Question Zen 4 builders thread

[Markfw]

I know its early, but as someone pointed out, I better get my ram now before the scalpers are all over it.

To go with a 7950x, what do you think of this ram:

https://www.newegg.com/g-skill-32gb-288-pin-ddr5-sdram/p/N82E16820374369?Item=N82E16820374369

(edited for my best recommendation, 12/5/2023)

 

[Fjodor2001]

I don't have the immediate need to use my PC as I do my phone. I'd rather have the piece of mind knowing the memory training at boot is going to make whatever PC gaming session I have problem free. I can wait an extra 20 to 30 seconds.

Some nublets equate it to a performance benchmark. Like it's some broader gauge of overall performance. As silly as that is. It's a party trick to show to your friends.

What about your car, TV, etc? Are you ok with that taking 60 seconds to start? Do you also just consider it a "silly party trick to show your friends" that your car and TV starts quicker than that?

Also, you did not answer the other two questions:
"And why not have it quicker if possible? Why do you want the startup time to take longer than necessary?"

Apparently other PCs are able to boot to Windows logon in 10-15 seconds while still working perfectly fine, so it's definitely doable.

 

[A///]

Further down:

" Arguably if the memory controller had a better memory training procedure it wouldn't have to retrain all the time as it would come to a set of settings that always work well enough."

So is the updated AGESA 1.0.0.7B just that?

I'm just wondering how much more stable a computer will be without context restore? I've used it since I've gotten it, and everything seems to run fine. I use Builzoids timings, so tight but not super tight timings. What are the MTBF difference between the two settings?

A better procedures fixes nuances between boots how now?

 

[biostud]

I don't have the immediate need to use my PC as I do my phone. I'd rather have the piece of mind knowing the memory training at boot is going to make whatever PC gaming session I have problem free. I can wait an extra 20 to 30 seconds.

Some nublets equate it to a performance benchmark. Like it's some broader gauge of overall performance. As silly as that is. It's a party trick to show to your friends.

The good thing is we can run our computers exactly as we want.

Personally I often has shorter sessions on my computer, and turn it off regularly, so the extra boot time is annoying.

If you usually have long sessions or just put your computer in sleep, then I can understand why it doesn't matter to you.

It would still be interesting to have some kind of comparison of stability between the two settings.

 

[biostud]

A better procedures fixes nuances between boots how now?

I'm not an electrical engineer, just quoting Buildzoid as you do.

My point being, if I don't experience any crashes with memory context enabled, even if it is 100x less stable than disabled then, I still have a perfectly stable system.

 

[A///]

I'm not an electrical engineer, just quoting Buildzoid as you do.

My point being, if I don't experience any crashes with memory context enabled, even if it is 100x less stable than disabled then, I still have a perfectly stable system.

Stability in the interim or stability over time?

 

[AdamK47]

What about your car, TV, etc? Are you ok with that taking 60 seconds to start? Do you also just consider it a "silly party trick to show your friends" that your car and TV starts quicker than that?

Also, you did not answer the other two questions:
"And why not have it quicker if possible? Why do you want the startup time to take longer than necessary?"

Apparently other PCs are able to boot to Windows logon in 10-15 seconds while still working perfectly fine, so it's definitely doable.

Different devices for different applications. You're comparing apples, grapes, and potatoes.

You're more than welcome to continue to use MCR. I personally find it an unnecessary feature brought about by those who incorrectly equate fast boot times as a gauge for a fast reliable PC. It's a fallacy. It has little appreciable real world benefit with a greater downside for long term stability.

 

[igor_kavinski]

It has little appreciable real world benefit with a greater downside for long term stability.

Regardless, our very own Biostud has volunteered to test the long term stability of MCR. Isn't that right, Biostud?

 

[A///]

Different devices for different applications. You're comparing apples, grapes, and potatoes.

You're more than welcome to continue to use MCR. I personally find it an unnecessary feature brought about by those who incorrectly equate fast boot times as a gauge for a fast reliable PC. It's a fallacy. It has little appreciable real world benefit with a greater downside for long term stability.

Adding to this and going forth with what Builderzoid said on the reddit is that much constant changing variables requiring a retrain a car's ecu for example will adjust ignition timing at a constant rate to get the best fuel to air ratio to achieve peak power and clean burn. Obviously the booting/starting and warming up have little to do with a pc but much like a more complex device (a desktop computers performing more calculations per second than a car can) there's constant adjustments.

