How is that For Flexibility?

As everyone is well conscious, the world is still going nuts attempting to develop more, more recent and better AI tools. Mainly by throwing unreasonable amounts of money at the problem. A number of those billions go towards developing low-cost or free services that operate at a considerable loss. The tech giants that run them all are hoping to bring in as numerous users as possible, so that they can capture the marketplace, and become the dominant or only celebration that can offer them. It is the classic Silicon Valley playbook. Once dominance is reached, anticipate the enshittification to start.

A likely way to earn back all that cash for developing these LLMs will be by tweaking their outputs to the liking of whoever pays one of the most. An example of what that such tweaking looks like is the rejection of DeepSeek’s R1 to discuss what happened at Tiananmen Square in 1989. That one is certainly politically motivated, however ad-funded services will not precisely be fun either. In the future, I completely anticipate to be able to have a frank and sincere conversation about the Tiananmen events with an American AI representative, however the only one I can manage will have presumed the persona of Father Christmas who, while holding a can of Coca-Cola, will sprinkle the stating of the terrible occasions with a cheerful „Ho ho ho … Didn’t you understand? The vacations are coming!“

Or possibly that is too improbable. Today, dispite all that money, the most popular service for code completion still has problem working with a number of easy words, despite them being present in every dictionary. There must be a bug in the „totally free speech“, or something.

But there is hope. Among the tricks of an approaching player to shake up the marketplace, is to damage the incumbents by launching their design free of charge, under a permissive license. This is what DeepSeek just made with their DeepSeek-R1. Google did it previously with the Gemma models, as did Meta with Llama. We can download these designs ourselves and run them on our own hardware. Even better, individuals can take these designs and scrub the predispositions from them. And we can download those scrubbed models and run those on our own hardware. And after that we can finally have some truly beneficial LLMs.

That hardware can be an obstacle, though. There are 2 options to select from if you wish to run an LLM locally. You can get a huge, effective video card from Nvidia, or you can buy an Apple. Either is pricey. The main spec that indicates how well an LLM will carry out is the amount of memory available. VRAM in the case of GPU’s, normal RAM in the case of Apples. Bigger is much better here. More RAM implies larger designs, which will dramatically improve the quality of the output. Personally, I ‘d state one needs at least over 24GB to be able to run anything beneficial. That will fit a 32 billion parameter design with a little headroom to spare. Building, or buying, a workstation that is equipped to handle that can quickly cost thousands of euros.

So what to do, if you don’t have that amount of cash to spare? You purchase pre-owned! This is a feasible option, but as constantly, there is no such thing as a totally free lunch. Memory may be the main concern, however don’t ignore the value of memory bandwidth and other specifications. Older equipment will have lower performance on those elements. But let’s not worry excessive about that now. I have an interest in developing something that at least can run the LLMs in a functional way. Sure, the current Nvidia card may do it quicker, however the point is to be able to do it at all. Powerful online models can be good, but one must at the minimum have the alternative to change to a local one, if the scenario calls for it.

Below is my effort to construct such a capable AI computer system without spending too much. I ended up with a workstation with 48GB of VRAM that cost me around 1700 euros. I could have done it for less. For example, it was not strictly necessary to buy a brand new dummy GPU (see below), or I could have found someone that would 3D print the cooling fan shroud for me, rather of shipping a ready-made one from a distant country. I’ll admit, I got a bit impatient at the end when I discovered I needed to purchase yet another part to make this work. For me, this was an acceptable tradeoff.

Hardware

This is the complete cost breakdown:

And this is what it looked liked when it initially booted up with all the parts installed:

I’ll offer some context on the parts below, and after that, systemcheck-wiki.de I’ll run a couple of quick tests to get some numbers on the efficiency.

HP Z440 Workstation

The Z440 was an easy choice due to the fact that I already owned it. This was the beginning point. About 2 years earlier, I desired a computer that might act as a host for my virtual devices. The Z440 has a Xeon processor with 12 cores, and this one sports 128GB of RAM. Many threads and a lot of memory, that must work for hosting VMs. I bought it secondhand and then swapped the 512GB hard disk drive for a 6TB one to save those virtual makers. 6TB is not needed for running LLMs, and therefore I did not include it in the breakdown. But if you plan to gather numerous models, 512GB may not be enough.

