Rampage V edition 10 no post but R5E does!? 3200cl14 quad channel

[TheGeneral]

Hey all,

So I have a R 5 Edition 10 and a 1680v3. I am trying to get my quad channel kit of gskill 3200 cl14 to post at xmp.

With bios at optimised defaults and nothing but xmp changed I am getting q code bd loops until q code cc appears. 2666 posts however.

The main twist to the story is, if I plug everything I to my spare R 5 Extreme it posts instantly at xmp.

Now call me stupid but surely as the R 5 E 10 is far newer you would think the memory algorithms are more up to date?

I've taken note of ALL the voltages the R 5 E has applied and will try to input them into the E 10 when I get time in the hope this may sort it.

R 5 E is at 3902 bios. E 10 is at latest bios released recently.

Only other think I can think of is to try a much older bios on the E 10?

Can anyone think of anything else?

Thanks

[Morkrah]

Hmm maybe,

I'm inclined to think that there's a timing somewhere that it doesn't like running that slot in - doesn't matter which stick I put in there its always fails training at xmp.

I haven't tried higher voltage or looser primaries yet so I will play with those.

Just odd that only Dimm B1 fails training at that speed and no others. I even tried dual channel B1 and D1 and B1 still failed training.

I've cleaned all dim slots, may try reseating cpu and cleaning socket (again).

Failing that I will need to source some tertiary timings from other QVL kits and see if it likes those.

The setup worked for about 24 hours at XMP, maybe the training is failing the offsets / skewing no

[Morlock]

I certainly understand the difference between my memory and G.Skill 3200CL14, definitely. )))
But I would like to object to you at the expense of my memory. This is not a server. Here, in your last link, the following is mentioned: SAMSUNG 8 GB DDR4 PC4-19200, 2400 MHz, 288 PIN DIMM, 1.2 V, CL 17 RAM MEMORY MODULE for M378A1K43BB2-CRC Desktop PCs. Not like unbuffered memory on the server side, not like unregistered memory, but like a desktop!
In addition, this memory was overclocked (not by me personally) by other people and quite successfully. Yes, the achievements there were not great, but, of course, they could exceed 2400 MHz. But I have a stop above 2400.
Tomorrow I will take a couple of G.Skill 3200c15 modules and try.
The main question that bothers me is: can Xeons 1600v3 work with straps?

[Vlada011]

Maybe they OC memory on other processors with different and better IMC. But It's not strange that Samsung 2400MHz 8GB Module only could work on 2400MHz.
You need to increase timings and voltage if you want to reach 2666MHz example... very rear memory could go level above specifications on same timings and 1.2V.
2800MHz is I think max frequency with default voltage 1.2V, fabric kits with higher frequency sometimes even 2800MHz work on 1.35V default.
I had two kits GSkill 2800 and Dominator 2666. Performance difference was 2-3% in favor of GSkill, but 2666 better fit with X99 and 100 strap.
I didn't saw memory kit 3200MHz with 1.2V. And it's very rear that any memory could be increased from 2400MHz to 2666MHz on same timmings, command rate, and voltage 1.2V.
First thing in such situation would be increasing voltage to 1.35V, increasing timings for 1 number save settings and leave to system boot.
Then when system boot with higher voltage and timings then you try to OC frequency. At the end even if you success result will be same as lower frequency but lower timmings.

I didn't OC my memory 2666 15-17-17... I only drop timings to 15-15-15 voltage was same and results is probably almost like little faster memory
2800MHz with 16-18-18.

[Morlock]

What is most strange and incomprehensible to me is that two G.Skill 3200CL15 2x16Gb memory modules could not work at their frequency.
XMP profile just did not load. It was selected, but besides the fact that it was indicated, not a single value has changed.
neither frequency nor voltage - nothing.
As a result, I was able to get 3200 / 16-18-18-18-38, but only with one module installed, at a voltage of 1.37. Two is only 2666, three or four is only 2400 otherwise the code is bd or b7.
In general, I decided to be content with my modules, which steadily work from 2400 / 12-13-13-22-22-1,3v.
Apparently in the future I will buy 5960X in order to unlock the possibilities of memory.

