Bullant Posted January 24, 2019 Posted January 24, 2019 Thought it be fun to start memory oc thread on 1156 platform, I've not long been testing this platform but I like it, here's a couple runs I did a while ago on cold PSC. I've spent little more time on it since then and will post some more results with PSC and hypers. All welcome to post some results, if you don't you know I will ?let us know your memory and board of choice, I currently using the P55a ud5 Gigabyte and have a P55a ud6 on the way to try 1 Quote
TaPaKaH Posted January 24, 2019 Posted January 24, 2019 (edited) Some of my Hyper runs from nearly a decade ago LN2 RAM on mini ITX before it got mainstream Edited January 24, 2019 by TaPaKaH 3 3 Quote
ground Posted January 25, 2019 Posted January 25, 2019 Sadly this CPU died from too much VTT (hint: these CPUs die instantly at 1.6V VTT on air, at least mine did) before I could finish my testing (and perhaps get a tuned OS up etc), but I've ran 2000 5-8-5-18 at 1.84V on dual Channel BBSE on air in the early stage of testing a, at the time, new to me BBSE kit (which does 2000 6-8-6-18 at 1.64V) . I've recently tried replicating the run with a new CPU (and better efficiency/subtimings), but couldn't get the board to boot cas 5 whatsoever, testing multiple kits even at super low frequency (even 800 5-5-5 refused working on Hypers). I suspect my board is the issue as it has many of those. May retry with older BIOS versions, I think the run was on one of those... Setup: H55M-UD2H, i5 750, GSkill ReaperX 2200 9-9-9-27 1 Quote
TaPaKaH Posted January 25, 2019 Posted January 25, 2019 (edited) I used to go mad about IMC binning on this platform, just to be able to do spec with newly-released 2300/2400/2500/2533/2625-rated PSC kits. I went through a few dozen of i7-8xx and highest I found was 1250. Don't remember how much VTT it took. Also no idea where this CPU is now. I think I sold it at XS, OCX or luxx. Edited January 25, 2019 by TaPaKaH 1 Quote
Bullant Posted January 25, 2019 Author Posted January 25, 2019 Love those pots sam ? they look awesome, vtt for me tops out on most my chips at 1.54v -30 CPU Quote
ground Posted January 25, 2019 Posted January 25, 2019 (edited) Yeah, my 750 kept scaling and I expected it to take similar voltages to 1366 (where I've even seen 32nm batches that took and scaled to 1.7V VTT without caring/degrading at ambient) which it obviously didn't like... And gotta agree, that setup looks amazing Edited January 25, 2019 by ground1556 1 Quote
Bullant Posted January 25, 2019 Author Posted January 25, 2019 I see little power button setup in the second pic ? Quote
Bullant Posted January 26, 2019 Author Posted January 26, 2019 Ran the 1156 Platform up today....870 ...P55a Ud5 with cold PSC....Memory temps ranged from -110 to -140....little more learnt today 2 Quote
Noxinite Posted January 26, 2019 Posted January 26, 2019 (edited) All the ice. Also, 2820C7. ? Why RCD 9 for the 2400MHz runs though? Edited January 26, 2019 by Noxinite 1 1 Quote
Bullant Posted January 26, 2019 Author Posted January 26, 2019 (edited) 10 minutes ago, Noxinite said: All the ice. Also, 2820C7. ? Why RCD 9 for the 2400MHz runs though? Was untested cold memory and memory wasn't fully prepped correct, you know heatsinks on. also keep dram volts little low.....When I run again ill use correct pot and memory with no Heatsink on. it should do trcd 8 then and maybe at 2500Mhz....got to save some stuff for improvements later hey? ahahahha Is quite possible that memory was under -100 as I had the probe in the top of the memory heatsink lol Edited January 26, 2019 by Bullant 3 Quote
kicsipapucs Posted January 26, 2019 Posted January 26, 2019 Nice thread Jayson! Can't contribute, but I read all you guys' posts. 1 1 Quote