 

[biostud]

Regardless, our very own Biostud has volunteered to test the long term stability of MCR. Isn't that right, Biostud?

n=1 should give us solid statistical data

 

[biostud]

Stability in the interim or stability over time?

My English skills are coming short, into understanding your question.

If for some reason my computer should show signs of instability, I can always disable context restore. But on the other hand reboots are really quick with CR enabled

Also every time there's a bios update, I will retrain my memory.

 

[A///]

My English skills are coming short, into understanding your question.

If for some reason my computer should show signs of instability, I can always disable context restore. But on the other hand reboots are really quick with CR enabled

Also every time there's a bios update, I will retrain my memory.

Stability in the interim is alright but stability X time may not be. What may seem stable to you in your eyes may not be truly stable but not unstable enough to cause a fault in regular use without pushing the hardware through external tools available in the community.

In the same vain as a "stable but not really stable" overclock that affects memory handling during read and write to disk a drifting memory setup may introduce errors over time. These would either be noticeable or not to you depending on the type of work you do or introduce issues that couldn't be traced back to the lack of retraining.

This is why I asked @AdamK47 if this was a new normal for higher power DDR generations or a DDR5 issue that will eventually be solved.

 

[AdamK47]

Regardless, our very own Biostud has volunteered to test the long term stability of MCR. Isn't that right, Biostud?

For JDEC speeds and timings I can see MCR working very well long term. Anything beyond that is a gamble.

 

[A///]

For JDEC speeds and timings I can see MCR working very well long term. Anything beyond that is a gamble.

Isn't MCR a jump in speeds that doesn't pay off until the end of DDR5's lifetime?

 

[eek2121]

Excellent. That means the chip is hitting the max frequency of 5.85 GHz without getting to TJmax of 95C. Very nice!

@Markfw should look into it.

It doesn't ever hit even 5.8. The chip hits 5.75. the temps at that workload are around 76-80C. The only time it actually hits 90+ is in a doomsday multi-core scenario.

I've actually never seen this chip have any type of sustained activity @ 5.75ghz (that is, "average effective clocks" or "effective clocks" min/max/average)

Core clocks will briefly touch 5.75ghz, but again, with the chip sleeping so much, it isn't "really" doing 5.75+ghz. One of the few complaints about Zen 4 (counter: the 7950x makes my 5950x look like a snail)

Effective clocks show most SC workloads at around 5.2-5.5, MC workloads were similar, but after prolonged use and BIOS updates, all core workloads are around 5.0-5.3ghz. Ignore effective clocks and the chip claims to hit 5.5ghz on all cores at least once a day.

AMD should really change how it markets this stuff. Overpromise and underdeliver does nothing to help perception. One of the reasons Zen 3 was so popular was you got more than what you purchased. My 5950x clocked faster than advertised! Rejoice! (never mind the fact every chip did that)

I can't really complain about the 7950X though. It beats my 144W 5950X with just 88W of power. Like...the competition isn't even close, and the 5950x was a FANTASTIC chip. When senior shareholders had nightmares prior to the Raptor Lake launch, it was of the 5950x taking over the world. 🤣

Hmm from what you've said my guess is that the template forces the heatsink to lay even more flush to the board and processor ihs through probably deflexing the board a little thanks to the template providing a more rigid area.

I was initially thinking that, however, while I don't watch LTT videos every day, I do keep an eye on his stuff. I found this little interesting video:

-- I suspect that since the CCDs are closer to the edge, the contact frame actually provides additional cooling area for the hotspots, which are right at the edge of the heat spreader. Why AMD thought that doing this was smart...

 

[A///]

-- I suspect that since the CCDs are closer to the edge, the contact frame actually provides additional cooling area for the hotspots, which are right at the edge of the heat spreader. Why AMD thought that doing this was smart...

Didn't Zen 2 and Zen 3 show problems with the CCD's too close to each other? Perhaps that is why. I don't know for sure. I think AMD's biggest mistake was trying to maintain cooler design uniformity with their z height. der bauer made some videos of delidded zen 4 and others with shaved down ihs videos but I think amd should have let the uniformity gone to the trash. It will remain an issue until zen 5 3d or zen 6 provided it lives in the am5 lifetime.