I have actually pertained to like this workstation. It feels all very strong, and I have not had any issues with it. At least, until I began this job. It ends up that HP does not like competitors, and I experienced some troubles when switching parts.

2 x NVIDIA Tesla P40

This is the magic component. GPUs are pricey. But, similar to the HP Z440, often one can discover older devices, that utilized to be top of the line and is still extremely capable, pre-owned, for fairly little cash. These Teslas were meant to run in server farms, for things like 3D rendering and other graphic processing. They come equipped with 24GB of VRAM. Nice. They fit in a PCI-Express 3.0 x16 slot. The Z440 has 2 of those, so we buy 2. Now we have 48GB of VRAM. Double great.

The catch is the part about that they were indicated for servers. They will work fine in the of a normal workstation, however in servers the cooling is handled differently. Beefy GPUs take in a lot of power and can run very hot. That is the reason customer GPUs constantly come geared up with huge fans. The cards require to take care of their own cooling. The Teslas, nevertheless, have no fans whatsoever. They get simply as hot, however expect the server to provide a stable flow of air to cool them. The enclosure of the card is somewhat formed like a pipe, and you have two choices: blow in air from one side or blow it in from the other side. How is that for flexibility? You definitely should blow some air into it, however, or you will damage it as quickly as you put it to work.

The option is easy: simply install a fan on one end of the pipeline. And certainly, it appears a whole cottage industry has grown of people that sell 3D-printed shrouds that hold a standard 60mm fan in simply the best location. The problem is, the cards themselves are currently quite bulky, and it is difficult to discover a configuration that fits 2 cards and 2 fan mounts in the computer case. The seller who offered me my 2 Teslas was kind adequate to consist of 2 fans with shrouds, but there was no chance I could fit all of those into the case. So what do we do? We buy more parts.

NZXT C850 Gold

This is where things got irritating. The HP Z440 had a 700 Watt PSU, which might have sufficed. But I wasn’t sure, and I needed to buy a brand-new PSU anyway since it did not have the best adapters to power the Teslas. Using this handy site, I deduced that 850 Watt would be enough, and I bought the NZXT C850. It is a modular PSU, meaning that you only need to plug in the cable televisions that you really need. It included a neat bag to keep the spare cable televisions. One day, I may offer it a good cleansing and utilize it as a toiletry bag.

Unfortunately, HP does not like things that are not HP, so they made it difficult to swap the PSU. It does not fit physically, and they also altered the main board and CPU adapters. All PSU’s I have ever seen in my life are rectangular boxes. The HP PSU likewise is a rectangle-shaped box, however with a cutout, making certain that none of the normal PSUs will fit. For users.atw.hu no technical reason at all. This is just to mess with you.

The mounting was eventually solved by utilizing 2 random holes in the grill that I somehow managed to align with the screw holes on the NZXT. It sort of hangs steady now, and I feel fortunate that this worked. I have seen Youtube videos where individuals resorted to double-sided tape.

The port needed … another purchase.

Not cool HP.

Gainward GT 1030

There is another issue with using server GPUs in this consumer workstation. The Teslas are intended to crunch numbers, not to play computer game with. Consequently, they do not have any ports to connect a display to. The BIOS of the HP Z440 does not like this. It declines to boot if there is no other way to output a video signal. This computer will run headless, however we have no other option. We have to get a 3rd video card, that we do not to intent to use ever, just to keep the BIOS happy.

This can be the most scrappy card that you can find, of course, but there is a requirement: we should make it fit on the main board. The Teslas are bulky and fill the 2 PCIe 3.0 x16 slots. The only slots left that can physically hold a card are one PCIe x4 slot and one PCIe x8 slot. See this site for some background on what those names suggest. One can not buy any x8 card, however, because typically even when a GPU is advertised as x8, the actual connector on it may be just as large as an x16. Electronically it is an x8, physically it is an x16. That won’t deal with this main board, we truly require the little adapter.