ground Posted January 30, 2019 Posted January 30, 2019 Been checking through some scores on the 32nm chips and I've observed something vaguely related to memory - few people are running high QPI, but it actually makes a massive impact. My own example is R15, but looking at @Bullants 32m Score on his 540 confirms this. link cb score cb % core core % dram Timings dram % QPI clock QPI % https://hwbot.org/submission/4039667_ground1556_cinebench___r15_core_i3_540_410_cb 410 100.00% 5379 100.00% 1024 7-9-7 100.00% 5123 100.00% https://hwbot.org/submission/2509986_joe90br_cinebench___r15_core_i5_670_408_cb 408 99.51% 5687 105.73% 842 9-9-9 82.23% 4213 82.24% https://hwbot.org/submission/2493991_der8auer_cinebench___r15_core_i3_560_436_cb 436 106.34% 6075 112.94% 971 7-7-7 94.82% 2916 56.92% https://hwbot.org/submission/2476431_michel90_cinebench___r15_core_i3_550_413_cb 413 100.73% 5711 106.17% 952 8-11-8 92.97% 2855 55.73% https://hwbot.org/submission/3377198_kintaro_cinebench___r15_core_i3_550_417_cb 417 101.71% 5757 107.03% 720 6-6-6 70.31% 3838 74.92% https://hwbot.org/submission/2474297_topdog_cinebench___r15_core_i3_550_406_cb 406 99.02% 5649 105.02% 706 6-7-6 68.95% 2624 51.22% https://hwbot.org/submission/2665450_ue50_cinebench___r15_core_i3_540_405_cb 405 98.78% 5451 101.34% 948 8-11-8 92.58% 4740 92.52% Of all these scores only the bottom one is somewhat close in efficiency, but its also the only one with fairly high QPI clocks. The same applies to 32m, that can be easily seen by just looking at the 32m rankings for the i3 540 1 Quote
Bullant Posted January 30, 2019 Author Posted January 30, 2019 (edited) Yeah qpi gives good boost on the "i3" chips but don't expect to see 350-400mhz boost in eff from qpi like seen in the 32m results ? Edited January 30, 2019 by Bullant 1 Quote
TaPaKaH Posted January 30, 2019 Posted January 30, 2019 (edited) Actually, QPI was known as the method to improve clock-per-clock efficiency on Clarkdale since its introduction It has to do with memory controller and core sitting on physically separate dies with QPI being the clockspeed of the bus between them. This also explains why Clarkdale is so terribly inefficient compared to contemporary Lynnfield / Bloomfield / Gulftown, all of which have IMC and core on the same die. Edited January 30, 2019 by TaPaKaH 3 Quote
Crew Don_Dan Posted January 30, 2019 Crew Posted January 30, 2019 (edited) Like Sam said, this was probably known by anyone when Clarkdale was benched hard. Just take a look at the i5 670 32M ranking for example. Edited January 30, 2019 by Don_Dan 1 Quote
ground Posted January 31, 2019 Posted January 31, 2019 I was mostly looking at the top scores for the i3s, where the QPI is usually quite slow from what I've seen... (are there no spoilers? ) https://hwbot.org/submission/2572691_der8auer_superpi___32m_core_i3_530_7min_8sec_937ms 3.2 QPIhttps://hwbot.org/submission/969132_jabski_superpi___32m_core_i3_530_7min_7sec_625ms 4.66 QPIhttps://hwbot.org/submission/979342_sacha35_superpi___32m_core_i3_530_6min_59sec_656ms 4.3 QPI https://hwbot.org/submission/4053113_bullant_superpi___32m_core_i3_540_6min_59sec_375ms 5.3 QPIhttps://hwbot.org/submission/969146_hicookie_superpi___32m_core_i3_540_7min_1sec_735ms 3.6 QPIhttps://hwbot.org/submission/969762_lodewijk1978_superpi___32m_core_i3_540_7min_3sec_156ms 3.7 QPI https://hwbot.org/submission/2503366_strunkenbold_superpi___32m_core_i3_550_7min_28sec_531ms 4.81 QPIhttps://hwbot.org/submission/2776005_wanted_superpi___32m_core_i3_550_7min_25sec_812ms 4.5 QPIhttps://hwbot.org/submission/2776005_wanted_superpi___32m_core_i3_550_7min_25sec_812ms 5.1 QPI https://hwbot.org/submission/971207_hicookie_superpi___32m_core_i5_650_6min_37sec_78ms 