 

[Hotrod2go]

After training my memory I've disabled memory training and boot times are around 15 secs

How did you disable it? is it under nitro training options?

Just wonder what Windows boot-up times you're getting with your AM5 Zen4 PC builds (and with what MB, CPU, and RAM)?

I found this review testing some different motherboards:

Screenshot from video review:
View attachment 85188

As can be seen the results vary from from 23 to 53 seconds! I would have expected more like 5-15 seconds.

Is this a common problem with AM5? And why does it differ so much between different motherboard manufacturers?

Can we expect future BIOS updates to bring down the AM5 boot-up time to around 5-15 seconds which I'm used to from other PC builds?

Once I've settled on a satisfactory OC with my B650 board, I'll disable nitro training options & put on auto then leave MCR & PDM on auto, my system boots up to the windows login screen in approx 10-12 seconds. Of course having a speedy PCIe4 NVME drive helps too, but the system just goes with the bios profile I've created in my OC when the first time training is done. The memory training thing can be an issue when you don't know or can't confirm what the system is doing when first booting with a given hardware profile. If my OC attempt is too aggressive or not properly configured with my particular hardware combo, the system will just hang & not boot. You can hear the case fans screaming away but nothing appears on the screen. Clear CMOS time! & re configure.

 

[biostud]

For JDEC speeds and timings I can see MCR working very well long term. Anything beyond that is a gamble.

I don't really understand what happens at the training, but shouldn't the result theoretically be pretty close every time as long as the hardware and settings are the same?

Is there a difference between running JEDEC, Buildzoids "safe" settings and at the "edge" settings?

So if you run at the "edge" settings it makes perfectly good sense to disable CR, but if you run more relaxed timings then you can use CR without risking system stability.

It would be interesting to see what values where saved after each memory training, to explore how much difference there is.

 

[JoeRambo]

Is there a difference between running JEDEC, Buildzoids "safe" settings and at the "edge" settings?

The difference is in margins that are left for stable operation. What is saved is "inputs" that configure mem I/O, like termination resistances, clock skews, variuos delay register values etc ( some visible in the BIOS and some are not ).
So CPU no longer needs to find those values on each boot ( itself a complicated process, with plenty of dark art, MB native values and cross param impacts ).
What it still needs to do is to "train and lock" the various clock domains to have signal that is acceptable and does not produce memory errors.

So obviously the above process is about margins - an artificial example of dropping voltages somewhere could result in signal no longer being acceptable. Same can happen for example if your original training was done at room temp and now you are cold rebooting after 2 days of video encode and you end up on edge that destabilizes the system maybe immediately or maybe it will become unstable real soon after CPU IOD and memory cools down.

So for "edge" settings, training each time provides additional warning about instability. For all others, not pushing anything anywhere, context restore will work just fine.

 

[Fjodor2001]

Different devices for different applications. You're comparing apples, grapes, and potatoes.

You're more than welcome to continue to use MCR. I personally find it an unnecessary feature brought about by those who incorrectly equate fast boot times as a gauge for a fast reliable PC. It's a fallacy. It has little appreciable real world benefit with a greater downside for long term stability.

Please explain how a car or a TV is different from a PC with regards to startup time. They are both using computer based systems (yes cars have infotainment system etc too!). You turn them on. You want them to be working as quickly as possible so you can start using them. And then you use them for quite some time. Very similar. Why are you ok with sitting idle for 60 seconds and waiting until you can use your PC, but not prepared to wait the same time before you start using your TV or car?

Also, I might add that I've worked with several different consumer electronic products, and one of the top requirements is always to get the startup time as quick as possible. Often it's being chased in 100 millisecond steps. Something like 60 seconds which you are ok with is totally unacceptable. It's DOA. Consumers don't want it. You won't get the contract.

That said, everyone is different. And if you're ok with 60 second startup times then good for you. But to me it just seems weird to accept that, when there are competing PCs that are able to boot up in 10-15 seconds. I still don't see why you'd rather have a 60 second boot-up time to no benefit. Maybe it's some party trick to show that your PC can survive a 60 second boot-up.

 

[In2Photos]

Please explain how a car or a TV is different from a PC with regards to startup time. They are both using computer based systems (yes cars have infotainment system etc too!). You turn them on. You want them to be working as quickly as possible so you can start using them. And then you use them for quite some time. Very similar. Why are you ok with sitting idle for 60 seconds and waiting until you can use your PC, but not prepared to wait the same time before you start using your TV or car?

