Jump to content
HWBOT Community Forums

Midnight703

Members
  • Content Count

    26
  • Joined

  • Last visited

Community Reputation

12 Good

About Midnight703

  • Rank
    maintenance bot
  • Birthday March 24

Converted

  • Location
    North Carolina, US

Recent Profile Visitors

92 profile views
  1. Midnight703

    i7 8086K

    I should have one available in a couple weeks. I am in US if you are still looking then and are willing to cover shipping.
  2. I did some experimenting with single rank v. dual rank memory in Time Spy and thought I would share my results here. Setup: 10900k at 5.3 core/4.7 ring XII Apex Single rank kit- Team Group 8-pack 4500cl18 xtreems Dual rank kit- G. Skill Trident Z 32gb 3200 cl14 Testing methodology: I used Time Spy cpu run only to compare results, and used the cpu score to measure gain in performance for each timing. I used maxmem to reduce the 16gb down to 5gb, and from 32gb to 10.4gb. For memory timings I used flat primaries with tRAS=tCL+rRCD(RD)+2, tWL= CL-2, tRDRD_dg and tWRWR_dg at 4, maximus tweak mode 2, and used 1t for all of the single rank timings, but for dual rank 1t only worked on the first test, and 2t was used at the rest of the timings. Everything else on the memory was kept at auto, I have attached Asrock Timing Configurator screenshots so you can see the other timings that were trained for each freq. Here are some screenshots from both single and dual rank kits showing the auto timings that were trained: https://imgur.com/a/hYNo2qC I used only as much voltage as necessary to complete a run and did not play with increasing voltage for additional stability or to see how it impacted a score. For this testing I only used a limited range of frequencies based on the limitations of the kits themselves, would be interesting to see if it keeps scaling linearly. With these limitations lets take a look at the numbers: 16gb Single Rank Speed/cl Result 1 Result 2 Result 3 Avg result Jedec 12164 12033 12363 12186.67 4000cl15 13408 13481 13198 13362.33 4100cl15 13555 13583 13519 13552.33 4133cl15 13596 13557 13516 13556.33 4200cl15 13853 13768 13810.50 4266cl15 13747 13789 13763 13766.33 4300cl15 13923 13926 13896 13915.00 32gb dual rank Speed/cl Result 1 Result 2 Result 3 Avg result Jedec 12873 12798 12855 12842.00 4000cl15 15852 15776 15796 15808.00 4100cl15 15880 15796 15845 15840.33 4133cl15 15958 15912 15859 15909.67 4200cl15 15856 15905 15954 15905.00 4266cl15 16001 15976 16004 15993.67 How much faster is Dual Rank? Speed/cl Jedec 655.33 4000cl15 2445.67 4100cl15 2288.00 4133cl15 2353.33 4200cl15 2094.50 4233cl15 2227.33 4300cl15 NA As you can see dual rank is slightly faster at jedec speeds (I didnt look at jedec timings to make sure they were the same, but how close those scores are seems to indicate they are the same), but it surprised me at just how much faster the dual rank kits are once they get up to 4000cl15. Some of this may be obvious to the more experience benchers, and looking at the world records they use a large amount of memory, but I thought it was interesting to have actual numbers to see just how much better the dual rank is. I would love to see someone with a quad or higher channel setup and see how their numbers scale with these.
  3. Out of curiosity, any idea how the 3900x does on ambient?
  4. Should be able to get 1t at that freq, try oil offset of 22 or higher and see if itll train, and see if the rtls will train all the same, at 62 or lower. I would also go through the secindaires, looks like some may be able to tighten and terts should be able to tighten significantly. What voltages are you on for those timings?
  5. Not in my experience, it doesnt require round trip latency to be enabled, but most of the time mode 2 will not train when you are pushing the limits of your imc or kit. If you can train mode 1 without a problem but not mode 2 I assume that's an issue with imc or kit not being able to handle the tighter subs, iol, and rtl values.
  6. Were the other 10900k's tested with cl15 only too?
  7. My understanding is the maximus tweak modes tighten subs, rtl, and iol values with mode 2 being tighter than mode 1. When testing limits I try to keep it on mode 2 and see how far it will go. I have also noticed on mode 2 the rtl values are as tight as they will train, but the iol values have some room for tightening most ost the time. If I am trying to push high freq I will sometimes back off to mode 1 if its having trouble training, only going to auto if it cant even train on mode 1.
  8. On core voltage my understanding is that the voltage limitations are the same as 9th gen, and I have consistently heard that 1.4v under full load was the limit for 9th gen. Though with the temps being lower on 10th gen I am not sure this holds true, and the speed of degradation, if any, could take longer given lower temps. That being said I will likely not go over 1.4v full load for daily.
  9. 0088 allows for v dimm up to 2.4, just checked it today. Not sure on core voltage side.
  10. I actually just flashed my bios to it to check it out and can confirm it does allow over 2v vdimm, up to 2.4v. I appreciate it. Gonna check out bios 0055 as well to see which is better.
  11. Good to know! I like the habe the flexibility of going over 2v on memory (not core) for mem OC. Not totally necessary but a nice option to have.
  12. Just got my z490 Apex, is 0088 the best bios so far? Anyone confirm if it allows over 2v vdimm without LN2 mode enabled?
  13. I was wondering at what memory speeds other bench? I know about 4666cl14-13 1t, 4800cl14-13, and 4133 cl12-11 1t but after after looking over some subs from memory bound benches like super pi and wprime, I am seeing a lot of top subs have lower memory speeds. So that made me wonder what others use as their bench memory OC. I have been working on a 4800cl14 memory OC profile but it's quite inconsistent in when it posts, it almost depends on the say as ti whether it will post or not. I feel like as a rookie I am missing something here. System I am working with is a 9900k on a XI Apex or XI Gene with a few different memory kits. Any other rookies have tips/tricks for this level of mem OC?
  14. Good to know, it was unexpected so I wanted to let other potential buyers know. Its unfortune are because the kit looks really good!
  15. Picked up a kit of these to play with and they are A1 pcbs. Just a heads up this kit is not all A2.
×
×
  • Create New...