3.9 QPIhttps://hwbot.org/submission/968980_ekky_jengkol_superpi___32m_core_i5_650_6min_43sec_219ms 4.3 QPIhttps://hwbot.org/submission/968980_ekky_jengkol_superpi___32m_core_i5_650_6min_43sec_219ms 4.4 QPI https://hwbot.org/submission/2226249_cl3p20_superpi___32m_core_i5_660_7min_14sec_625ms 4.3 QPIhttps://hwbot.org/submission/959023_topalof_superpi___32m_core_i5_660_7min_1sec_140ms 3.2 QPIhttps://hwbot.org/submission/962339_3oh6_superpi___32m_core_i5_660_6min_11sec_438ms 5.34 QPI https://hwbot.org/submission/964253_hiwa_superpi___32m_core_i5_670_6min_8sec_859ms 4.6 QPIhttps://hwbot.org/submission/968355_nickshih_superpi___32m_core_i5_670_6min_2sec_625ms 4.9 QPI From 17 top 3 results, a whopping 3 are above 5GHz QPI, not even half above 4.5 which appears a little odd from the above mentioned. Just found it weird when my random-ish chip (best out of a dozen chips) manages 5.1 easy on dice... Quote
Bullant Posted January 31, 2019 Author Posted January 31, 2019 (edited) Yeah my couple i3 chips do 5.3 at -30, not tested water.....i know also after now testing one other board that there is a big difference in qpi scaling....my same chip can't do as high qpi as in my current board. So I'll prob look to find some other cheap boards to try Edited January 31, 2019 by Bullant Quote
Bullant Posted February 1, 2019 Author Posted February 1, 2019 (edited) I wonder if 3000mhz memory is possible cl7-8, might chuck on the ek memory pot soon to see how far it can go ? Edited February 1, 2019 by Bullant 1 Quote
Bullant Posted June 11, 2019 Author Posted June 11, 2019 New chip seems ok imc, well better then my i7 870.....cpu water and memory air...will cold memory this soon Quote
Noxinite Posted June 11, 2019 Posted June 11, 2019 Stable at those settings? Nice regardless though. Quote
Bullant Posted June 11, 2019 Author Posted June 11, 2019 7 hours ago, Noxinite said: Stable at those settings? Nice regardless though. No not stable mate, looks good to me for 1156......I'm sure the memory will fly once both CPU and memory are cold. Hopefully the chip does good freq too 1 Quote
unityofsaints Posted June 17, 2019 Posted June 17, 2019 On 1/31/2019 at 12:23 AM, ground1556 said: (are there no spoilers? ) [spoiler][/spoiler] works fine Quote
ground Posted July 23, 2019 Posted July 23, 2019 (edited) Anyone here has experience getting hyper to run with 32nm 1156 chips? We have a couple 650s and 550s here that won't post with hyper at anything beyond 800 MHz if even that while handling PSC at >1050 or so... edit: and once again noting that almost nobody else running high QPI which makes it easy Edited July 24, 2019 by ground1556 Quote
unityofsaints Posted August 16, 2019 Posted August 16, 2019 On 7/23/2019 at 8:23 PM, ground1556 said: Anyone here has experience getting hyper to run with 32nm 1156 chips? We have a couple 650s and 550s here that won't post with hyper at anything beyond 800 MHz if even that while handling PSC at >1050 or so... I will give them a run soon, will let you know how I go. Although I only 1 stick of Kingston and several kits of Corsair Hypers so my results may not end up being representative. Quote
mirco69 Posted February 5, 2020 Posted February 5, 2020 On 1/25/2019 at 1:13 AM, TaPaKaH said: I used to go mad about IMC binning on this platform, just to be able to do spec with newly-released 2300/2400/2500/2533/2625-rated PSC kits. I went through a few dozen of i7-8xx and highest I found was 1250. Don't remember how much VTT it took. Also no idea where this CPU is now. I think I sold it at XS, OCX or luxx. Hey TaPaKaH, can you give a brief summary of your procedure to bin CPUs for good IMC? cheers Quote
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.