Ram read and write speed is slow.

[Zana]

Hello,
So i have been having performance issues for a while now with my pc and have done a lot of things and got it to run pretty decently now but still not how its supposed to. recently i have overclocked my cpu in the bios through PBO and set max memory frequency to 200 also pbo scalar is set to 3x and xfr enhancements or something like that, in that option the PBO there is also enabled with the scalar set to 3x. So i recently did a benchmark on my memory with OCCT and it seems like thats the part that is underperforming. The ram is in the correct slots number 2 and 4 in dual channel, xmp profile is enabled. i have tries resetting the bios to its default settings but that didnt help. Looked it up everywhere but i cant seem to find a solution. can somebody help me please? it would be very much appreciated thank you.

Specs:
B550m aorus elite
Ryzen 5 5600x - Coolermaster hyper 212 aircooler.
Rtx 3060 gigabyte gaming oc 12gb
32gb 3200mhz corsair rgb vengeance.
Coolermaster 750watt gold 80plus psu.

 

[Harlow]

i have overclocked my cpu in the bios through PBO

First of all, why bother with PBO? What do you hope to gain? A few more FPS in a game perhaps?

I left my 5600X, 3800X and 7950X at stock (no PBO) but I have applied XMP memory overclock where applicable (my 2x32GB DDR5-4800 DIMMs in my 7950X rig do not have any faster XMP settings). I do run significant overclocks on older AMD and Intel CPUs.

Unfortunately I'm not familiar with OCCT Personal, so I've no idea what the numbers 889.48 (Read) and 617.97 (Write) mean. Are they abstract benchmark figures or do they have specific values, e.g. MHz, MT/s (Mega Transfers per second), memory bandwidth (Gigabits per second), etc?

I'm assuming you're running XMP settings in your DDR4-3200 Corsair Vengeance? If so, what speed?

If, for example, you're running XMP 3600, that means the true memory clock is 1,800MHz and the DDR Transfer Rate is 3,600MT/s.

It would be helpful if you posted screen shots from CPUID CPU-Z, or CPUID HWMonitor, or Aida64, rather than OCCT Personal.
I run memory benchmarks in Aida64 to give similar comparisons with other machines, as shown in your screen shots.

Chances are your memory is running just fine and OCCT Personal is not providing clear, unambiguous information.

 

[freeagent]

First of all, why bother with PBO? What do you hope to gain?

Its a free bump in performance if you can figure out how to use it properly.

I left my 5600X, 3800X and 7950X at stock (no PBO)

That is YOU, not everyone else.

i have overclocked my cpu in the bios through PBO and set max memory frequency to 200 also pbo scalar is set to 3x and xfr enhancements or something like that, in that option the PBO there is also enabled with the scalar set to 3x.

You do not need to touch any of that stuff. Just your power limits, your curve, and boost override. And if you are using Ryzen Master, uninstall it, clear your cmos and start fresh.

 

[Dr. Dro]

PBO scalar = auto
Curve optimizer = all core as far as it'll go without crashing, optionally you can reduce your prime cores (marked as a star in Ryzen Master) a bit further, also again until it crashes
RAM timings: Low bandwidth can be caused by incorrect timings, do not use XMP or DOCP; download AIDA64 (trial version works) and check the XMP timing table.

In your case, look for the timing tables regarding 1600 and 1500 MHz, they are a good start. Set DRAM voltage to no higher than 1.45 V

Afterwards test with memtest86 (free works, alternatively use FOSS memtest86+, PassMark's works better on newer machines though)

MemTest86 - Download now!

Download your free copy of MemTest86 here for USB.

www.memtest86.com

OCCT (free version works) for 1hr

Maybe run Realbench

All goes well - your rig is probably stable for the long haul

 

[Harlow]

That is YOU, not everyone else.

When you say "everyone else", does that mean I'm the only (idiot) in the whole wide world who isn't using PBO. Surely there must be at least one other person?

Its a free bump in performance if you can figure out how to use it properly.

