|
Post by Michael Mueller on Jun 20, 2020 18:23:06 GMT
Hello,
I have sold my pc, it was a reliable system, but I needed something smaller. It was a big casing.
I thought about a laptop, but it seems the new ones aren’t built to stay alive for more than three years. What are your experiences? And are they not suited for crunching boinc? Would you recommend a laptop?
My new system could look like this, it should be a budget system, not high end. What do you guys think about it? And do I need cuda for boinc or are the amd graphics card good too? Because at machine learning are cudas needed.
Power: be quiet! Pure Power 11 500 Watt (80+) Mainboard: MSI B450M Pro-VDH Max, AMD B450 Processor: AMD Ryzen 5 3600 6x 4.2GHz Cooler: Original AMD Wraith Stealth Cooler Memory: 16GB DDR4-Ram PC-3000 (2x 8GB) Graphic card: Nvidia GeForce GTX1660 6GB, Palit StormX M.2/Optane: NVME M.2 SSD 500GB Kingston A2000 Hard drive/SSD: 1000GB S-ata3,600MB/Sek.
+Windows 10 Pro, its for "free" at configurator.
Cost: 741,70€ with 19% taxes. 828,24$ 670,7 pounds. Thanks. Greetings.
|
|
|
Post by Bozo on Jun 20, 2020 23:07:03 GMT
My Laptop has been running boinc 24/7 for years without a single issue but I would always go with a desktop pc. I do have a couple of newer laptops running but i don't know if they will stand the test of time (I don't run run them flat out)
I am a bit behind the times on the latest hardware but if it helps the power supply on my desktop I have run 24/7 for years is the be quiet pure power l8 530w. I have not had any issues running it in a case without the side - Never had a problem and yes they are quiet. My video card is the only noise i can hear now i changed cpu fan.
AMD cards are better at some projects, any projects that use opencl run better on AMD (milkway@home for one). other projects (gpugrid) run better on cuda.
I am hoping to finally get around to a new pc myself, will probably be going Ryzen this time around.
|
|
|
Post by 888 on Jun 21, 2020 14:29:35 GMT
Personally I wouldn't recommend a laptop for crunching. The cooling system isn't up to dissipating the heat from a CPU running at 100% continuously, and would probably shorten the life of the processor and other components. The system that you've shown looks good, and the Ryzen processors seem great, although I've never ran a system with one. All I could say is to make sure that you can upgrade some of the components at some time in the future, if needed. The motherboards looks as though it can take faster CPU's, and more memory. The 500W power supply should be ok, as long as you dont go crazy with a new mega graphics card.
|
|
|
Post by jenesuispasbavard on Jun 21, 2020 20:23:07 GMT
I love my laptop (Surface Book 2) - it has a pretty fast 4-core, 8-thread Intel Core i5-8250U CPU, and cooling-wise it's probably the best laptop I've ever used. Even running BOINC non-stop, the CPU remains near its long-term power limit of 15W, and it's basically a slab of metal so either the fan doesn't even come on or runs very quietly. I've never heard this laptop make any noise. And in the newer one they've managed to also cram an Nvidia GTX 1660Ti in there, but that's super expensive. That said, unless you need the portability, you're much better off with a desktop for crunching. Without the 15W limitation, you'll get more sustained performance from almost any desktop processor than most laptop ones. The Ryzen 5 3600 is a great CPU, and if you're not too concerned about noise you can also run the CPU at higher wattage - I run my 3700X at 95W instead of the default 65W - but the bundled cooler will quickly get loud at higher power. And 500W should be plenty for that system. I had accidentally left my CPU and GPU (3700X + 2080) in their high-power-consumption gaming settings and they still only pulled a total of ~475W from the wall running CPU+GPU computing projects instead of the usual ~350W. Regarding the graphics card, you're probably better off with Nvidia at this time, until AMD gives Nvidia any serious competition at the high end (hopefully by the end of the year when the new consoles are out?). Regarding machine learning applications, Nvidia's RTX cards (direct competitor to the GTX 1660 would be the RTX 2060) support higher-performance applications with FP16 code (see here and here).
|
|
|
Post by prokaryote on Jul 10, 2020 9:42:20 GMT
IMO, if you'd like to do distributed computing, probably the RTX 2060 is a good buy. For ML, maybe price out cost of projects done on the cloud?
|
|