Posts by ProDigit

1) Message boards : Questions and problems : Computer type not supported (Message 111644)
Posted 23 Apr 2023 by ProDigit
Post:
From my android phone,

World community grid, Einstein, Universe and Rosetta are the only 4 projects supporting arm A50 (efficiency) series cores.

If you're running an A70 series (performance) core, results might be different.

Universe usually has the most WUs, but occasionally have months of nothing.

WCG is second,

third Einstein and Rosetta. These 2 only very sporadically have WUs for arm.
2) Message boards : Questions and problems : Which phone. (Message 111643)
Posted 23 Apr 2023 by ProDigit
Post:
Personally I think the Samsung Galaxy A72 were the best phones. You can still find them online. Upgrade the os to android 11 or 12, enable power saving (lower performance) and battery protect (charge to 85% & slow charge only), then tell boinc to run 6 out of 8 threads.

The snapdragon inside the A72 has 2P and 6E cores. That's 2E cores more than most other phones.

My recommendation is to keep the phone in a cool place. Like all phones, the battery swells up when too hot.

I set mine to 36C max, which is already too high if you ask me.
But in winter time (below 60F ambient) you can run boinc at full speed (disable power saving).
3) Message boards : Android : Loses Wi-Fi connection after a few hours (Message 111642)
Posted 23 Apr 2023 by ProDigit
Post:
Unless jailbroken, you can't force a reboot without human interaction, unless you remove the battery, and have the ac adapter plugged in a timer.. Just set the wus to 0.5 days and an additional 0.6 days in settings, so your phone will buffer up WUs.
4) Message boards : Android : Crunching from battery on old Android (Message 111641)
Posted 23 Apr 2023 by ProDigit
Post:
Check to see if minimum battery level is set.

On such old phones, it's generally better to throw them in the garbage. Not many projects still support older slower hardware.
5) Message boards : GPUs : GPU questions. (Message 111640)
Posted 23 Apr 2023 by ProDigit
Post:
1. I have observed if you have one RTX 30x or RTX 40x and have the build in intel GPU processing working units the rate of processing goes down ~10 %. IE if not running an RTX 3070 ti without intel GPU the processing rate ~31% if we add the intel processor the rate drops to ~19-21%. Anyone know what the underlying cause of might be the slow down.



Cpu bus or ram bandwidth bottleneck. Many tests have shown that 20-22 threads on older ryzen cpus start to bottleneck the ram at 3600-3800Mhz ram speed.
Faster ram (5000Mhz DDR4) somewhat alleviates the issue, (can run up to 30 threads fine), but using the IGP aggravates it in 2 ways:

1- An IGP heavily makes use of ram memory both in read and write, and it bottlenecks the cpu interconnects. This causes delays and data collision in ram access, read/writes. Additionally, the igp uses up 4pcie lanes. Not a lot of study is done to see if pcie lanes can cause other lanes to slow down at an interconnect point to the processors, as in some cases it takes 1 to 2 full cpu threads to process 16 lanes. So the more pcie lanes used, the less cpu threads will be available.

2- using the IGP causes the cpu to increase in internal temperature, which in turn lowers boost frequency of both the cpu and igpu core frequencies, and returns lower CPU and IGP PPD. Lower cpu frequency may also adversely effect pcie transfer speeds, causing additional lag on the dgpu, resulting in lower PPD, though the slowdown should be barely registerable. Still, it's a slowdown nonetheless.

My recommendation would be, to just use a DGPU for boinc, and use up to ~80% of your available cpu threads (depending from cpu to cpu), and use the IGP only for display out. If you plan on installing 2 dGPUs, it might be best to turn off the IGP in bios, to free a few PCIe lanes for the dGPUs. Running the monitor from your primary dGPU is ok, if you're not using the pc. But if you also want to play games or watch videos, you might want to keep the igp enabled and connect the monitor through that.
6) Message boards : Questions and problems : Are most of the Projects dead? (Message 109274)
Posted 17 Jul 2022 by ProDigit
Post:
I removed Rosetta from my computers, as they only have 2GB of RAM, and have about 3,5GB available on free disk space (this includes Swap partition).
Since they have older, low power 1,44Ghz quadcore CPUs, they're not suited for those massive WUs anyway.
7) Message boards : Questions and problems : Is WCG (World Community Grid) back? (Message 109273)
Posted 17 Jul 2022 by ProDigit
Post:
If yes, do we need to update the address?

If so, what's the best way?

I notice the past 5 or so months, my PPD results are very low.

WCG is not fully back yet. Small test batches are sent out, but most volunteers will probably not get any work.
It will probably take weeks before we can consider WCG as being fully back.
There's still tons of issues that they have to fix before WCG goes full bore.

Regarding the address, If you see a message from WCG in the BOINC event log that says:
"This project is using an old URL. When convenient, remove the project, then add https://master.worldcommunitygrid.org/" IGNORE that message, and do NOT do what it says.
The problem is with WCG, and not with your BOINC clients settings. It's up to the WCG admins to fix the problem, so the erroneous message goes away.


