Jump to content

X_X

Member
 View Profile  See their activity

  • Posts

    426

  • Joined

  • Last visited

 Content Type 

Everything posted by X_X

  1. X_X
    If your at 1.1V for 4GHz that's as good as I can get. Remember I was hoping for 5GHz lol so wonder if I also lost the lottery. This will be my last post on this forum, was supposed to be yesterday but a couple of loose ends to tie up. Been nice chatting to you and hope it all works out well.
  2. X_X
    In that one you can check that P_State_Limits isn't limiting your ratio but a lot of the settings you want are int the SDM mostly under the MSR section such as MSR 0x198, 0x199, 0x1A0. Limits may or may not exist in both MSR and MMIO so you would need to check especially as some might not be documented.
  3. X_X
    Notebookcheck is good for someone looking for a laptop and what to expect out of the box while HWBOT is more about maximum achievable performance. Intel gifted their Haswell mobile quads and their top Haswell mobile dual core with the ability not to only run a limited bin overclock but also the ability to fully unlock ratio's so they can run as high as power, cooling and stability permits and a Haswell at it's real limit draws a lot of power. Example
  4. X_X
    This. The specification is the brand string read directly from the CPU while the other is a LUT in CPU-Z software. @ScubaSteve404 Try HWiNFO to see if any different.
  5. X_X
    X_X

