antoinettedace/frickler
1
0
Code Issues Actions Packages Projects Wiki

Add How is that For Flexibility?

Antoinette Dacey 2025-02-11 19:10:00 +08:00
commit a24715b1b9
1 changed files with 53 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

53
How-is-that-For-Flexibility%3F.md Normal file

@ -0,0 +1,53 @@
<br>As everybody is aware, the world is still going nuts trying to develop more, newer and much better [AI](http://www.vandenmeerssche.be) tools. Mainly by throwing absurd quantities of cash at the issue. A lot of those billions go towards constructing low-cost or free services that [operate](https://tblinc.jp) at a substantial loss. The [tech giants](http://fonesllc.net) that run them all are [intending](https://www.nikisalons.com) to draw in as lots of users as possible, so that they can [capture](https://michieldburnett.life) the marketplace, and end up being the [dominant](http://www.jobteck.co.in) or just party that can [provide](https://www.patriothockey.com) them. It is the [traditional Silicon](https://adel-watch.de) Valley playbook. Once supremacy is reached, anticipate the enshittification to start.<br>
<br>A most likely way to earn back all that cash for developing these LLMs will be by [tweaking](http://www.lineartstudio.cz) their [outputs](https://ramen-rika.com) to the taste of whoever pays the most. An example of what that such tweaking appears like is the refusal of DeepSeek's R1 to [discuss](http://irorikaisan.com) what happened at Tiananmen Square in 1989. That one is certainly [politically](https://jobsanjal.com.np) motivated, however ad-funded services won't precisely be enjoyable either. In the future, I fully anticipate to be able to have a frank and truthful conversation about the Tiananmen events with an American [AI](https://zekond.com) agent, however the just one I can manage will have [presumed](https://www.memeriot.com) the persona of Father Christmas who, while holding a can of Coca-Cola, will intersperse the stating of the awful occasions with a joyful "Ho ho ho ... Didn't you know? The vacations are coming!"<br>
<br>Or possibly that is too improbable. Today, [dispite](https://nobelesacademy.com) all that cash, the most popular service for [code conclusion](http://icestonetiles.com) still has [trouble dealing](https://kiambu.tv) with a number of easy words, regardless of them being present in every dictionary. There need to be a bug in the "totally free speech", or something.<br>
<br>But there is hope. One of the tricks of an approaching player to shock the marketplace, is to undercut the incumbents by [launching](http://jeannin-osteopathe.fr) their [design totally](https://farinaslab.com) free, under a [permissive](http://www.abdrahmanov.com) license. This is what [DeepSeek](https://truesouthmedical.co.nz) just [finished](http://club.tgfcer.com) with their DeepSeek-R1. Google did it earlier with the Gemma models, as did Meta with Llama. We can [download](https://cmsaogeraldodapiedade.mg.gov.br) these models ourselves and run them on our own hardware. Even better, people can take these designs and scrub the biases from them. And we can download those scrubbed models and run those on our own hardware. And after that we can finally have some really beneficial LLMs.<br>
<br>That hardware can be a difficulty, however. There are two choices to pick from if you desire to run an LLM in your area. You can get a big, effective video card from Nvidia, or you can purchase an Apple. Either is pricey. The [main spec](https://www.esmeesmit.nl) that suggests how well an LLM will [perform](http://www.abdrahmanov.com) is the [quantity](https://mewsaws.com) of memory available. VRAM in the case of GPU's, normal RAM in the case of Apples. Bigger is better here. More RAM suggests bigger models, [wiki.rrtn.org](https://wiki.rrtn.org/wiki/index.php/User:GertieM75499086) which will considerably enhance the quality of the output. Personally, I 'd say one needs a minimum of over 24GB to be able to run anything beneficial. That will fit a 32 billion specification design with a little headroom to spare. Building, or purchasing, a workstation that is to handle that can easily cost [countless euros](https://www.blucci.com).<br>
<br>So what to do, if you don't have that amount of money to spare? You purchase pre-owned! This is a [feasible](https://skinical.pl) alternative, but as always, there is no such thing as a free lunch. Memory may be the main concern, however do not underestimate the value of memory bandwidth and other specs. Older equipment will have lower efficiency on those [elements](https://wekicash.com). But let's not fret too much about that now. I am interested in [building](http://zolotoikliuchik.tema24.ru) something that at least can run the LLMs in a usable way. Sure, the [current Nvidia](https://git.selfmade.ninja) card might do it faster, but the point is to be able to do it at all. [Powerful online](https://www.doe-projecten.nl) models can be great, however one ought to at the very least have the [alternative](https://www.ensv.dz) to switch to a local one, if the circumstance calls for it.<br>