Nvidia Tesla Cooling Fan Kit

As said, asteroidsathome.net the obstacle is to discover a fan shroud that suits the case. After some searching, I discovered this set on Ebay a bought two of them. They came delivered total with a 40mm fan, and it all fits completely.

Be warned that they make an awful great deal of noise. You don’t desire to keep a computer system with these fans under your desk.

To keep an eye on the temperature, I worked up this quick script and put it in a cron task. It regularly reads out the temperature level on the GPUs and sends that to my Homeassistant server:

In Homeassistant I included a graph to the control panel that displays the worths with time:

As one can see, the fans were loud, but not particularly efficient. 90 degrees is far too hot. I searched the web for an affordable upper limitation but might not discover anything specific. The paperwork on the Nvidia website discusses a temperature of 47 degrees Celsius. But, what they indicate by that is the temperature of the ambient air surrounding the GPU, not the determined value on the chip. You understand, the number that actually is reported. Thanks, Nvidia. That was useful.

After some more browsing and checking out the viewpoints of my fellow internet residents, my guess is that things will be fine, provided that we keep it in the lower 70s. But do not quote me on that.

My very first effort to correct the situation was by setting an optimum to the power intake of the GPUs. According to this Reddit thread, one can decrease the power usage of the cards by 45% at the cost of only 15% of the efficiency. I tried it and … did not see any distinction at all. I wasn’t sure about the drop in efficiency, having only a couple of minutes of experience with this configuration at that point, however the temperature level qualities were certainly unchanged.

And after that a light bulb flashed on in my head. You see, prior to the GPU fans, there is a fan in the HP Z440 case. In the photo above, it remains in the best corner, inside the black box. This is a fan that draws air into the case, and I figured this would work in tandem with the GPU fans that blow air into the Teslas. But this case fan was not spinning at all, since the remainder of the computer did not need any cooling. Looking into the BIOS, I discovered a setting for the minimum idle speed of the case fans. It ranged from 0 to 6 stars and was currently set to 0. Putting it at a greater setting did wonders for the temperature level. It likewise made more sound.

I’ll reluctantly admit that the third video card was practical when changing the BIOS setting.

MODDIY Main Power Adaptor Cable and Akasa Multifan Adaptor

Fortunately, often things simply work. These two products were plug and play. The MODDIY adaptor cable linked the PSU to the main board and CPU power sockets.

I used the Akasa to power the GPU fans from a 4-pin Molex. It has the great feature that it can power two fans with 12V and 2 with 5V. The latter certainly reduces the speed and hence the cooling power of the fan. But it likewise minimizes noise. Fiddling a bit with this and the case fan setting, I found an appropriate tradeoff in between sound and temperature. For now at least. Maybe I will need to revisit this in the summertime.

Some numbers

Inference speed. I collected these numbers by running ollama with the– verbose flag and asking it five times to write a story and averaging the result:

Performancewise, ollama is configured with:

All designs have the default quantization that ollama will pull for you if you don’t define anything.

Another essential finding: Terry is without a doubt the most popular name for a tortoise, followed by Turbo and Toby. Harry is a preferred for hares. All LLMs are loving alliteration.

Power intake

Over the days I kept an eye on the power consumption of the workstation:

Note that these numbers were taken with the 140W power cap active.

As one can see, there is another tradeoff to be made. Keeping the model on the card improves latency, however takes in more power. My existing setup is to have actually 2 models packed, one for coding, the other for generic text processing, and keep them on the GPU for up to an hour after last usage.

After all that, am I pleased that I began this task? Yes, I believe I am.

I spent a bit more cash than prepared, however I got what I wanted: parentingliteracy.com a way of locally running medium-sized designs, entirely under my own control.

It was a great choice to start with the workstation I already owned, and see how far I might feature that. If I had actually started with a new maker from scratch, it certainly would have cost me more. It would have taken me a lot longer too, as there would have been a lot more choices to select from. I would likewise have actually been very lured to follow the hype and purchase the current and greatest of everything. New and glossy toys are fun. But if I purchase something new, I want it to last for years. Confidently anticipating where AI will go in 5 years time is impossible today, so having a cheaper maker, that will last a minimum of some while, feels satisfying to me.

I wish you good luck by yourself AI journey. I’ll report back if I discover something new or fascinating.