AMD Announces BIOS Fix for Ryzen 3000 Boost Clocks - hits September 10th

Status
Not open for further replies.
LOL! Thank you man, I do appreciate it, and I do try for our end-users. :D
I think I'm doing bad if I don't get a laptop back to a user within 48 hours with a full load and a full custom rebuild needed - month and a half, damn, you deserve better!!!


Most systems it is done in the background, but there are quite a few where the task running can impact work performance and have an impact on the end-users, hence why we want those single-thread boosts to be running at their advertised rates, regardless of whether it is AMD, Intel, etc. :)

May I ask how many spywares are pushed onto the end users from your organisation too ? W10 vanilla is ~4GB while our custom image (which support ~30 models) barely fits onto a 64GB key lol.
So many processes ran to monitor user actions.... may want to cut on that first lol.

That might not be your case too, that's from experience.
 
May I ask how many spywares are pushed onto the end users from your organisation too ? That might not be your case too, that's from experience.
That's not the case, and no spyware outside of Windows 10 itself... :D
 
Last edited:
Yet you obviously don't have enough hands on experience to know that there are still applications in and out of enterprise/personal use and OS agnostic, that are still single-threaded.

I never said there weren't single threaded tools. (strawman) In fact I have enough experience to know that Audacity is actually mutli-threaded. Albeit a UI thread and a work thread. I actually worked on it and multi-threaded their equalizer code. It still sits in a dormant state as certain other priorities have taken precedence. They're cool guys and I do understand their least common denominator approach.

Your example was predicated on a decryption case. Yes hashes must be computed in order but that doesn't prevent pipelining operations. Moreover take a typical application say GnuPG. It runs at a rate around 100MB/s. So in your case you're getting 40min decrypt times. Doing the math that's 40 * 60 * 100 = 240gb. Yeah introduce compression into the picture and you may get a 10th of that, but 10MB/s is still faster than most fiber and 240gb is a hell of a lot of data to be deploying on a regular basis.

The more I anaylize your edge cases I'm sticking with shenanigans.

Certain proprietary softwares cannot be altered without voiding warranty and license agreements.
Again, this shows how ignorant you are when it comes to enterprise software, at least that is licensed.

I'm not saying you don't know what you are talking about with scripting, but I am saying there are scenarios your opinions aren't applicable to, and this is one of them.

Yeah. I'm not an it guy, but I have enough computer sense and experience to know when others are blowing wind up my ass.
 
Last edited:
When single-threaded workloads can take longer than a few minutes, a few hundred MHz could mean the difference between hours to days.
Apparently no one here is thinking about long-term tasks and only about short-term personal usage - there is a world outside of your personal desktop...

You obviously never worked in enterprise where a few hundred MHz could make the difference between hours, and even days, depending on the single-threaded workloads.
There is more to single-threaded apps and performance than a personal gaming desktop running Windows.

The ignorance in this thread is pretty astounding...
Haha, no I work in enterprise where we use server grade hardware; cores and memory with performant I/O... I have no clue what kind of "Enterprise" you work at where you are purchasing user/personal grade PC equipment (this is a discussion about 3000 series CPU's...) and complaining about single core performance. If that's how your IT department runs, I'm sorry for you, but that doesn't change my argument. You're comparing some really odd "Enterprise" setup with some software that sucks and using that to combat like 8 different people saying the same thing. Keep using your argument, but I'm not even sure what it is at this point, you are trying to tell me a few hundred MHZ for a short period of time are going to magically save you minutes, hours and even days. If that is the case, you may want to find or write some better software rather than using a 12/24 CPU to run a single thread.

"there is a world outside of your personal desktop" -> I don't purchase 3000 series Ryzens for my job... we use Xeon's (and if I can convince management to let me get some Rome stuff I would). You know there is an entire world outside of your Enterprise that doesn't use Zen2 consumer CPU's for their single threaded speed... I am going to even hazard a guess that it's MORE common that they don't use them as you imply you are.
 
The more I anaylize your edge cases I'm sticking with shenanigans.
Yeah. I'm not an it guy, but I have enough computer sense and experience to know when others are blowing wind up my ass.
Do what you gotta do man, it won't change what I have to do, and deal with on a daily basis.
If we do ever get an improvement to make it at least lightly-threaded, I will share the info with you, and will look into it with the company we license the software from.

Hey, if there is a way to do it and we've been wrong all these years, I will admit that and will give you credit where credit is due. ;)
Going to look into it further...
 
You aren't wrong, and not all enterprises are equal.
If it were up to me, everyone would be using EPYC or Xeon workstations - that, unfortunately, isn't the case, and I have to use the tools at hand that we are given.
 
Yeah, who can ever forget the internet outrage from all those people that actually got exploited by spectre & meltdown, and how Intel didn't do a thing to fix it.

Yes and if a tree falls down in the forest and no one is around to hear it does it make a sound?
 
You aren't wrong, and not all enterprises are equal.
If it were up to me, everyone would be using EPYC or Xeon workstations - that, unfortunately, isn't the case, and I have to use the tools at hand that we are given.
I'm still saying that you're in a thread that is targeted at consumer level CPUs complaining of some Enterprise task that you have to do and then telling others that things exist outside of their home... well, they bought a consumer CPU for consumer loads... and we're arguing that it's what they are normally used for out of some odd use case you seem to have. Sure, there are some legacy software we have to support that may not be threaded, but we don't run them on a 12 core system by themselves, and it's still not common, especially for the hardware we're discussing.
 
I'm still saying that you're in a thread that is targeted at consumer level CPUs complaining of some Enterprise task that you have to do and then telling others that things exist outside of their home... well, they bought a consumer CPU for consumer loads... and we're arguing that it's what they are normally used for out of some odd use case you seem to have. Sure, there are some legacy software we have to support that may not be threaded, but we don't run them on a 12 core system by themselves, and it's still not common, especially for the hardware we're discussing.
The thing is, I have to work with what I am given, in which case single-core boosts do greatly assist our operations, and my point of all of this is that there are still existing scenarios where it does matter.
Someday we will have EPYC and/or Xeon workstations to work with...
 
Last edited:
The thing is, I have to work with what I am given, in which case single-core boosts do greatly assist our operations, and my point of all of this is that there are still existing scenarios where it does matter.
Someday we will have EPYC and/or Xeon workstations to work with...

Thing is, I'm thinking you're confusing single-core boost and all-core boost. If you disable boost, yeah well it will impact everything you do. Single-core boost is another beast... AMD have 1 star core which is known to boost higher than others. When you say higher max boost would benefit well the overall all-core boost should remain similar while the star core may boost a little more. The current "issue" isn't about all-core boost, it's about one core boosting a little higher than others... Ideally single threaded app would be assigned to that star core while others spread around but I doubt it's happening.

That why I said we would have more benefit to have baked auto parking core features based on AI (I guess?) than one core boosting higher...
When is the last time you saw 7 cores at <=3.6GHz and one at 4.3GHz on a 2700X ? (example here)...
 
Status
Not open for further replies.
Back
Top