Jump to content
HWBOT Community Forums

Alex@ro

Members
 View Profile  See their activity

  • Posts

    1834

  • Joined

  • Last visited

  • Days Won

    27

 Content Type 

Everything posted by Alex@ro

  1. Alex@ro
    2220 ramu placi...f tare samsungu
  2. Alex@ro
    Your eff is way off man,maybe you need a reinstall of win ...
  3. Alex@ro
    I got to play with the OC Formula and it is a solid board for sure.Quick testing 4-dimm MFR kit from x99 could perform easy 3200 C12 tight or 3333 C13,also samsung 3733 15-18-18 or 3866 16-19-19 was a breeze with a r retail G.skill X99 3000 15-15-15 kit. I might add that the volts needed on x99 are the same like Z170 so i guess it is off to little bit of tuning to get the results to x99 and ofc higher.
  4. Alex@ro
    L524B663 shit shit shit shit shit cpu 1.4V VID ,how is this possible,fails 4700 cinebench 1.425
  5. Alex@ro
    Ok,you guys are on your league
  6. Alex@ro
    payment received, can be closed.
  7. Alex@ro
    thanks,need to see them go
  8. Alex@ro
    All4 for 100e
  9. Alex@ro
    Hello,i have for sale 2 kits i purchased recently but have no use for. Both 2200 7-10-10 Pi series,both in mint condition with box and fan,K0-8155. Perfect shape,they both do 2600 8-12-8 very tight at 1.87-1.88 I preffer to sell them both at 100E + shipping 15E in Europe,25$ to USA.
  10. Alex@ro
    Thanks,i did quick tests on 32M on some values just to see if things are really benchstable and not only for bragging!
  11. Alex@ro
    As i was playing with the Maximus VIII Gene and tweaking ram i came to notice how things work regarding training and having the right RTL/IO depending on frequency and CAS latency. It is important to understand that having properly set RTL/IO will also gain stability versus non optimized training and of course give you that small amount of performance needed to beat damn bullant's 32M run Having said that i tested the following method on basically all memory IC's on the market and found out that it works good . What have i done here ? Basically i found how to set the proper RTL/IO combination that fits given frequency/cas timings. First of all,Skylake is a much more nicer platform then X99 on DDR4 so if you are a memory fanatic this suits you perfectly.There are enough secondary and tertiary and also RTL variables to make you spend days tweaking and discovering new scenaries . Asus did a good job with Maximus VIII Gene as always and the motherboard behaviour in memory fiddling is great. Hardware used: Core i7 6700K Retail,nothing extraordinary Maximus VIII Gene 0004 Bios Kingston 3466C16 ES Kit(AFR) The reason of using AFR was that besides being my best memory it accomodates a larger range of CAS/Frequency,especially the ones that we are interested of,meaning tight CL of 11,12,13 Now what is this topic really about.In the RTL IOL CONTROL you will find beside the usual RTL/IOL settings also two new areas : IO Latency OFFSET IO Latency RFR Delay. Ofc with 2 tabs used for the 2 memory channels. How does this work: Say you plug your good old Hynix MFR and aim for DDR4-3000 C11. All set and go the usual RTL/IOL for this will be 44/45/6/6 In order to keep this values and avoid bad training you might want to set RTL Initial Value to lowest value that can boot and you will be good. If you are not happy you can improve this.How ? Move from standard IO Latency offset of 21/21 one value higher.Apply and reboot.You will notice tighter RTL/IO.Repeat this until you fail to POST.Go back one step and you have the tightets RTL/IOL combination that your system can handle at this scenario.In my case : 43/44/1/1 . Now all you have to do is type manually the values and you are good to bench. Test Scenarios : CAS 11 : DDR4-3000 C11 Auto Set RTL/IOL = 44/45/6/6 DDR4-3000 C11 IO latency= 24==>44/45/2/2 DDR4-3000 C11 IO Latency =25==>43/44/1/1 DDR4-3200 C11 Auto Set RTL/IOL = 45/56/3/13 (Bad training this is the case when stability fails because of missed training) DDR4-3200 C11 IO latency 24==> 45/46/2/2 DDR4-3333 C11 Auto Set RTL/IOL = 54/47/15/6 (again missed training) DDR4-3333 C11 Auto set RTL/IOL = 46/47/7/6 Standard settings DDR4-3333 C11 IO Lantecy 23 = 46/47/4/3 CAS 12 : DDR4-3333 C12 Auto Set RTL/IOL = 48/49/7/6 DDR4-3333 C12 IO Latency 23 = 48/49/4/3 DDR4-3466 C12 Auto set RTL/IOL = 48/51/6/8 DDR4-3466 C12 IO Latency = 48/49/3/3 