920XM.

    Second, Dell set the current to the processor at a rather low value, so under load the CPU would throttle to 3GHz. Put it under Cinebench load or wPrime load, it will possibly go down to 2GHz. The reason Dell/Alienware chose to do this might be to limit the temperature of the CPU to below 80C (actually the GPU is kept to 77C max temp, I think that they want to keep the CPU at around that temp).

    Third, to fix the problem temporarily with the current, we use Intel XTU, but Intel XTU has some wonky interaction with UEFI/BIOS settings, especially if you're running Windows 8/8.1.

    Fourth, on top of 4930MX running hot, we have a fan table that's too slow to react to the spikes in CPU temperature. And this could cause overheating and thermal shutdown. I wouldn't mind if the GPU fans spin up more either. You know what, just let us control the fan profile. Someday I feel like having a silent machine, someday I want to play some latest games but I don't want the GPUs to catch fire. :D

    Fifth, 780M SLI + 4930MX overclocked, easily pulls more than 330W. Performance is also hampered by the limitation of the AC adapter, especially the GPU overclock at the high end.

    Sixth, the screen has burn in issues, like plasma TV. I read that some IPS screens has this problem, but seeing that we paid a lot I expected better. It goes away after a short period of time, but I could see that it might be annoying to some.

    Seventh, which might be irrelevant to some. The M18x R2, if I am not mistaken, has space to accommodate 4 2.5inches HDD/SSD (3 in the disk bay, 1 in the ODD bay if you remove the ODD). Now this AW 18 (only 2 in the disk bay, and another 1 in the ODD bay if you remove the ODD) is bigger wider and heavier, there's clearly enough space in the ODD bay to have another HDD/SSD in there. Likewise, there's easily space for extra mSATA drives. Why they did not include the connector for the 4th HDD/SSD, is beyond me.

    The SATA connector for the ODD can only be used for 1 HDD/SSD. I think if we replace that with this: FMJH7 HDD FPC cable, which has 2 SATA connector for 2 2.5" drives, it should work. Just need to stick or tape the drives somewhere in there.
     
  50. kh90123

    kh90123 Notebook Deity

    Reputations:
    964
    Messages:
    989
    Likes Received:
    423
    Trophy Points:
    76
    Speaking of which, Mr. Fox, how come I get slightly better scores than you in wPrime 1.55 even though I clock mine lower, at 4.1Ghz?

    This is yours, at 4.3, http://i.imgur.com/zZGvoSB.jpg

    This is mine, at 4.1,
    View attachment 104622

    Also, why is there another version that's wPrime 2.10? My scores are noticeable worse in 2.10 compared to 1.55. The wPrime website said 1.55 is the HwBot version.
     
← Previous pageNext page →