    25-Mar-2019 - No longer active on the forum

  6. X_X
    Sorry @skycat2216 I have no experience with v2 and can only comment on my findings with v3. Haswell is perhaps a special case.
  7. X_X
    Runs at a fixed 1600Mz which is tied to BCLK so while you cannot adjust ratio's AFAIK you can use the 125MHz strap for 2000MHz.
  8. X_X
    @Christiaan21-03 Problem is percentages can be so ambiguous. My old 1080 could hit over 200% for instance. GPU-Z has a bug when it calculates my 1050Ti power in Watts from percentages but then Nvidia have admitted the 1050Ti readings are crap so the newer drivers no longer report power for them.
  9. X_X
    Like I said, that's userbench for you. I don't have a ST CBR15 score to hand but have an MT one of 896 cb. Considering MP ratio will usually be under 5 for same clocks then at least 180cb for ST, probably more. We can take SuperPi from HWBOT, a well known single thread bench and compare how many seconds each takes to complete. Size 8265U 4700MQ 1M 9.636 7.846 4700mq 23% Faster 32M 572.638 433.774 4700mq 32% Faster For multi threaded we can use the well known wPrime bench with sizes of 32M and 1024M and again compare how many seconds each takes to complete. Size 8265U 4700MQ 32M 6.064 5.131 4700mq 18% Faster 1024M 271.675 163.316 4700mq 66% Faster 66% with 1024M! Does the performance of the 8265U drop off that much on benches that last a fair bit more than 30 seconds? Maybe someone can double check. Now if you want to run efficiently with crippled performance then the 8265U would win of course. For some though it's about maximum performance and you cannot get that being efficient unfortunately.
  10. X_X
    If it's done like that then it's not done properly ;)
  11. X_X
    I agree. Should be on Corsair to fix under warranty.
  12. X_X
    Seems to me you want to flash the ME region, by default it is locked. To unlock see if there is a BIOS option to upgrade ME, or a GPIO setting to unlock it or as a last resort shorting 2 pins on the audio chip, see here. You might find Intel FPT a little easier to use as you can easily specify just the ME region and not have to flash the BIOS region. This means you don't wipe your serial / MAC numbers but keep a backup just in case. I've not heard of RAM size affected by ME before but loss of overclocking can happen if corrupt.
  13. X_X
    Be warned that it ca be addictive :) Nothing wrong with software overclocking if done properly.
  14. X_X
    It means it's typically better to do the modifications before POST. For instance with the E5 v3 all core turbo the Asrock X99 Taichi hides some items such as the FIVR menu so we can unhide those. We can also add our own code, I have added a new menu "V3 - All Turbo" in my own BIOS This sub-section offers a choice of microcodes since people were claiming one microcode worked better than another. Microcodes are made available for CPUID's 306f2, 306f1 and 306f0. Some people were experiencing failure to POST with 306f1 when the microcode was removed so this modification would make the necessary changes for all core turbo AND apply the microcode afterward but before the troublesome module that the BIOS would usually halt on. Unfortunately after showing people how to run all cores on the Anandtech forum, in particular "The Stilt" prices of second hand v3 Xeon's became too much for my budget so never got to do any real testing except for my original 14 core E5-2683 v3. Might have shot myself in the foot with that one. Microcodes
  15. X_X
    Meh, I said other than power draw. Do be aware though that 15W rating is for base clock which is 1.6GHz while the 4710 is for a base clock of 2.5GHz. I'm not a fan of userbench, only seems to publish average or mediocre results. For example here's a couple of results I found that were done a while back https://www.userbenchmark.com/UserRun/224475 http://www.userbenchmark.com/UserRun/613476 Perhaps some results from HWBOT would be better, especially those benches that last a lot longer than 30 Seconds. I think you'll find the 8265U gets beaten by a mobile CPU from over 5 years ago.
  16. X_X
    I had a H100i stop cooling recently after about 3 years use and google seems to show the same thing time and again, blockages. There were also some stories of the warranty replacements not lasting long either so I decided to void my warranty (5 years) and open it up. Liquid port obstructed Channel for copper cooling fins blocked. It looks to me like something has crystallized in the liquid, perhaps some silicone grease that wasn't right for the job. I would think Corsair would be very aware of the problem after receiving returns and have hopefully addressed the issue in their newer units. At some time I'll try to put it back together minus the blockages and see how it goes. Pump was still turning but maybe starved of flow or dead heading into the blockage. After all it was fine when it was working properly so maybe worth a shot. Possibly in the Op's case there was some build up in the cooling circuit and moving the unit around dislodged it into a blockage. Just a guess.
  17. X_X
    It really depends on you. How much do you have to gain, are you trying to beat some record, etc...
  18. X_X
    The E5 v3 all turbo works the same way as desktop/mobile Haswell. IOW finding a early microcode patch that doesn't fix the all core errata or not using any patch until the full core turbo is properly set. Never heard of the CPU not being recognized from such deeds, perhaps you have done something extra special to achieve that. @skycat2216 The original v3 EFI file was for people who were kind of lost without BIOS buttons to push, really more of a proof of concept. People were supposed to modify their BIOS instead but seems nobody really took that path
  19. X_X
    It's a business move. Shareholders want maximum profits, not much different for instance from Microsoft imposing limits such as memory depending if you bought starter, home, premium... Doesn't anyone remember Intel's upgrade service? For some dollars it was possible to gain extra bins (clock higher), increase cache and in one case enable Hyperthreading.
  20. X_X
    @Hallinger_G See https://linustechtips.com/main/topic/1044306-question-about-alienware-laptop-cpu/?tab=comments#comment-12410214 Better in what way other than power draw?
  21. X_X
    What does HWiNFO show for the ISL6388 under sensors? It is an old board so was surprised you still had warranty, shame really as it's seems clearly a board failure. Please also confirm if disabling BD PROCHOT in ThrottleStop fixes the limit reason. Not sure if you need to disable it while it's not active or not.
  22. X_X
    If the heatsink on the power stage isn't making good thermal contact it might not be obvious that things are getting hot. If you have warranty that might be a good way to go unless they give you back something crappy HWiNFO supports the ISL6388 on your board, perhaps it will show the actual TM temperature and at what value VR_HOT# is activated. The temperature may also be used as part of the compensation circuit so better if it's working properly. Your original BIOS also has an option under "Digital Power" to set "CPU VRM Over Temperature Protection" from 85C - 120C, default is Auto. Actually it has an option to disable it too but it's hidden from the user. Check see if those HWiNFO temperature registers are shown, might also have the power in/out values too, and you can use these possibly as ammo for your warranty claim.
  23. X_X
    Are you sure it's not power throttling? 60C with CBR is good, wouldn't think AIDA would be much higher unless it's because it's running much longer. How are you finding the 2 extra cores, are they helping a lot more with real world use?
  24. X_X
    Hwinfo CPU Tab example For undervolting there is a guide, software and help here
  25. X_X
    If you use that link provided there's lots of help on there. For the extra 2 bins from 32x (3.2GHz) to 34x (3.4GHz) I would start with 32x ratio and undervolt it as far as you can while retaining stability, so lots of testing to ensure that. Then see what VID is required at normal idle temps for the 32x ratio and take that voltage, subtract the offset and add say 50mV which should be enough for 2 bins but test again for stability. An example... VID at 32x ratio is undervolted by -80mV and now shows as 1.000V. Subtract -80mV, add 50mV and we now get 1.130V and this is what you would set the adaptive (static) voltage to so that last the unlocked bin (34x) runs at 1.050V. The 33x ratio should run at 1.025V, all subject to temperature and AVX2 loading of course.
×