• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Share your AIDA 64 cache and memory benchmark here

3200@3600 cl14
cachemem3600gdmoff.png
 
i5-6500T with DDR3

126094
 
Old xeon cache results. Not bad considering the age.

126106
 
Old dual xeon 6 channel 6x16Gb DDR3 1066 ECC REG CL7
127014
 
Ryzen 7 3800X, Patriot Viper 3600 RAM 17-19-19

1563276595226-png.126969
 
Last edited:
07182019-132140.jpg
 
127593


Bonus:
127594


127595


127596


127597
 
Some classic silicon..

Capture.JPG
 
128560
 
Ryzen 5 3600 (Stock), Memory HyperX FURY OC 3333MHz 18-20-20-42 (Original 2666MHz 16-18-18-39)
128618
 
My 3700x w/ G.skill tridents @3600 14-16-15-30
RAM.PNG
 
129414
 
phanbuey said:
what the hell is that stable?

Nice!
Click to expand...

adata d41 spectrix 3600 cl7 kit @3733 cl14 1.5v and 4000 cl16 1.45v but can't pass 1866 FCLK so worthless.
 

Attachments

  • 4000cl16-435-1005.png
    4000cl16-435-1005.png
    91.8 KB · Views: 1,209
I got myself a kit of 3200C14 B-Die Trident Z and started fiddling around with the DRAM Calculator.

System:
MSI B450M Mortar
Ryzen R5 3600 @ stock (no PBO)
Trident Z RGB 3200C14 2x8GB
Corsair RM650X

3200 14-14-14-28-42-1T-GDD @ 1.37:
130346

130347

A bit high on the Latency and not very impressive Writing bandwidth

3600 14-15-14-28-42-1T-GDE @ 1.42
130350

130360

Quite a lot better!

But now it got funky. I discovered a difference in performance when entering timings directly into bios or through Ryzen Master. Above is Ryzen Master. Below is BIOS
130361


Exact same timing and settings! That's a hefty drop! The only way to restore performance is setting timings on auto in bios again and use Ryzen Master to override them again.


Then I gave it another shot at 3733 at CL14 which didn't boot and at CL16 it gave me worse performance than 3600CL14. I almost gave up until I decided to give the DRAM calculator another shop at 3800MHz. Here the settings and results:

3800 16-17-16-32-50-1T-GDE @ 1.42v
130362

130363


And that's good enough for me. I'm currently testing if it's stable for 24/7 and so far it devoured everything I threw at it. Prime and Memtest and Gaming. 1.42 is a really nice low voltage and not much heat is coming from it but when the GPU is heating up the are touching 50 degrees. I'm going to switch around the layout of the cooling a bit or get a Dominator Airflow cooler on top of it to make sure they stay cool. I'm working in an M-ATX case and AIO cooler not much airflow options.

I think with some better cooling and a bit more voltage this memory is not out of breath yet and I could tighten timings a bit here and there but for now I'm really satisfied with the performance increase.

I failed to make an Aida64 bench screenshot at XMP baseline but it was lower than my first screenshot here for the ease of things lets compare the first screenshot with the last:

Read: 48100 MB/s -> 55500 MB/s = +7400 MB/s
Write: 25600 MB/s -> 30400 MB/s = +4800 MB/s
Copy: 46200 MB/s -> 53100 MB/s = + 6900 MB/s
Latency: 74.2 -> 65.8 = - 8.4 ns

How this translates to real world performance depends on the application it doesn't help with GPU bound resolution gaming but might help with some other things.
 

Attachments

  • 3600-14-15-14-28-42-1T_RM.jpg
    3600-14-15-14-28-42-1T_RM.jpg
    317.1 KB · Views: 648
130446
130447


Yeah, similar issues here. Every calc I've tried just bootlooped me insensible. Only way I could get it to even POST anywhere near that mythical 1900MHz was to enable XMP, set MEM SPD to 3800 manually and let everything else run Auto. She'll crank the CPU right up there and throughput is lower but still in the same ballpark; however latency crunks out a good 10ns slower.

I'll admit I really don't know what I'm doing in there anymore, though; I haven't done this shit for close to a decade. :-/

I'm waiting for some more real-world info on my build; I'm sure somebody with more knowledge can get these numbers down. I had it down to 64ns/54Ksomething READ ONE TIME and it crashed when I saved, and refused to POST again after that.

I know this RAM is capable of more; I found out AFTER I bought on a "Deal of the day" that it is record-setting gear: https://www.crucial.com/usa/en/ballistix-elite-world-record
 
cachemem.png
 
Really can't knock this $89 (equivalent cost) memory kit from Patriot. I wasn't overly impressed when I started, but as more mature UEFI updates have come up, the memory has coped with higher clocks and tighter timings. The only one I can't change is tRCDRD, as then the system won't boot for some reason.
It runs on lower Voltages and at slightly tighter timings that the DRAM calculator suggests at these speeds.

1567234305202-png.130478


1567234448633-png.130481
 
130524
130525


130526
130527


Inspired by the successes of Cidious and Lost Swede seen above, I decided to take another stab at it with Ryzen Master. I must say, I'm definitely much pleased; managed to break over 54K and under 66nS with CPU OC at full Auto.

Would love to see something like Swede's numbers, but can't seem to shake the "No POSTies" if i drop ANYTHING any lower at either 3667 or 3733. 3800 is just totally out to lunch with latencies in the high 70s as seen in my last post. :facepalm:

I'm currently beating on the machine at 3667; will compare both to see which is better in gaming vs content creation vs crash-resistance.

Cheers,

mnem
Don't read this.
 
Latency to 34.0

130580
 
130596
 
7980xe + terrible quality RAM from corsair = mess.
130609


So XMP2 decided to be stable, well I won't complain - reran five times and no BSODs (unlike my z270 system or my carbon)
130610
 
You must log in or register to reply here.
Back
Top