Info 64MB V-Cache on 5XXX Zen3 Average +15% in Games

[Kedas]

Well we know now how they will bridge the long wait to Zen4 on AM5 Q4 2022.
Production start for V-cache is end this year so too early for Zen4 so this is certainly coming to AM4.
+15% Lisa said is "like an entire architectural generation"

 

[Schmide]

It's called a joke. I know exactly what it is.

It was a bad or horribly timed joke. I was making a very serious point.

IMPORTANT - Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO

Ryzen CPUs for AM4 platform rely on external, motherboard sourced telemetry to determine their power consumption. The voltage, current and power telemetry is provided to the processor by the motherboard VRM controller through the AMD SVI2 interface. This information is consumed by the processors...

www.hwinfo.com

In case you need a reminder, many AM4 boards are guilty of feeding incorrect power levels to CPUs. The 5800X3D is the first CPU that has a firmware vcore lock to protect key components. There must be a good reason why AMD won't even let "xtreme" overclockers give extra vcore to the chip. Nobody seems to know yet what , if any interactions these power deviation "cheats" may have with the 5800X3D. In case I have to spell it out for you, IT ISN'T YOUR FAULT YOU DIDN'T DO ANYTHING WRONG.

Power is not voltage. I understand power delivery. I know how voltage regulators work. None of that has anything to do with my circumstances. The chip never got the chance to do any work other than boot to bios three times and windows once.

There are tons of capacitors on a chip to smooth out the power delivery. A voltage spike isn't going to make it to critical components unless it's one hell of a spike. The taichi is a 12 phase board which is total overkill for a 105w processor that is designed to boost well above that.

If I had to conjecture. Most likely reason for death. Chip had a flaw in its packaging. After being cool for months, something separated when thermal expansion was reintroduced to the chip.

 

[igor_kavinski]

If I had to conjecture. Most likely reason for death. Chip had a flaw in its packaging. After being cool for months, something separated when thermal expansion was reintroduced to the chip.

It might be why it took so long for AMD to get this chip out of the gate. Maybe they had a lot of failed chips before they got the process just right.

 

[DAPUNISHER]

It's called a joke. I know exactly what it is.

IMPORTANT - Explaining the AMD Ryzen "Power Reporting Deviation" -metric in HWiNFO

Ryzen CPUs for AM4 platform rely on external, motherboard sourced telemetry to determine their power consumption. The voltage, current and power telemetry is provided to the processor by the motherboard VRM controller through the AMD SVI2 interface. This information is consumed by the processors...

www.hwinfo.com

In case you need a reminder, many AM4 boards are guilty of feeding incorrect power levels to CPUs. The 5800X3D is the first CPU that has a firmware vcore lock to protect key components. There must be a good reason why AMD won't even let "xtreme" overclockers give extra vcore to the chip. Nobody seems to know yet what , if any interactions these power deviation "cheats" may have with the 5800X3D. In case I have to spell it out for you, IT ISN'T YOUR FAULT YOU DIDN'T DO ANYTHING WRONG.

Precisely.

Homie getting saltier than the dead sea.

 

[igor_kavinski]

I can understand Schmide's feelings. He is angry and disappointed that he waited so long for this chip and then when he got it, it was a dud. Now he has to wait more. It's like having the girl of your dreams die in your arms after just a little kissing.

 

[igor_kavinski]

The few games where it shines even at 1440P and 2160P.

 

[igor_kavinski]

Few more. When it can't do anything about the average fps, it ups the 1% low fps.

 

[igor_kavinski]

Two more.

 

[Mopetar]

Honestly this is like slut shaming. I explained every step very clearly and I have been building systems since the 80s. I've killed my share of hardware but in no uncertain way did I expose this chip to anything but the best environment.

Edit: and btw the main reason I water cool is for silence.

The following two ideas are both true and explain why people are posting the way they do:

The failure rate for technology is fairly low and a lot of times it may not actually be cooked.

