HN user
Two things: a repaired 3090 24gb can be sold for 1000+ euro in Germany at least. It's still crazy expensive, so if you have the skills and devices, it can be done with a profit when you buy 3 broken cards and can sell 2 repaired one.
I never reballed any IC, so I would not dare to try it on my card.
CPU and GPUs are kind of a different game. I would rather like to have a shop, which has a good reputation on RTX gpu repairs. But thanks :)
Good idea. Haven't thought about mikrocontroller.net, I even have a login there for quite a long time.
I am open for shops with a good reputation
Apple didn't design the PowerPC or make custom variances. Motorola and IBM did it. Especially Altivec was added by Motorola, and IBM didn't like to add it to their PowerPC CPUs when Apple asked for help, when Motorola had the 500 MHz glitch bug back in the day.
There is a nice coverage on this topic at https://www.youtube.com/watch?v=Tld91M_bcEI (Why the Original Apple Silicon Failed)
FTP, Email, Webspace and newsgroups access was the norm. good old times at uunet ...
I used a similar budget and build something like this:
7x RTX 3060 - 12 GB which results in 84GB Vram AMD Ryzen 5 - 5500GT with 32GB Ram
All in a 19-inch rack with a nice cooling solution and a beefy power supply.
My costs? 1300 Euro, but yeah, I sourced my parts on ebay / second hand.
(Added some 3d printed parts into the mix: https://www.printables.com/model/1142963-inter-tech-and-gene... https://www.printables.com/model/1142973-120mm-5mm-rised-noc... https://www.printables.com/model/1142962-cable-management-fu... if you think about building something similar)
My power consumption is below 500 Watt at the wall, when using LLLMs,since I did some optimizations:
* Worked on power optimizations and after many weeks of benchmarking, the sweet spot on the RTX3060 12GB cards is a 105 Watt limit
* Created Patches for Ollama ( https://github.com/ollama/ollama/pull/10678) to group models to exactly memory allocation instead of spreading over all available GPUs (This also reduces the VRAM overhead)
* ensured that ASPM is used on all relevant PCI components (Powertop is your friend)
It's not all shiny:
* I still use PCIe3 X1 for most of the cards, which limits their capability, but all I found so far (PCIe Gen4 x4 extender and bifurcation/special PCIE routers) are just too expensive to be used on such low powered cards
* Due to the slow PCIe bandwidth, the performance drops significantly
* Max VRAM per GPU is king. If you split up a model over several cards, the RAM allocation overhead is huge! (See Examples in my ollama patch about). I would rather use 3x 48GB instead of 7x 12G.
* Some RTX 3060 12GB Cards do idle at 11-15 Watt, which is unacceptable. Good BIOSes like the one from Gigabyte (Windforce xxx) do idle at 3 Watt, which is a huge difference when you use 7 or more cards. These BIOSes can be patched, but this can be risky
All in all, this server idles at 90-100Watt currently, which is perfect as a central service for my tinkerings and my family usage.
If you want to see the included images, jump back to 2022: https://web.archive.org/web/20221218011802/https://akapugs.b...
I am a German native speaker as well. That is horrible idea!
A simple explanation can be seen here: https://www.youtube.com/watch?v=7j_NE6Pjv-E
cheap easy eats
and mostly unhealthy in comparison to self hunted / self-planted food.
Thanks for the clarification. Surely, If I run hashcat benchmark the power consumption goes nearly to 1400 Watt, but I also limited the max power consumption for each card to 100 Watt, which worked out better than limiting the max gpu frequency. To be fair, the most speed comes from the RAM frequency - as long as this is not limited, it works out great.
I took a fair amount of time to get everything to a reduced power level and measured several llm models (and hashcat for the extreme) to find the best speed per watt, which is usally around 1700-1900 mhz or limiting 3060 to 100 to 115 watt.
If I planned it in the first run, I may got away with a used mac studio, thats right. However, I incrementally added more cards as I moved further into exploration.
I didn't wanted to confront someone, but it looks like you either show of 4x 4090 or you keep silent
For optimal performance, we recommend the sum of VRAM + RAM to be at least 80GB+.
Oh nice! So I can try it in my local "low power/low cost" server at home.
My homesystem does run in a ryzen 5500 + 64gb RAM + 7x RTX 3060 12gb
So 64gb RAM plus 84gb VRAM
I dont want to brag around, but point to solutions for us tinkerers with a small budget and high energy costs.
such system can be build for around 1600 euro. The power consumption is around 520 watt.
I started with a AM4 Board (b450 Chipset) and one used RTX 3060 12gb which cost around 200 Euro used if you are patient.
There every additional GPU is connected with the pcie riser/extender to give the cards enough space.
After a while I had replaces the pcie cards with a single pcie x4 to 6x PCIe x1 extender.
It runs pretty nice. Awesome to learn and gain experience
That was a good one! Keep up your humor. It's a tough environment out there.
I really like build123d[1], which is based on CadQuery. There are a few nice IDEs for it and to make it simple: It's python based code-based CAD.
https://www.mengelke.de/Projekte/FritzBox-JSTool allow changing the domain of the fritz box routers.
It should be enough to fix that issue
"Principal Software Engineer" with 18 years of experience in Software development ... Nope. That's not a lower tier engineer at Microsoft.
I am writing a comment as a kind of love letter to the Asianometry YouTube channel. The author does such a great work, and I am really pleased with the details and structure of the essays.
If you haven't heard of Asianometry, give it a try!
And then you have Teams and a large Miro board open :-D
SCNR!
Exactly the same experience for me!
You can not join, when you have a commercial account (like me using google workspace). Wow! Thanks Google!
me too! Running BBS for several lines on one pc was awesome at that time.
I have female friends
It's good
I have male friends
It's okay
My boss is female
It's good
My boos is male
It's okay
+1 The Brother printers are awesome and just work as printers should do.
Well, it doesn't help you, but supporting independent online "magazines" give you much more quality articles than in the good old days. I support "The Pod", a german-based video- and computer game podcast. The quality is beyond everything I had in the years before and I bought magazines since the mid 80s.
I am pretty sure there are far more magazines/podcasts/blogs which work with just supporting readers, but you have to find them.
Thats super awesome! Hopefully it will spread around. Thanks for creating this adapter!
Because: medium.com ...
This is really awesome!
It looks pretty mature. Why is this compression not implemented? I checked out the youtube channel of ZFS, but it is nowhere mentioned in the years of monthly sync ups.
@nathanaldensr please bring up the options.