I.M.O.G.

Damn. I think you got me. EDIT: Nice work. Good job. *insert other socially appropriate nice things*

Thanks for the input. The multi limit and the memory ratios are the kickers. I will probably go ahead and sell it and get something else.

What are the problems with using ivy on the MIVE? I have heard of problems, but I dont remember what they were. Looking for my next full pot ln2 board, and I have one of these I am trying to sell currently. I have 2 dead mvg boards currently, so looking at cheap options.

Back on the L204B321, cold bug a bit below -110C. Another crippled Asus MVG... Ran great for about a month. I don't think I even want to subject myself to Asus RMA process. I'll probably eat the $400 in crap boards and just get something that hopefully doesn't fail after a few weeks. On the upside, hopefully one of these other chips I have here don't suck on whatever board I run next. Wasted 3 months now on Asus junk, and overall my Ivy Bridge experience has been nothing short of insanely frustrating as a result.

3218B957 doing the same thing as 3218B958 - crap. Warming it back up, tearing down, and putting known-good L204B321 back in to see if its the board or the chips.

Good deal for you Reggie. I had the same cold issues on a Maximus V Gene. Known good 3770K, but every chip I dropped in cold bugged at the same point and the board would act funny... Since it was covered in vaseline I didn't bother pursuing RMA, and just bought a second board. You are lucky!

I had 3 with low cold bugs, couldn't believe it so I then bought a 2nd MVG, and found it was the 1st MVG that was causing the problem, and that my 3rd chip was a decent one (6.3GHz) and there was no coldbugging. My first 2 chips might have been fine too, but I sold each one to buy the next one and never got to test them on a good board. So I have a known good MVG now, and a known good L204B321. The 3218B958 yesterday coldbugged - either its the chip, or something went bad with my board/insulation. I'm going to test my next CPU (3218B957) today, and if it coldbugs as well... Then I'll put the L204B321 in and see if that still works without any problem.

This was 3218B958: CBB at -105C, coldbug at -117C. First tried the exact profile which worked fine for my 6.3GHz L204B321, then did the following trying to eliminate cold bug. Tried 1.25V VCCSA and VCCIO, 1.6V to 1.7V core, 107.5-112.2MHz bclk, 2c/2t-4c/8t, high ram freq (2700) to low freq (2100). Tried keeping it loaded with sp32m, tried ucbench, wprime, nothing has any difference on when the CB hits. Best I saw on the chip was 6.1GHz+, at 1.7V -105C... So it isn't a bad chip, probably even a pretty solid DICE chip, but the cold limit puts a major damper on things. Only ran it for about an hour. Gonna try 3218B957 tomorrow, hoping for better luck.

There's a bug logged in the tracker. Massman just posted the other day in reply to my question, letting me know it was not scheduled to be implemented, but its out there. EDIT: Here's the posts (I asked in a separate thread because the tone of this one felt icky, and I was just interested in what will happen to big globals because its relevant to my spot in the US rankings):

Mostly unrelated, but I always had a vague curiosity of why CPUID.com distributed the application, but validation was handled at canardpc.com. Never was curious enough to ask or look into it, but I always thought that was weird.

Thank you. I'm going to keep doing what I've been doing, but will try that as well to see if it makes any difference. I found LN2 mode on MVG or not, I have the same clocks on my L204B321 (6.3GHz). LN2 mode increases the "auto" voltages on various things, so I've run non-LN2 mode to take it easy on the chip. I haven't messed with memory settings much to see if the other voltages would help memory speeds, but for core frequency alone, the only thing that made any difference for me was vcore.

I have a question for you guys... What settings are you changing in BIOS to bin chips? On my L204B321, the only settings I changed were dram voltage and vcore for 6.3. Everything else was left on auto... I tried changing other settings, but it seemed to make no difference on my chip for max cpuz. Are you guys changing anything other than vcore when testing for max cpuz? Great volts!

You sure the 2000's you got are duds? I don't know as I haven't ran these kind of c2d's much, but I had a HELL of a time getting UCBench to run on my e2200. Something weird it seemed like where UCBench would bomb out with the e2200 - haven't seen it act that way on any other chip I've benched (not that many). I did 4.4 in ucbench, but I could do 4.7 in wprime1024, 4.9 in wp32, 5GHz in CPUz. I didn't end up spending much time on it since it wasn't worth anything, but if you are judging your chips currently based on ucbench, figured I would mention my personal experience. I plan on working on ucbench more this time around since its worth more time due to the competition.