When technology does fail the most likely times for it to happen are right away out of the box or after many years of use.

It does sound like it's cooked because you've done about everything imaginable to isolate the fault and ensure it isn't something else. Equating it to slut shaming (which itself is a stupid concept) though is going a bit far. Most people here are just trying to be helpful and whatever did fail with your CPU didn't in the typically obvious or spectacular manner that makes it apparent that it may not be a weird board or RAM issue.

 

[Schmide]

I'm not angry at the processor failing. That happens. I can only report on that.

I am a bit miffed with statements so I will clarify

If you were letting the board handle voltage and boost with its default profile, I suggest taking control of it this time around.

This certainly implies that allowing the hardware to make decisions about the processor is somehow negligent. I tested ram on the 5800x since using 4 sticks can be kind of tricky. My choice to let XMP define the voltage is a conservative one. Was I expected to somehow go below that to keep the processor from frying?

Quick poll who under volts their processor or ram before initial testing? Is this a thing? If so please make a complete guide.

I'm betting you still killed it somehow. Water sounds like overkill for 5800X3D.. did you turn PBO on?

This is victim blaming. (fine I wont use slut shaming anymore but the analogy stands) I reported a few posts above this exactly what I did. None of which involved PBO

This is also complete hyperbole. Having the best possible cooling is somehow wrong and implies I'm reckless with my hardware. Just to be clear I have never overclocked my 5800x nor used PBO.

Precisely.

Homie getting saltier than the dead sea.

I think my bit of salt was justified.

If you want to alleviate my grievances, this is probably not the best statement to make

 

[BTRY B 529th FA BN]

Sometimes tough love can teach a person to lighten up a little and that's what's going to eventually happen and you'll be over it.

 

[Schmide]

Sometimes tough love can teach a person to lighten up a little and that's what's going to eventually happen and you'll be over it.

Sometimes empathy can teach a person. Honestly I argue less for myself and more thinking of what if someone with less experience was placed in the same situation.

 

[DAPUNISHER]

You know what, never mind, ain't worth the time.

 

[igor_kavinski]

It is weird that people assumed user fault. I mean, come on guys. This is not some random dude who just joined today with the first post about an issue.

 

[wilds]

I was able to find one other user on overclock.net who has same motherboard I do and they are able to hit 4550 MHz on 1-2 cores; while I still can't get above 4450 MHz. Pretty perplexing issue. Did message them to hopefully get some details.

 

[igor_kavinski]

Did message them to hopefully get some details.

Ask them about their heatsink choice and how they applied thermal paste etc. Could be their cooling is better and CPU is clocking higher coz of that. Also, do they have four RAM sticks? The IMC could be limiting your clocks due to the load of four sticks.

 

[DrMrLordX]

Power is not voltage.

It's either current or voltage, take your pick. Your system can't accurately monitor either one if the motherboard firmware deliberately misreports that data. It's also pretty well-known from The_Stilt that some "safe" voltages at low current (like 1.5v for Zen2 and Zen3 cores) can become unsafe at higher current levels, which is why standard Zen3 can have high transients @ idle without damaging the CPU.

Again, what implications all this has for Zen3D is anyone's guess. If yours is just a one-off and your next CPU works just fine in the same board, we can chalk it up to being just a bad part, but it would not surprise me at all if some boards haven't gotten rid of "review day" firmware hacks that were safe-ish for a 5950X but fatal to a 5800X3D.

 

[BTRY B 529th FA BN]

Sometimes empathy can teach a person. Honestly I argue less for myself and more thinking of what if someone with less experience was placed in the same situation.

And then there's the time when you see that it's not helping and judgement comes into play

 

[Schmide]

And then there's the time when you see that it's not helping and judgement comes into play

What is it?

 

[wilds]

Ask them about their heatsink choice and how they applied thermal paste etc. Could be their cooling is better and CPU is clocking higher coz of that. Also, do they have four RAM sticks? The IMC could be limiting your clocks due to the load of four sticks.