I looked at the graph below which shows zero improvement in Premiere Pro on my 7950X. Better still, Lightroom Classic is 1% to 2% faster with PBO switched off! Similarly there's only a 1% variation in Photoshop. For my specific applications, PBO/CPB is pointless and a waste of power at $0.32 per kWhr, but I would certainly enable PBO/CPB if I was running Unreal Engine.

https://www.pugetsystems.com/labs/a...ermals-and-Content-Creation-Performance-2373/

Is Precision Boost Overdrive and Core Performance Boost Worth it for Content Creation?​

In most of the workloads we tested, it is very clear that using Precision Boost Overdrive (PBO) and Core Performance Boost (CPB) is not worth it. Photoshop, Lightroom Classic, Premiere Pro, After Effects, and DaVinci Resolve all showed no difference in performance when we disabled these settings from their "Auto" default, yet the CPU temperature dropped as much as 30C, or even 40C!

Note the statement that CPU temperature dropped by 30 to 40°C for these apps with PBO/CPB switched off. All that wasted heat for no effect!

I'm sure there are apps where PBO provides worthwhile gains, but not "everyone" benefits.

 

[freeagent]

When you say "everyone else", does that mean I'm the only (idiot) in the whole wide world who isn't using PBO. Surely there must be at least one other person?

Lol no that is not what I meant. Most enthusiasts have PBO enabled. As far as reviews go, just because they enabled it, does not mean they know how to used it.

There are gains to be had across various workloads.

But.. I am still on Zen 3, I need to move up to show you guys how its done

 

[Dr. Dro]

PBO can be left on safely, it compliments the curve optimizer. Just don't get delusions of "set scalar to 10X and -30 on all cores"; keep expectations in check. Next to no (and i'm willing to bet no) chips do that - claims on forums about it tend to be exaggerated and/or embellished to help make some individuals feel good about their purchases.

Simply treat it as an "overclocked" system - make sure you have cooling and go to town on it

 

[Sarajiel]

PBO can be left on safely, it compliments the curve optimizer. Just don't get delusions of "set scalar to 10X and -30 on all cores"; keep expectations in check. Next to no (and i'm willing to bet no) chips do that - claims on forums about it tend to be exaggerated and/or embellished to help make some individuals feel good about their purchases.

Curve optimizer = all core as far as it'll go without crashing, optionally you can reduce your prime cores (marked as a star in Ryzen Master) a bit further, also again until it crashes

Couple of nitpicks here, PBO is the prerequisite to actually enable curve optimizer and all the other fancy stuff like the scalar. From my experience with potato Zen3 CPUs, the preferred cores typically have less headroom on the curve than the rest. I assume most folks never try to experiment with it enough to notice, or their CPUs are so good that they hit other limits first.
The effective values you can achieve with curve optimizer without running an unstable system also depend on your memory overclock. If you run "balls to the wall" memory timings, you won't be able to run those mythic co -30 settings anyway, because your CPU isn't actually waiting long enough for its data anymore, to put it very simplistically.

For example, I'm happy, if I can use a -10 all-core stable on most of my Zen4s with tightened Hynix timings. I can go higher for benching when stability only requires finishing the benchmark, but I doubt that the OP is looking for that sort of thing.

So i recently did a benchmark on my memory with OCCT and it seems like thats the part that is underperforming.

Run a benchmark that is useful to you and that actually measures something computationally and not just some willy-nilly memory benchmark, especially one that doesn't even tell you what it is actually measuring. Measuring data transfer speeds and latency may be useful to gauge the theoretical peak performance of a memory sub-system, but it's fairly useless to get an idea how a modern system actually performs. Also avoid AIDA on Ryzen, because it tends to spit out physical impossible values since it usually can't get around hitting the caches. The authors basically reverse engineer their benchmark for each CPU architecture to measure something that modern CPUs try to hide with their complex cache hierarchy.

If you are out of ideas what to use as a benchmark, you can try collections like BenchMate that are used by extreme overclockers to measure the insides of their pants, to break overclocking world records.