Thank you.
It would save me a lot of time, since I'm still running about 48 CPU units, most of them doing Universe@home right now.
And whatever other work there's available.
Anything but Qchempedia, which apparently needs better hardware.
8) Message boards : GPUs : Any more Intel GPU projects? (Message 109272)
Posted 17 Jul 2022 by ProDigit
Post:
Most Intel 9th gen or earlier CPUs have very lousy IGPs.
They crunch data about as fast as the CPU itself (have similar FLOPS performance), but increase the CPU TDP usage.
If your 9th gen (or older) intel CPU crunches at 100%, with a temp of 60-75C, adding the GPU to crunch at 100% will increase the temps to 90+C.
Not recommended.
Newer 11th and 12th gen Intel IGPs perform about as lousy as a GT 1030.
I would recommend to just spend some money on a GPU, and plug it in the system.
And forget about CPU crunching if PPD is what you care about. Even if you have the CPU available.
The amount of $$$ you'll waste a year running an Intel IGP (about $35/year/IGP) should easily allow you to buy a small GPU that will fold a lot faster for the same power usage.
I get 5x speed from my i5 8600K's GPU+1core doing an Einstein Gamma than I would just using the core, so it's like having 10 cores on a 6 core CPU. Might aswell use all parts of the chip at once. Extra cooling is easy enough, I stuck a larger heatsink on it and used SYY thermal paste, it's running at 80C in a 19C room. With the stock cooler and provided transfer compound, it was bouncing off the 100C limiter. 80C is my preferred limit for everything, 90C if I have to. Higher than that errors and crashing occur.


Too bad prices of GPUs are still high.
Before the shortage, you could easily snatch a GT1030, GTX1050, or 1060 from $50 to $150, which would net you A LOT more PPD than any CPU at the time could (take note, the Ryzen 1000 series were still very uncommon.
9) Message boards : Questions and problems : Most computers not getting work. (Message 109264)
Posted 16 Jul 2022 by ProDigit
Post:
Try adding einstein, LHC, Cosmology, Gaia, and Universe@home.

Most of my units are running Universe@home right now.
10) Message boards : Questions and problems : Is WCG (World Community Grid) back? (Message 109130)
Posted 10 Jul 2022 by ProDigit
Post:
If yes, do we need to update the address?

If so, what's the best way?

I notice the past 5 or so months, my PPD results are very low.
11) Message boards : Projects : What does the future of BOINC look like? (Message 108869)
Posted 4 Jul 2022 by ProDigit
Post:
F@H and Boinc are incompatible, and will never merge.
I do hope that Boinc will start running on more devices. Specifically more server based ARM or RISC CPUs; but for that massive parallel CPUs need to become more mainstream.
It's about time, I can buy a 100 core ARM or RISC CPU, that runs at 3 to 4Ghz, and consumes only 30-65W.
12) Message boards : GPUs : Any more Intel GPU projects? (Message 108867)
Posted 4 Jul 2022 by ProDigit
Post:
Most Intel 9th gen or earlier CPUs have very lousy IGPs.
They crunch data about as fast as the CPU itself (have similar FLOPS performance), but increase the CPU TDP usage.
If your 9th gen (or older) intel CPU crunches at 100%, with a temp of 60-75C, adding the GPU to crunch at 100% will increase the temps to 90+C.
Not recommended.
Newer 11th and 12th gen Intel IGPs perform about as lousy as a GT 1030.
I would recommend to just spend some money on a GPU, and plug it in the system.
And forget about CPU crunching if PPD is what you care about. Even if you have the CPU available.
The amount of $$$ you'll waste a year running an Intel IGP (about $35/year/IGP) should easily allow you to buy a small GPU that will fold a lot faster for the same power usage.
13) Message boards : Questions and problems : Recurring issues with grayed out remove option "they don't seem to care about. Best ask at their site" (Message 108724)
Posted 26 Jun 2022 by ProDigit
Post:
You can't remove a project that's been added via account manager.
You either log off from account manager, and then remove the project, or go to the project website, and remove it there from the account manager page.

Don't use an account manager, if you plan on changing your setup frequently.
Boinc BAM is easy to add or remove.
After you update the website, just update the project page on the client/manager, and the removed projects will remove from your client.
14) Message boards : Android : GPU on Android? (Message 108496)
Posted 16 Jun 2022 by ProDigit
Post:
On android, no.
Most games underperform what the hardware can do.
15) Message boards : Android : GPU on Android? (Message 108457)
Posted 15 Jun 2022 by ProDigit
Post:
PC games max out the GPU. What's the point of a powerful GPU you can't fully use? They might aswell have put in a slower one. You should have written to the phone manufacturer and asked for a return of their piece of crap hardware. Imagine if your car overheated if you tried to drive at maximum speed.