Also, I might add that I've worked with several different consumer electronic products, and one of the top requirements is always to get the startup time as quick as possible. Often it's being chased in 100 millisecond steps. Something like 60 seconds which you are ok with is totally unacceptable. It's DOA. Consumers don't want it. You won't get the contract.

That said, everyone is different. And if you're ok with 60 second startup times then good for you. But to me it just seems weird to accept that, when there are competing PCs that are able to boot up in 10-15 seconds. I still don't see why you'd rather have a 60 second boot-up time to no benefit. Maybe it's some party trick to show that your PC can survive a 60 second boot-up.

Smart phones and smart TVs take time to boot. They are only "instant on" devices once they have booted. How long does your phone take to fully start from being powered on? Your TV? How long does each take to reboot? A car has many dedicated computers, not a single one controlling everything. Your PC will spend more time applying Windows updates than it will booting up everyday.

 

[Fjodor2001]

Smart phones and smart TVs take time to boot. They are only "instant on" devices once they have booted. How long does your phone take to fully start from being powered on? Your TV? How long does each take to reboot? A car has many dedicated computers, not a single one controlling everything. Your PC will spend more time applying Windows updates than it will booting up everyday.

My TV does not take 60 seconds to boot, no! Not my car either. Counting cold boot. More like <10 seconds.

And no, my Windows PC does not apply updates everyday. What kinda weird stuff is MS pushing on your PC if you get Windows updates every day?

 

[In2Photos]

My TV does not take 60 seconds to boot, no! Not my car either. Counting cold boot. More like <10 seconds.

And no, my Windows PC does not apply updates everyday. What kinda weird stuff is MS pushing on your PC if you get Windows updates every day?

How about your phone? Noticed you left that response out.

I didn't say that Windows updates my PC daily. But a weekly Windows update takes longer to apply than spending 60 seconds booting your PC a few days a week.

 

[AdamK47]

Please explain how a car or a TV is different from a PC with regards to startup time. They are both using computer based systems (yes cars have infotainment system etc too!). You turn them on. You want them to be working as quickly as possible so you can start using them. And then you use them for quite some time. Very similar. Why are you ok with sitting idle for 60 seconds and waiting until you can use your PC, but not prepared to wait the same time before you start using your TV or car?

Also, I might add that I've worked with several different consumer electronic products, and one of the top requirements is always to get the startup time as quick as possible. Often it's being chased in 100 millisecond steps. Something like 60 seconds which you are ok with is totally unacceptable. It's DOA. Consumers don't want it. You won't get the contract.

That said, everyone is different. And if you're ok with 60 second startup times then good for you. But to me it just seems weird to accept that, when there are competing PCs that are able to boot up in 10-15 seconds. I still don't see why you'd rather have a 60 second boot-up time to no benefit. Maybe it's some party trick to show that your PC can survive a 60 second boot-up.

Someone is taking my comments personally.

Booting 10 seconds. 20 seconds. 30 seconds. It has little to no importance over stability and performance.

It is a silly party trick to impress non computer enthusiast friends who are less versed in what true PC performance entails.

 

[Markfw]

Someone is taking my comments personally.

Booting 10 seconds. 20 seconds. 30 seconds. It has little to no importance over stability and performance.

It is a silly party trick to impress non computer enthusiast friends who are less versed in what true PC performance entails.

Exactly... Now 5 minutes or more like my Genoa boxes takes get annoying, but since they are virtually never rebooted and run all the time who cares. As for my 7950x's when I do reboot, or power up, I hit the power button, then go the fridge for a drink, and when I get back its on, so who cares ? As Adam says, anything else is a party trick for non-computer people.

 

[biostud]

Exactly... Now 5 minutes or more like my Genoa boxes takes get annoying, but since they are virtually never rebooted and run all the time who cares. As for my 7950x's when I do reboot, or power up, I hit the power button, then go the fridge for a drink, and when I get back its on, so who cares ? As Adam says, anything else is a party trick for non-computer people.

Example:

So when you turn your computer on a cold Monday morning, it does it's training and we then reach a hot Thursday after your computer has been crunching for a couple of days, then why is the training which is done a cold morning more stable than a restored training from a hot afternoon?