Memory Type
DDR4
Capacity
32GB (8GBx4)
Multi-Channel Kit
Dual/Quad Channel Kit
Tested Speed
4000MHz
Tested Latency
18-22-22-42
Tested Voltage
1.35V
Registered/Unbuffered
Unbuffered
Error Checking
Non-ECC
SPD Speed
2133MHz
SPD Voltage
1.20V
Fan Included
No
Warranty
Limited Lifetime
Features
Intel XMP 2.0 (Extreme Memory Profile) Ready
Additional Notes
Rated XMP frequency & stability depends on MB & CPU capability.
I turned off XMP and the gamemode stuff in the bios, the ram should be at a stately 2133mhz now. It still gave errors in the windows memory test, then when I ran checks on the sticks by removing two, they all checked out, as did both pairs of slots. WTF.
Check the MB bios for an update. Are you running Multi Core Enhancement or any other tweaks on the CPU?
none. I updated the BIOS almost the first thing when I installed it and got it working on a single stick. It’s been a journey.
I like “normal” hardware that does not push limits by too much. There’s a market out there for people that squeeze 1% out of everything, but it’s just not big enough to be critical in mass.
With RAM I am not even sure if high clock speed with long waiting times is faster than slightly lower clock speed and less waiting cycles.
Looks like maybe you have a heat and/or voltage issue. Not necessarily on the RAM sticks, but maybe the stuff that drives it? Hard to tell.
Its possible you just got unlucky with another bad stick. The only time I’ve ever had issues with ram and XMP is with OC’ing the cpu pretty hard. Is the voltage showing 1.35 for XMP and 1.2 without?
yep I’m regretting my choices. On the other hand. When it works… It was so freakin’ smooth!
according to CPUID they are
I’m running a slower ripjaws kit (3200) and its great ram. Defnitely a gremlin in there. If the CPU clocks are untouched, and the ram is untouched, then the issue is either the stick or the slot.
CPUID is known to be fuzzy sometimes with voltage calculations, but if BIOS settings confirm it, I’m out of ideas other than sending the sticks back again, and if that doesnt do it, its gotta be the board.
Just so you have a reference, here’s my CPU-Z:
Thinking of buying another kit of XMP certified ram and sending this kit back.
Can you get a screen shot of the bios settings for the ram?
SPD is basically what the ram stick is saying it’s capable of, identifying itself.
Maybe the MB is misinterpreting something or a setting isn’t ticked.
sure, not easily but I can. DCS is sucking down tens of jiggerbytes so I’m not rebooting it now.
I put it on default settings last time I saw it today. Made sure XMP and the Extreme Game Mode CPU overclocking thing were turned off.
With that and the BIOS update you can also reset BIOS to factory defaults again, just incase something is out of whack like what @Gunnyhighway is suggesting. It may accomplish nothing, but it shouldn’t hurt anything.
From what I have read, this is the case - latency will matter more for us as the RAM is not changing data much - load map and done. Now, if you don’t have sufficient RAM - say 16GB or less, then the clock speed bonus may be a benefit as the data in RAM changes more often. So even before latency, having sufficient RAM is key.
The only cases I have heard of where speed mattered was in large databases - data is all stored in RAM and manipulated there for high speed - then dumped to disk when changes are confirmed. This is still below XMP values as well, as you can imagine overclocking your critical database server would be frowned upon.
I been away today, day out with the kids. While away, I let the box run idle, just to see what it did. Well, not entirely idle, I had it suck down all of DCS again, to make sure there were no leftovers from old shader mods and whatnot.
When I got back, fed the children and put them to bed, I ran a memory test (the built-in windows one). It ran to completion on its default settings without errors.
Next, I googled for and found a GPU testing tool named OCCT. I stress-tested the GPU and VRAM for a bit. They check out. Next up, a complex DCS scenario. Let’s see how that shakes out.
If everything is working on default, there may be a stability issue between the ram and cpu when either or is overclocked. Probably just have to find the sweet spot through testing. Sometimes with higher speed ram, the CPU needs to be OC’d as well or they wont like each other. I’ve been doing a little research for your because I literally have nothing else to do.
I would turn XMP on back to 4000 and see if the problems persist. If it does, I would go to Historical Binning Statistics – Silicon Lottery and set the CPU to the 100% settings for a base OC. If that solves it, carry on at 4000mhz. If it doesn’t, put the CPU back to stock, and manually set your ram (non XMP) to something lower like 3200 and 1.35v and see what happens. You can get the clock timings from any Ripjaws kit (check Pony’s above) rated for 3200, they should work, could possibly work at the same timings you are using now. It’s possible there may be an issue with the CPU not being able to keep up with the ram.
EDIT: Just realized the 9600kf isnt listed, you can probably find decent setups online if you aren’t comfortable testing on your own. I think the general rule of thumb is set the vcore to 1.35v and let it rip.
Cool, thanks! That’s some pretty interesting information.
I will probably start experimenting with upclocking the thing a while later. For now I’m just glad the blasted thing works. I flew a bit just now, and I could feel it wasn’t running full tilt, but it didn’t BSOD or CTD and that’s a win.