There are so many designs for each type of SOC. Every brand makes them different.
Some brands make em run at their highest frequency (for performance reasons) and build in a cooling solution, while others focus on low power consumption.
Sometimes this has to do with the phone size (form factor), that is not able to extract enough heat from the CPU.
Hence a game needs to work acceptably on a certain SOC on all phones; which is why some games use sync (30 or 60fps), and any extra computing power is untapped.
It would nearly be impossible to make a game utilize 100% of the same resources, though one device is faster than the other.
Phones are usually low power devices. They focus on low heat, long battery life.
So whenever a game utilizes only 30FPS, but they can easily display 50, the phone internals use clever trickery to still save that power. Mostly in the form of duplicating screens (like 1 out of 6 screens, so a game running at 50FPS, can display that it's running the game at 60FPS, even though 1 out of 6 frames is a duplicate...

It's only one of the so many tricks manufacturers bypass the limitations of the hardware.
Another is to not offer ray tracing, or reduce the LOD on games.
Other limitations could be reduce the screen resolution of the device, so it can never play games at too low frame rates...

It makes it nearly impossible for developers to make a game 100% utilize all resources on all phones.
They rather make fewer variants of the game that are compatible with both faster and slower versions of the same chipset..
16) Message boards : Android : GPU on Android? (Message 108410)
Posted 11 Jun 2022 by ProDigit
Post:
most games don't max out the gpus.
I used to have a LeEco 3 Pro, with a Snapdragon 820 system on chip.
When I played Robot wars, that thing got so hot you could burn yourself on it.
I wrote robot wars about the phone heating up, and they fixed it with a next update.
The game still looks the same, save for fewer ray tracing and bullets flying.
The ordinary person wouldn't notice unless looking at screen shots.
But the phone ran much cooler.

Another thing that heats up the SOC is lots of read/writes to the L-Cache.
I can run Boinc fine on my Galaxy A72 on 5 cores (runs around 30C), but at 6 cores the temperature goes up to 36C (and sometimes exceeds 42C if I combine projects with too many Einstein WUs).
It's not the added CPU, it's lack of L-Cache that causes more Read/Writes, and increases temperature.

Games are optimized to keep the most accessed code within the L-Cache. But OpenCL might need much more access to L-Cache; many more read/writes, as the memory results always change after a calculation.
Unlike games where a lot of that data is static data (like textures, or geometric data).
17) Message boards : Questions and problems : Issues after every [L]Ubuntu system update. (Message 108392)
Posted 10 Jun 2022 by ProDigit
Post:
There is no need to always update Ubuntu and variants.
I just recently updated my 18.04.1 to 18.04.4, which was mainly security updates, and these systems have been running this configuration for about 6 months now.
No need to fix what isn't broken, unless you are a daily user, and need the upgrades for what or what not.

In most cases, upgrades of the linux kernel will result in slower performance, as they get more and more bloated.
The only reason I can see why you would want the absolute latest, is if you're running the absolute latest CPUs (those that come with performance and efficiency cores).
18) Message boards : Android : GPU on Android? (Message 108372)
Posted 9 Jun 2022 by ProDigit
Post:
Their CPUs aren't bad, how come the GPUs suck?

Presumably people max out the phone's GPU playing games, probably on their lap without active cooling.

Their GPUs are actually more powerful than their CPUs (at least the Cortex A50 series).
However, they also consume a lot of power (much like a Cortex A70 series CPU).
Thus heat is the main reason why they don't do Boinc GPU projects on ARM (mainly because they are built in phones).
There's currently too few ARM devices out there that offer sufficient cooling. Some IOT devices, like singleboard computers, and Apple laptops.
No other laptops, desktops or servers out yet, and if there are, the low availability on Boinc often doesn't justify writing projects in OpenCL for ARM.

Aside from heat problems (where phones could potentially run GPU WUs meant for servers and overheat), there could also be compatibility issues, where GPU WUs work for one type of ARM (say Snapdragon IGPs), but won't work on mediatek or kirin devices, and where they basically error out every WU.
R&D lacks time and money for such projects.

We will be happy that ARM or RISCV will come out with a nice server CPU for the end user, like the much anticipated Ampere Altra server CPU.
It's promised for nearly a decade, but still today is hard to nearly impossible to buy for regular people like us.
19) Message boards : Questions and problems : Does the extra cache of the new 5800X3D increase crunching? (Message 108346)
Posted 8 Jun 2022 by ProDigit
Post:
.. and just as I write this, news from ryzen 7000 cpus (Zen4) comes out...

Perhaps a 7950x or 7900x might be the best buy.
20) Message boards : Android : GPU on Android? (Message 108343)
Posted 8 Jun 2022 by ProDigit
Post:
There are no GPU WUs for ARM devices.
I believe neither for Apple...
OpenCL capable devices don't mean they'll perform well.
The mobile GPUs in our phones have between 2 to 12 shaders for gaming phones.
Compare that to 2k to 8k shaders on modern GPUs.
You'd have to run your Cellphone a 1000+ days or years to have the equivalent of 1 day or year of dGPU folding.

Also, if your phone has 12 EUs in it, they either need active cooling, or used in an extremely cold room, like 50F or below, or so, as most cellphone GPUs are more power hungry than their CPU.


Next 20

Copyright © 2024 University of California.
Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.