Share your AIDA 64 cache and memory benchmark here

[zebra_hun]

Doing 54000mb/sec in 10.5ns!
BE WARNED, MIGHT CAUSE DROOLING! I believe this is a world record! It took me over 7 months and MANY MANY hours of long hard work, leaving no stones unturned to get this. The current world record is 13.5ns but using a 2Gb stick and he has to use millions of volts to get it to posts. As long as it posts to complete the benchmark then it's good. If you play the game on it, then there is no question it will blow up. Mine is rock solid! And passes every test out there. It passes 30 hours of memtest86, prime95 AVX/SSE 24HRS STR8! And last but not least, IMO is the hardest stress test to pass is OCCT AVX LARGE/ SSE MIX. Zero errors for 14 half hours. Playing on it is like to me feels like playing on a CPU that was overclocked to 7ghz cryofreeze to kept it cool that I've tried out at a friend's house in the past. It is simply stunning. It took me many trials and errors. I have post 3 results over a 6months period.

I do not really understand you. If you think of the last picture, I’m not saying it’s stable. On the bonus image, you can also see that I did not attach my stress test result.
For me, that’s enough, if the test runs on it, I’m not going to run the P95 for 24 hours.
I have a normal 24/7 profile at 4133MHz, 1.38V.
That's enough for me to play.
I made a 4400 MHz profile yesterday, yes, stable.
I need such extreme profiles for 3DMark, Superposition, for example.
Well, this is an absolute world record here in the RTX3080 category. 16700 points is the peak if you search for it. (Everithings are OC'ed)

 

[justiceleague]

I do not really understand you. If you think of the last picture, I’m not saying it’s stable. On the bonus image, you can also see that I did not attach my stress test result.
For me, that’s enough, if the test runs on it, I’m not going to run the P95 for 24 hours.
I have a normal 24/7 profile at 4133MHz, 1.38V.
That's enough for me to play.
I made a 4400 MHz profile yesterday, yes, stable.
I need such extreme profiles for 3DMark, Superposition, for example.
Well, this is an absolute world record here in the RTX3080 category. 16700 points is the peak if you search for it. (Everithings are OC'ed)

I have been a Ram Junky/Enthusiast since 2000 and have the luck of owning some super bin rams. There is something about them that drives me and many others nuts. I spend more money on my rams than my CPU. To me, your CPU is ONLY as good or fast as your Ram. I have the honor of tuning in some of the best ram kits in the world. Because I talked about it so much and as you know, not many people really think much about rams. They thought it is not worth going through all the hassle for a measly 5fps improvement. But little as they know, it was never about fps that you wanted super fast rams. It's about response time, the smoothness of how it makes your mouse respond to your computer and so so many other things that can't be measured. Most people thought that the more fps the better and that's about all that matter to most!
I have tuned many kits for friends, family, neighbors, etc., and have NEVER seen a CL 17-17-17-35 at 4600mhz that can run error-free for more than 3hours. However, one came very close to 2hrs 39 mins before it lit up with errors on memtest86. You post that picture for your own gratification of some sort of superiority to people. And to people that just started in this hobby will admire it and want to get that results in one day. Little do they know at this moment, I can say with experience that it is NOT possible! Why would you have that profile knowing that it is not just unstable, it is VERY, EXTREMELY not stable? The only reason is, for benching and I don't blame you because I used to do that. But I would always let them know that it is not stable and I would NEVER run it on my computer for any task but benching.

 

[zebra_hun]

Please believe this, it exists 4600MHz Cl16 error free, I saw. It doesn’t work for me unfortunately, but since it’s not my photo, I won’t share it now. I will be sharing the image soon if I manage to fix the bugs. Control of the TM5 Extreme1 Anta777. 1.5 hour stress test, that's enough.

 

[Mr. Fox]

Considering the platform, this seems like a pretty decent result. I wish AMD could find a way to get FCLK to run 1:1 above 3800.
It's really lame that they can't get that right.

 

[Det0x]

1:2 and default TM5 config at what seems to be XMP settings with GDM enabled on single rank memory ?
Seems excessive ^^

Btw guys, this is anta777 new TestMem5 cfg: Absolut.
(its even harder then normal Anta777 extreme and 1usmus cfg o_0)

He posted it here and here.

For 6-core processors time in main section in config need to increase 1750%.
16 threads -1250% (the config was written)
24 threads - >=830%
32 threads - >=625%
12 threads - >=1670%

Completed my run with twice the runtime: (1250% instead of 625% for 32threads 5950x)

Mistakes was made on the last screen.. (didnt notice r15 was hiding TM5 mainwindow)
Reason to show valid benchmate -> no trickery with windows timers

 

[justiceleague]

Please believe this, it exists 4600MHz Cl16 error free, I saw. It doesn’t work for me unfortunately, but since it’s not my photo, I won’t share it now. I will be sharing the image soon if I manage to fix the bugs. Control of the TM5 Extreme1 Anta777. 1.5 hour stress test, that's enough.

I'm sure there are, but I have not seen it myself is all I'm saying. I apologize if I come up as rude or condescending. This hobby isn't as big as others and we need more people to make it more fun for everyone. I was looking at your lastest Aida bench results and you did a very good job and a very good latency esp at 4400 cl17-18-18. Normally it takes at least 4500mhz to get to 6800mb/s at cl17. There's a simple way to kinda tell if your motherboard is happy or not with your timings or your RLT/IOL is looking at your read/write. I found that your motherboard is most happy when the read/write is perfectly in the middle or 90-98 percent read/write. But always the write speed has to be leading and not trailing behind. If it is, meaning that your RTL/IOL is a bit loose than what she wanted. Of course, she is happiest when it is half and a half though. I post some samples of mine.