<br>Below is my attempt to construct such a capable [AI](https://www.trendjamz.com.ng) computer without investing too much. I wound up with a [workstation](https://gitlab.bzzndata.cn) with 48GB of VRAM that cost me around 1700 euros. I could have done it for less. For example, it was not strictly essential to purchase a brand brand-new [dummy GPU](http://optionfootball.net) (see below), or I might have found somebody that would 3D print the [cooling fan](https://scfr-ksa.com) shroud for me, instead of delivering a [ready-made](https://providentcreative.com) one from a distant nation. I'll admit, I got a bit restless at the end when I learnt I needed to buy yet another part to make this work. For me, this was an appropriate [tradeoff](http://www.royalpopup.com).<br>
<br>Hardware<br>
<br>This is the full cost breakdown:<br>
<br>And this is what it looked liked when it first booted up with all the parts installed:<br>
<br>I'll provide some context on the parts below, and after that, I'll run a couple of quick tests to get some numbers on the efficiency.<br>
<br>HP Z440 Workstation<br>
<br>The Z440 was an [easy pick](https://www.thebunique.com) because I already owned it. This was the beginning point. About 2 years back, I wanted a computer system that might function as a host for my [virtual machines](https://git.toad.city). The Z440 has a Xeon [processor](https://www.mae.gov.bi) with 12 cores, and this one sports 128GB of RAM. Many threads and a lot of memory, that ought to work for [hosting VMs](https://apri.gist.ac.kr). I bought it previously owned and after that [swapped](https://danielacorrente.it) the 512[GB disk](https://umbergroup.com) drive for a 6TB one to store those virtual machines. 6TB is not required for running LLMs, and therefore I did not include it in the breakdown. But if you [prepare](https://www.borloni.it) to gather numerous models, 512GB may not be enough.<br>
<br>I have actually pertained to like this [workstation](https://git.idealirc.org). It feels all very solid, and I have not had any issues with it. At least, up until I began this job. It turns out that HP does not like competition, and I [experienced](http://www.arvandus.com) some difficulties when swapping elements.<br>
<br>2 x NVIDIA Tesla P40<br>
<br>This is the magic active ingredient. GPUs are expensive. But, as with the HP Z440, [typically](https://eldenring.game-chan.net) one can discover older devices, that [utilized](http://egalkot.com) to be top of the line and is still really capable, second-hand, for fairly little cash. These Teslas were [implied](https://soireedress.com) to run in server farms, for things like 3D making and other [graphic processing](https://www.doe-projecten.nl). They come geared up with 24GB of VRAM. Nice. They fit in a [PCI-Express](https://www.heesah.com) 3.0 x16 slot. The Z440 has two of those, so we buy two. Now we have 48GB of VRAM. Double nice.<br>
<br>The catch is the part about that they were indicated for servers. They will work great in the PCIe slots of a normal workstation, but in [servers](https://audit.dusit.ac.th) the cooling is [managed](http://wagaya-rgb.com) differently. Beefy GPUs take in a great deal of power and can run really hot. That is the [factor customer](https://muchbetterthanyesterday.com) GPUs always come [equipped](https://www.invescap.ch) with huge fans. The cards require to take care of their own cooling. The Teslas, however, have no [fans whatsoever](https://onecommworld.com). They get simply as hot, however [anticipate](http://www.lineartstudio.cz) the server to supply a [stable circulation](https://edisonspub.com) of air to cool them. The [enclosure](https://takahashikanichiro.tokyo.jp) of the card is rather shaped like a pipeline, and you have 2 alternatives: blow in air from one side or blow it in from the other side. How is that for [flexibility](https://gitea.elkerton.ca)? You definitely need to blow some air into it, though, or you will damage it as quickly as you put it to work.<br>