[Vlada011]

I believe as well Xeon is limitation. If you wait little more i7-6950X will become cheaper and more affordable, price is still high for his performance.
Special because new CPU with more cores, stronger cost 500$ now and people ask 500$ for used i7-6950X.
I didn't run XMP because timings are higher then I could set them.

[Morlock]

So far, in our country (Ukraine), the 5960X sells for $ 350, and the 6950X for $ 650. I thought it was such savagery in our country, but looking around the world, I realized that it was just universal human greed. Especially considering that 10 X99 cores in 2016 cost $ 1,750, and in 2017 10 X299 cores - already $ 1,000. Therefore, they do not want to reduce prices as much as possible. Prices too sharply and unexpectedly fell then.
* Considering that 1650v3 cost me only $ 100, I was not very upset, although my friends have the same processor (1650v3) with a strap of 132.4 MHz and 3178 MHz of memory. I will stay with him a little and gain experience with v3 processors, I use different modes. Later I will purchase an I7 line processor, not a damned E5. Once to buy 6950X - it's worth it. But a little later. )))

[Vlada011]

Yes it's greed. You can't explain to owners of i7-6950X that CPU new with 3 years warranty worth 350$ beat him in single and mult threaded applications.
In my country very poor people can't buy Xeon with 6 cores for LGA 1366 (they cost 20 euro) because owners of old X58 motherboards with SATA III and USB 3.0 ports asked more then 150$ for motherboards. They would be able to build whole system for less then 100 euro with memory but people with X58 motherboards wait to mobo die to someone who have i9-990X or similar and 12GB DDR3 and they have no choice. 150 for motherboard on new system for 500-600-700.
i7-5960X used worth 200. i7-6950X worth 350 max.


i7-5960X overclocked to 4.5GHz give score 1700-1800 in Cinebench R15
i7-6940X on 4.4GHz on Youtube I saw score 2.329, 1000 points less then R9-3900X Even single score is weaker then AMD 190-200 example.
AMD Ryzen R9-3900X give 3200 score.

And how someone could ask 150 more for CPU without warranty seriously weaker then new CPU with warranty with better IMC and newer PCI-E Interface.
I don't understand that. Performance of i7-6950X say that he should be 100$ or euro cheaper then R9-3900X. But if we talk about new CPU in warranty.

People charge Haswell-E and Broadwell-E used like AMD not exist.
They forgot what happen on market and that every generation Intel have bigger problem to stay competitive with AMD.
Now can't follow them any more and no news on horizon that situation could be different. They advertise new platform with obsolete PCI-E Interface and people will figure out all flaws after upgrade on Intel.

Don't even to talk about stores who still sell old stocks and offer such models. That's price 1K or more. You can build R9-3900X and motherboard with memory for price of i7-6950X.

[Morlock]

There is nothing to add to what has been said. I completely agree.:)

[Morlock]

Tell me what else needs to be changed so that the memory overcomes the border of at least 2666 MHz in four-channel mode. The system boots at a frequency of 2666 MHz (up to 17-17-17-39 - unchanged), but with a loss of channel B. At a frequency of 3200 MHz, the loss of channels B, C and D.
At 2400 MHz - even at 12-13-13-22 everything is perfect.
Screenshots are attached (there are no final voltages - during experiments with memory it is loaded up to 1.38 V).
https://drive.google.com/drive/folders/1U7_NPegpApQg-gRWBDmj3zw2INL4muCF

[Morlock]

In general, as I understand it, for Haswell it is R5E, and R5E10 for Broadwell.
Hence the misunderstanding of XMP when working with the v3 process and other sadness.
It is sad. Mobo is not bad, but absolutely not universal.
Although if we forget about XMP support and use exclusively manual overclocking, then everything will be approximately the same.
By the way, how far does the 5960X in the maximum reasonable overclock move away from the 6850K in the same maximum reasonable overclock for a constant time?
And how hot is Broadwell?