I will ask about how they did thermal paste. I couldn’t achieve 4550 with 2 sticks of RAM before upgrading to 4 sticks. Was on 2 sticks for about a week. Tried without XMP at just 2133 as well.

For thermal paste, I did a slightly bigger pea sized drop than I usually do. 4550 MHz should only be achievable under lightly threaded tasks I think. Once more than 3 cores are utilized, 4450 should be its max clock speed.

Because of that, I have a feeling thermals are not the issue; but I could be wrong.

I did look into HWiNFO’s Power Reporting Deviation since my motherboard is not accurate OOB. But once I lowered my PPT, TDC, and EDC to some 5800X users, I saw it hovering around 99% under heavy load. I put values back to Auto for time being.

One thing I haven’t figured out in HWiNFO is its Frequency Limit - Global value is 4450 MHz. Not sure if important.
Here is most of hwinfo expanded 50 ms interval, average system usage for a few hours and running Fallout 4 (sorry for big screenshot)

 

[igor_kavinski]

4550 MHz should only be achievable under lightly threaded tasks I think. Once more than 3 cores are utilized, 4450 should be its max clock speed.

Because of that, I have a feeling thermals are not the issue; but I could be wrong.

You could try using Winrar benchmark with/without multithreading to see how fast a single core goes vs. all cores. How much is the max temp when Winrar is working on a single thread?

 

[wilds]

You could try using Winrar benchmark with/without multithreading to see how fast a single core goes vs. all cores. How much is the max temp when Winrar is working on a single thread?

I've tried a few benchmarks limiting them to one core. I make sure no background software is running except for monitoring tools. 7zip has a built in benchmark where I can set it to 1 thread. It will bounce around the Star and Silver thread, but never above 4449 MHz according to Ryzen Master.

 

[wilds]

On a whim I just tried downloading one of the Ryzen Custom Power Plans hosted on overclock.net:

Ryzen Custom Power Plans for Windows 10/11 (Snappy...

CPUDoc now features a custom dynamic power plan with ultra low power in standby: https://github.com/mann1x/CPUDoc/releases/latest These are the custom power plans I've made for my 5950x. I have tested them as well on a 3800X and 5600G (not very thoroughly). They should not interfere with PBO...

www.overclock.net

Downloaded the Balanced LowPower v8 plan (was using Balanced vast majority of time before this).



And within a couple minutes, HWiNFO is showing 4550 Mhz hit on a core. Very interesting!


Seeing lower idle minimum clocks to boot. There may be a better power plan in that thread I linked, but I'm happy I made some progress!

 

[DrMrLordX]

I did look into HWiNFO’s Power Reporting Deviation since my motherboard is not accurate OOB. But once I lowered my PPT, TDC, and EDC to some 5800X users, I saw it hovering around 99% under heavy load. I put values back to Auto for time being.

You can only accurately read Power Reporting Deviation at stock settings. Not sure that it's going to be a big issue for the 5800X3D. It can't hurt to start asking questions and see if anyone has more info.

As an aside, if you know from taking readings at stock that your board is feeding more power than reported, it's certainly appropriate to lower PPT regardless of which AM4 CPU you use.

 

[dr1337]

Sometimes empathy can teach a person. Honestly I argue less for myself and more thinking of what if someone with less experience was placed in the same situation.

I think a big part of empathy is context and this thread is so off topic already sorry, but your original likening of slut shaming to computer overclocking was quite humorous to me. Im not sure jp was being nice to you but one negative opinion in a sea of people helping you isn't as much of a moral attack as you're making it out to be, in my opinion.

 

[Schmide]

Beyond all this fiddle faddle.

AMD requires a picture of the CPU in socket to get RMA. Glad I took one randomly before. It would suck to disassemble and put it in the motherboard for a sad vanity shot.

So take pictures of your CPUs when commissioning them.