 

[Mr. Fox]

1:2 and default TM5 config at what seems to be XMP settings with GDM enabled on single rank memory ?
Seems excessive ^^

Btw guys, this is anta777 new TestMem5 cfg: Absolut.
(its even harder then normal Anta777 extreme and 1usmus cfg o_0)

He posted it here and here.


Completed my run with twice the runtime: (1250% instead of 625% for 32threads 5950x)
View attachment 219938
View attachment 219940

View attachment 219939
Mistakes was made on the last screen.. (didnt notice r15 was hiding TM5 mainwindow)
Reason to show valid benchmate -> no trickery with windows timers

Those are great results and you have a nice silicon sample. My 5950X is a mediocre bin. I've definitely seen worse, but it is pretty meh, and I am disappointed with the card I drew in the silicon lottery. You've got some nice RAM, too. My B-die is not binned as well. I am able to get it to post and load Windows with the same settings, but it has severe errors and ends up having a BSOD if I try to run TM5 or AIDA64 memory benchmark.

What kind of settings are you using to get one core to boosts to 30GHz? That is insane. I suspect that is why your memory read/write/copy speeds are so high, because the RAM clock and timings almost certainly wouldn't produce the same crazy good result.

 

[HD64G]

My seemingly stable results after updating to the latest UEFI. I will try another TRFC setting later if it helps or not.

 

[Ibizadr]

This is my 3200mhz @cl16 I change some timings according to some guides and dram calculator. What you guys think of it

 

[HD64G]

I would try to reach at least 3466 or even 3600 with your 5800X. Use dram calculator for safe settings first and if ultra stable go for fast.

 

[Ibizadr]

I would try to reach at least 3466 or even 3600 with your 5800X. Use dram calculator for safe settings first and if ultra stable go for fast.

I try with 3600 cl18--22-22-48 @1.4v and computer boot but after 5sec on windows I got bsod. So I only change some timings and keep the 3200mhz. The only thing I dislike it's the difference between read and write

 

[HD64G]

I try with 3600 cl18--22-22-48 @1.4v and computer boot but after 5sec on windows I got bsod. So I only change some timings and keep the 3200mhz. The only thing I dislike it's the difference between read and write

This difference is typical of the 6 or 8-core Zen3 CPUs but don't make them weak in anything to bother with. Copy speed and latency are the important numbers. As for the test on 3600, try 3466 with higher SOC voltage. Should be pretty easy and stable and will give better performance in most workloads.

A bit lower voltage for SOC results in a bit better copy speed. Weird me thinks. If that is proven stable I am ok with that.

 

[freeagent]

Just a pair of 3200C14

 

[HD64G]

Just a pair of 3200C14

View attachment 221138

Great numbers there but what is the VDIMM?

 

[freeagent]

Great numbers there but what is the VDIMM?

I think I had it at 1.55, might have been 1.6.. I was trying a couple of different things..

Edit:

Back to my beastie.. I was actually going to sell it but keep getting low balled so I dropped it back in the socket.

 

[97pedro]

Considering it's the cheapest b die kit you can buy, being 3600mhz 17-18-18-38 stock at 1.35v, and also being a Ryzen 3000 platform, think it's preety nice

 

[harm9963]

About the same ,since a year ago.

 

[Miky2147]

Hi, I have a 3950x, 2080Ti, Crosshair VIII Formula, G.Skill Trident Z Neo F4-3600C16D-32GTZN. All STOCK (except the simple automatic DOCP of the ram from bios). I have Windows 11, all updated to latest version (Windows patch and AMD chipset). I noticed that the benchmarks with AIDA64 6.33.5700 of the L3 cache are slower than normal. On the AIDA64 website there is a benchmark of a 3950x better than mine: https://www.aida64.com/news/finalwi...icelake-rx5500-threadripper-3950x-3970x-3990x. Attached there are two of my benchmarks. What can I do? Thank you

 

[dopemoney]

looking at other results, i'd say mine are respectable. good times!

 

[Kyton72]

Hi all.

does this look within spec?

 

[JCL]

I did this with one CCD disabled eliminating the latency between the two. (like a 56/5800)
5180 Mhz (51.5 x 100.6) H20 cooled.
Max offsets, 200 over and 10x

 

[DR4G00N]

4x8GB B-Die @ 3600 C15 2T, 1.40V. Board doesn't OC mems well, any more with 4 dimm's and the speeds drop to half.

Gained around 8000MB/s Read over 3200 C14 XMP so it seems worthwhile.

 

[freeagent]

Windows 11 is so pitiful. Ugh. They sorta fixed it, but now its not. This is an all core clock. I think I will bail on 11 tomorrow.

Booooo.

I didn't bother waiting until tomorrow

First run on a clean install..

 

[tbob22]

Got rid of the Silicon Power because it wouldn't pass memtest86 rowhammer test. Picked up some Hynix DJR 3200/cl16 on sale, at 3600 I needed push the uncore voltage quite a bit higher so 3466 is a good compromise. Haven't done any major tuning but I need stability, this is 7 pass memtest86 stable (about 96h), 20,000% karhu and 24h p95 blend.

 

[JCL]

Windows 11 is so pitiful. Ugh. They sorta fixed it, but now its not. This is an all core clock. I think I will bail on 11 tomorrow.

Booooo.

View attachment 225582

I didn't bother waiting until tomorrow

First run on a clean install..

View attachment 225591

↑↑ nice L3 bandwidth ↑↑