(DDR4) Micron D9RGQ

[TaPaKaH]

some SuperPi 32M test runs with four Crucial 2133C16 OEM modules:

DDR4-2666:



DDR4-3000:



DDR4-3333:



Getting 1500MHz on 8GB modules is already hard. Only one of four modules I have would even POST at 1666MHz.

[Charles Wirth]

Looks good, nice to see 2666C10 with DDR4

[Hyperhorn]

I see, you've also set Sensor=0 and SPD=0, because it takes forever when enabled.
Can you use DDR4-2666 with 100 MHz BCLK strap? I can.

I suppose these 8 GiByte modules are dual ranked. So how are your raw performance numbers at Aida64? These are mine with SK Hynix single-ranked sticks from G.Skill (F4-3000C15Q-16GRR):

Core: 4,375 MHz
Cache: 4,000 MHz
BCLK: 125.0 MHz
RAM: 4 x 4 DDR4-3000 15-15-15-36 2T RFC 390 (not maxed out, I've used these timings for comparison reasons)

Read: 73,472 MiB/s
Write: 62,210 MiB/s
Copy: 72,470 MiB/s
Latency: 54.1 ns

Now at DDR4-2133 15-15-15-36 2T RFC 360:

Read: 57,592 MiB/s
Write: 59,849 MiB/s
Copy: 53,203 MiB/s
Latency: 67.9 ns

System: Core i7-5960X, Asus X99-Deluxe (UEFI 1601), Geforce GTX 780 Ti; Win 8.1U1 Pro, Geforce 337.88 (WQHL), Aida64 v4.60.3120 Beta

I have a MSI X99 SLI Plus at hand, but no time to compare the results with the X99-Deluxe. UEFI 1601 brought a huge improvement over earlier versions, before this version the performance numbers were way off.

[Charles Wirth]

On early BIOS it helps to run higher bclk to run higher ratios. Setting ratio to 1.25 helped make the 2400+ jump.

BIOS's are still being updated, I see the new 0502 will allow R5E to use 30x ram ratio with 100 bclk

[TaPaKaH]

100:2666 was fixed on MSI back in July, but I still prefer to pretest with 125:2666 as same profiles can be directly used afterwards on LN2.
I couldn't get AIDA to run on XP (it just blanks when I start benchmark) but I remember read numbers being around 60-63k on Win7.
Main problem on this platform, so far, is the NB clock. Very few makers can actually run it above 3800 and 37xx is insufficient to see the true gain from 1333 to 1500.

[andressergio]

Amazing job bro binnes or just random ?

[NEOAethyr]

What kind on numbers where you guys seeing on the older bios'es in aida?

[Don_Dan]

Thanks for sharing, Sam!

[vario]

Hi, i bought Some cheap crucials, and they have micron D9RGV chips, anyone knows what to expect and what to aim for ? Im still waiting for mobo so i will be testing monday at the soonest

[Charles Wirth]

Shoot for 125:2666C13 1T @ 1.35v, go 1.4v if needed.

[XBrytanX]

Thanks Sam ,.. dual rank 8GB OEM inbound

[l0ud_sil3nc3]

Shoot for 125:2666C13 1T @ 1.35v, go 1.4v if needed.

2666 c10 good with 1.55v

[l0ud_sil3nc3]

From the other night. I think if these dimms react well to cold, results could be decent maybe even comparable to single sided MFR.

[TaPaKaH]

Is that 4GB or 8GB sticks? I could never tun tRAS and tRFC so tight on mine at 1333.

[l0ud_sil3nc3]

Is that 4GB or 8GB sticks? I could never tun tRAS and tRFC so tight on mine at 1333.

8GB sticks Sam, so given the density clocks are pretty good for OEM memory with no heat spreaders I think

[vario]

Shoot for 125:2666C13 1T @ 1.35v, go 1.4v if needed.

So, i was in AMD for like forever and these new intels arent as easy as expected.For the time being i want to test my mem max stable and efficient settings, and cpu at 125 got funky.And at 100/2666 even c14 wasnt stable(after 2hours of prime but still) so i figured these 4 channel bad boys have plenty of bandwith, maybe they need some low timings tho.
Anyhow, i tested somewhat 2400 C12-10-10-40-300 and seemed stable, now testing 11-10-10-31-280 all at 1.4v
A question, any best and fastest memory testing method preferably in windows ?Currently im employing a 2xmemtest windows plus 2xthreads of prime95 blend and normal pc usage at the same time.C10 at 1.4v was a no boot so probably not gonna happen .
Anyway, what is the sweet spot for these systems ?I want a 24/7 rock stable machine that will do gaming/encoding (also at the same time) What is more important speed/timings ?
Also one weird thing, i bought 4 modules ,all of them look the same ,packaged the same, 3 of em are read CRUCIAL by any kind of software but one is read MICRON and has slightly different SPD entries, normal/weird ?

[iboomalot]

I find it interesting the faster the mem. clock but worse the timing = slower computer

3333 was slower than 2666 kinda sad

[Don_Dan]

I find it interesting the faster the mem. clock but worse the timing = slower computer

3333 was slower than 2666 kinda sad

You have to keep in mind that timings had to be loosened a lot to go from DDR4-2666 to DDR4-3333 plus the DDR4-3333 result was done in single channel instead of quad channel.

[Shiranui Gen-An]

Picked up 4x8GB of these, definitely not as good as the ones you guys are working with.

[TaPaKaH]

tight subtiming tests at 1.46V with X99 SLI Plus and X99 Xpower

[Charles Wirth]

I have some D9RGV inbound, hopefully some decent numbers.

[TaPaKaH]

Micron ICs are very weird. I have twelve 8GB modules in hands right now, 4pcs week 1424 (USA) and 8pcs week 1428 (Puerto Rico).

On MSI boards (X99 SLI / Mpower / Xpower) all w1424 sticks run 1500 flawless (boot in single, boot in quad, 32M in quad). But only four out of eight w1428 sticks can boot 1500 in single (mem slot doesn't matter, boot in quad or stability I didn't check).

On R5E neither of w1424 sticks can even boot above 1200 irrespective of timings/voltage/slot. Out of eight w1428 sticks, six boot 1500 in outer red slots but none boot 1500 in inner red slots. All w1428 sticks can do 1400 in inner slots hence 1400 is the max for me on R5E in quad.

What is more funny is that I have sticks that boot 1500 on MSI but not on ASUS and sticks that boot 1500 on ASUS and not on MSI so binning 8GB Micron is tough.

[Shiranui Gen-An]

2750 is the max I can get out of my D9RGQ sticks (in quad channel at least), which lines up with Hardware.info's tests:
http://us.hardware.info/productinfo/...ab:testresults