TaPaKaH

I ran high memory voltages (2.5V+ DDR3) with every generation of Intel's IMC without any signs of damage to either memory or the processor. So I see no reason to think that things are going to be different with Skylake.

I haven't yet seen any Samsung-based Kingston DDR4. Samsung has always been a rare sight on Kingston for business reasons so it's highly unlikely that we will get retail E-die HyperX.

My kit scales up to 2.2V on air. More I can't set in board's BIOS.

On Skylake 32M times can vary by more than 5 seconds without you touching any of the settings. So I wouldn't refer to 32M times as a sort of reliable performance comparison. Have you tried raising tRCD/tRP to 21 or 22?From my experience on E-die, tRCD and tRP responds to voltage far more than what we've seen on any generation of Samsung, DDR3 or DDR4.

What memory multipliers are available with these i3 (and locked Skylake in general)?

I've tested 3466C16 TridentZ and 3600C17 RipjawsV today, both were D-die. So you have to be really lucky to get E-die.

Yes. Can even do an HCI Memtest at DDR4-3600 CL11 CL10 boots but 2.20V is not enough.

Because result reliability of XP is crap with newer platforms (Skylake especially). I can run same settings 5 times in a row with/without in-between rebooting and get results that are 4+ seconds apart.

You need to press F5 right after windows setup loads for the first time (the bit before formatting drives) and select MPS Multiprocessor mode.

2666C15 1.2V is actually the hardest spec in Kingston's current retail range.

German 1-2-3, too boring

$1K CPU and $1K VGAs deserve $1K thermal paste, obviously.

Finally, somebody has noticed!

good morning

1466 9-12-12 boot is easy on many kits, but 32M is a totally different ballgame.

good afternoon

Whiteboard behind leaks important internal info

Up for sale are some hardware leftovers: 1) Core 2 Duo E4300 (M0) = sold to Leeghoofd Batch L026B356. Is above average on FSB (462MHz on air) and is very good on MHz/volt (3.6GHz 32M at 1.38V) for an Allendale. 2) Core 2 Quad Q6600 (G0) = 25 Euros Batch L734B442. Is quite decent on FSB (514MHz on air). 3) Core 2 Quad Q6600 (G0) = sold to willemc700 Batch L737B477. Is also quite decent on FSB (520MHz on air). 4) 2x4GB G.Skill TridentX 2600C10 (Samsung) = sold Two modules from a quad kit. One module was damaged in shipping (broken corner) and does not POST. The second module works fine, needs around 1.98V for 1400MHz 9-12-12 tight. In all cases, price includes international shipping.

AIDA64 can be used to show SPD with XP.

I don't know of a simple way to flash SPD with DDR4 so you might as well add the CPU-Z SPD tab requirement

This is partially true. 1.65V on 2000C9 Tridents implies BBSE (or 2Gbit BCSE if you're unlucky) but 1.60V can be anything.

There are many G.Skill kits that can do 877 6-7-6. Pretty much anything that is based on Elpida Hypers and BBSE can reach these numbers. Some PSC and BDBG might be capable too.

Failure to obtain a job in the hardware industry, even one that is not related to overclocking, is only one of the subreasons why I decided to part with the competitive side of it (e.g. hwbot). The main reason is money. I haven't been doing too well financially for the last couple of years and I no longer see any overclocking direction where I could entertain myself on the cheap: my 2D hardware points are at a dead end, 3D hardware points are too expensive since most point-heavy cards are still CPU-dependent and rejects are impossible to resell, the latter also counts 8.5GHz+/200%+/DDR1/DDR2 projects out, DDR3/DDR4 binning has been killed off by manufacturers seeding special kits and putting rejects on retail, etc. Since my interest is now elsewhere, I thought it doesn't make sense to waste time on "account maintenance" and asked the staff to get it wiped ... also letting people like Alby to have a chance now

I had the same, you have to ask hwbot to remove your account/points manually. Takes about a week.

Kit sold and shipped long ago.