DDR4-3600 C12 Auto set RTL/IOL = 53/50/11/6 DDR4-3600 C12 IO Latency 23 = 49/50/5/4 CAS 13 : DDR4-3333 C13 Auto Set RTL/IOL = 49/53/6/8 DDR4-3333 C13 IO Latency 23 = 50/51/4/3 DDR4-3466 C13 Auto Set RTL/IOL = 50/51/6/6 DDR4-3466 C13 IO Latency 23 = 50/51/3/3 DDR4-3600 C13 Auto Set RTL/IOL = 51/54/7/8 DDR4-3600 C13 IO Latency 23 = 51/52/5/4 DDR4-3733 C13 Auto Set RTL/IOL = 51/53/6/7 DDR4-3733 C13 IO Latency 23 = 51/53/4/6 I used AIDA64 as a quick performance measure to see the gains however you should feel small boost in the memory dependent benchmarks such as XTU,S-pi 32M,Maxmemm,Geekbench,etc. Please note that you can adjust IO latency separate on each channel altough they have to stay 1 value close to each other so for example if you found that 23/23 works you should try 24/23 or 23/24 as ultimate tuning. Also you can try tweaking IO Latency RFR Delay also,there are scenarios when if you are on edge you can try going higher with this as well and it will have the same effecto of training lower RTL/IOL. By doing this guide i wanted to help people having a starting base for basic tuning of memory aspects in DDR4 early times using Asus Motherboards based on Z170 . Have fun playing with your DDR4 !
  12. Alex@ro
    Do it anyway,you might get suprized
  13. Alex@ro
    First to 2015 has a beer
  14. Alex@ro
    Thanks,i hope this solved many people that bench chillers or single-stages dilemmas.
  15. Alex@ro
    Everyone was assuming that this paste would have good thermal capability i guess...You are basically right,put no paste and have no CB ,or put a sheet of armaflex between cpu and pot and still CB free
  16. Alex@ro
    As you already know Skylake overclock depends greatly on TIM used.It is now more important than ever for both ambient and extreme users to use a quality TIM for maximizing overclocks. Since overclocking Skylake is pretty much apply voltage and frequency and start i made a short test with stock tim versus two quality solutions from Gelid and Grizzly . I used for this a good single-stage made by nachtfalke who kept a steady temperature of -56 idle -53 load on xtu benchmark . Having a steady temperature and avoiding variations helps greatly test data so i have to give credit to nachtfalke for being a great builder and also a good friend Long story short: SINGLE STAGE -56 IDLE -53 LOAD Intel I7 6700K @ 1.58V Frequency for completing XTU benchmark : Test #1 Cpu not delidded,stock TIM between DIE and IHS,Gelid GC EXTREME between IHS and EVAPORATOR. Results : TERRIBLE Stock TIM is a disaster and Intel should be sued for letting chips like these out of factory,maximum temperature was 54 degrees,maximum frequency 5070... Test #2 Cpu not delidded,stock TIM between DIE and IHS,Grizzly Kryonaut between IHS and EVAP Results : POS but better Kryonaut improved the result by 12 degrees and maximum frequency is now 5120 ... Test #3 Cpu delidded,Gelid GC-Extreme between DIE and IHS,Grizzly Kryonaut between IHS and EVAP Results : Good Replacing stock TIM with Gelid GC Extreme slashed another 19 degrees and maximum frequency is now 5190 for XTU Test #3 Cpu delidded, Grizzly Kryonaut between DIE and IHS,Grizzly Kryonaut between IHS and EVAP Results : Excellent Grizzly Kryonaut between DIE and IHS slashed another 10 degrees and gained 100 mhz more for a decent frequency of 5290 XTU test... Conclusion : Kryonaut is an excellent product and should be taken into consideration by anyone who knows what he is doing when overclocking. A good tool for overclockers deserves credit and this goes to Roman !
  17. Alex@ro
    100$ for 50 grams of paste to have some common sense tough would be maximum i would spent . The thing is the more is made the cheaper it gets for supplier,so a proper way of doing this would be to create a preorder list of interested people then see the final price .
  18. Alex@ro
    you don't care if you could have a WR too
  19. Alex@ro
    Haha,true.The Japanese so a pretty good job and good tweaking,well done !
  20. Alex@ro
    You are a luckbox...
  21. Alex@ro
    it can be squeeeezed more for sure
  22. Alex@ro
    you've done it,nice job
  23. Alex@ro
    it does 5300 at same voltage but this was a first run
  24. Alex@ro
    Maybe L519C087 otherwise it means you are back in time with your skylake sample
  25. Alex@ro
    4970 xtu on air?great chip
×
×
  • Create New...