Nice score, looks well tweaked!

My current chip is L204b321, and does 3d at 6.2ghz@1.72v. sp1m at 6.38@1.8v. Just picked up today at microcenter: 3218b957 and 3218b958 Testing this weekend and will report back.

Yes, you can, and it works. However, if using the unlocked dual core CPU in the x2 stage, you must run it with only 2 cores activated.

Did you see my reply above? You are filling out a lot of fields which are not required, which make submitting harder at first to get right: Fill out only the required fields. Then if the submission is accepted, you'll be taken to the result page - you can edit from the top right, and fill in additional field like ram settings, motherboard, or whatever else you want to do. You can edit and fill them in one by one, to see which ones are giving you problems... This way you aren't trying to fill out 150 fields at a time.

That's a stupid error, I would ignore it and change the way you are trying to submit. For CPU frequency submissions, these are the only fields you have to fill out: Benchmark score (frequency) Processor (begin typing your CPU model, it will drop down a list, click on your exact processor model) Compare URL (This is where you past in the valid.canardpc.com URL for your validated result) You also have to post a picture of your rig since you are on air cooling, which you do within the "pictures of your system" box near the bottom. If you just fill out those fields, it should accept your submission and not give you an error. All other fields are optional, and filling out one of them is probably causing the stupid error.

Thanks, looking for more competition at the top end of the UCBench rankings. Currently its pure CPU frequency at the top, but it seems there are tweaks that have been proven out for many people, but no one is using yet on LN2, or the tweaks don't work the same on LN2.

Yes, thats always permitted in pcmark rankings, so its valid in these stages. 555 is dual core, so it can be submitted as dual, tri, and quad.

Nice work. Last I tried to validate using 1.60, it wouldn't accept it at valid.canardpc.com.

Is there more than one version of version 1.61? I thought the installable and zip were the same. I think the bug he's describing was the same problem with the 8.5GHz validation that was pulled.

+1 for profiles. I'm often changing resolutions after the bench finishes and before I take the screenshot, depending on what I'm running exactly. Profiles would be essential so that the tool quickly works for whatever I'm doing currently. Also would be nice if it could be configured to apply a desktop background when it launches the apps. I'm always benching with no background, but that isn't what I want to use when I take the screenshots. Changing desktop backgrounds is cumbersome, so I often don't bother to change the background for the screenshot... If this tool could also quickly set the background, that would be perfect. I don't do watermarks like a lot of guys do, but I'd like to use a background with my site logo and my username. Other guys might like it to watermark the screenshot, more so than set the desktop background.

That's an issue for every team DopeLex. Curiously, it only seems to be an issue for users who have submitted recently... Users who have not submitted recently, still show more UP than league points. For example, my UP and league points match on my team members tab. However Jiccman1965 hasn't been active for a few months, and his UP are still higher than his league points on the team members tab.

That has been a red herring in my experience for LN2 benching. Different rules may apply on air/other cooling when you aren't stressing the limits of stability. On the second run, if your clocks were maxed, you should not be stable enough to run 8-16-32-64 threads sequentially... If your clocks were maxed you could complete 1 out of those 4 tests if selected alone, but one after another would cause instability... Similar to the ability to complete wprime32 vs wprime1024. The shorter tests run at higher clocks - independently 8, 16, 32, or 64 may complete when ran alone, but will fail when ran in succession. I've never been able to get higher scores at lower clocks on ucbench by tweaking... It's best to tweak, then do your run at the optimal settings you have found with your max clocks. If I'm wrong and running more thread tests increases the score more than maximizing clocks, someone should be able to beat me in this challenge: http://hwbot.org/challenge/i.m.o.g.s_ucbench_2011_global_challenge___jul_4_2012_until_aug_3_2012/ As is usually true with hwbot, lower in the rankings there are often many exceptions with scores versus clocks and how things scale... But at the top of the rankings where the competitors typically all have learned how to optimize well, there is more consistency with clocks vs results. EDIT: There's no guarantee I'm right. But the only 2 scores beating mine in 4x CPU class currently are CPUs that can run higher clocks: http://hwbot.org/submission/2291870 If UCBench were worth globals, there would be more people pushing it to its limits. Unfortunately, there are few people really trying hard at the top of the UCBench rankings.