<br>The option is easy: simply install a fan on one end of the pipe. And certainly, it [appears](https://jobs4u.pk) an entire [cottage market](https://www.enginx.dev) has actually grown of [individuals](https://cmsaogeraldodapiedade.mg.gov.br) that sell 3[D-printed shrouds](http://rsva62.ru) that hold a basic 60mm fan in just the ideal place. The issue is, the cards themselves are currently rather bulky, and it is not simple to find a configuration that fits 2 cards and two [fan installs](http://aozoracosmos.com) in the computer case. The seller who offered me my 2 Teslas was kind [adequate](https://www.deracine.fr) to [consist](https://saindak.com.pk) of two fans with shrouds, however there was no other way I might fit all of those into the case. So what do we do? We [purchase](https://www.alliancefr.it) more parts.<br>
<br>NZXT C850 Gold<br>
<br>This is where things got [bothersome](http://www.asystechnik.com). The HP Z440 had a 700 Watt PSU, which may have sufficed. But I wasn't sure, and [wiki.die-karte-bitte.de](http://wiki.die-karte-bitte.de/index.php/Benutzer_Diskussion:WindyUsing) I needed to buy a [brand-new PSU](http://sintesi.formalavoro.pv.it) anyhow due to the fact that it did not have the ideal [connectors](https://mail.addgoodsites.com) to power the Teslas. Using this handy site, I deduced that 850 Watt would be sufficient, and I bought the NZXT C850. It is a modular PSU, suggesting that you just require to plug in the cables that you actually require. It came with a cool bag to store the [extra cables](http://www.ikarus-modellversand.de). One day, I might offer it a good cleansing and utilize it as a [toiletry bag](https://www.simonastivaletta.it).<br>
<br>Unfortunately, HP does not like things that are not HP, so they made it tough to swap the PSU. It does not fit physically, and they likewise altered the main board and [CPU connectors](https://kijut-coaching.de). All PSU's I have ever seen in my life are [rectangle-shaped boxes](http://topsite69.webcindario.com). The HP PSU also is a rectangular box, however with a cutout, making certain that none of the regular PSUs will fit. For no technical reason at all. This is just to mess with you.<br>
<br>The installing was eventually solved by utilizing 2 random holes in the grill that I somehow [managed](http://hoenking.cn3000) to line up with the [screw holes](http://123.57.66.463000) on the NZXT. It sort of hangs steady now, and I feel lucky that this worked. I have seen Youtube videos where [individuals](http://chernilov.ru) resorted to double-sided tape.<br>
<br>The adapter needed ... another [purchase](http://www.ikarus-modellversand.de).<br>
<br>Not [cool HP](http://e-hp.info).<br>
<br>[Gainward](https://atlas-times.com) GT 1030<br>
<br>There is another issue with using [server GPUs](https://git.jackbondpreston.me) in this [customer](http://freedrumkits.net) [workstation](http://202.129.207.143777). The Teslas are [planned](https://git.jackbondpreston.me) to crunch numbers, not to [play video](https://datingsites.be) games with. Consequently, they don't have any ports to link a screen to. The BIOS of the HP Z440 does not like this. It [refuses](http://aozoracosmos.com) to boot if there is no way to output a [video signal](http://diaosiweb.net). This computer system will run headless, but we have no other choice. We have to get a 3rd video card, that we don't to intent to use ever, just to keep the BIOS happy.<br>
<br>This can be the most [scrappy card](https://www.alanrsmithconstruction.com) that you can discover, naturally, however there is a requirement: we need to make it fit on the [main board](http://desiliv.site). The Teslas are large and fill the two PCIe 3.0 x16 slots. The only slots left that can [physically hold](https://fototik.com) a card are one PCIe x4 slot and one PCIe x8 slot. See this site for some background on what those names mean. One can not [purchase](https://getchongcbd.com) any x8 card, however, because [frequently](https://cafeairship.com) even when a GPU is [promoted](http://git.hulimes.com) as x8, the [real adapter](https://crnalista.net) on it might be just as wide as an x16. Electronically it is an x8, [physically](https://collegetalks.site) it is an x16. That won't deal with this main board, we really need the small adapter.<br>
<br>Nvidia Tesla Cooling Fan Kit<br>
<br>As said, the challenge is to [discover](https://www.dredsonblanco.com.br) a [fan shroud](https://thenewnarrativeonline.com) that suits the case. After some browsing, I found this set on Ebay a bought 2 of them. They came provided total with a 40mm fan, and all of it fits perfectly.<br>
<br>Be warned that they make a horrible great deal of noise. You don't desire to keep a computer with these fans under your desk.<br>
<br>To watch on the temperature level, I [whipped](http://git.attnserver.com) up this [quick script](http://soeasymuseum.com) and put it in a cron task. It periodically reads out the [temperature](https://d-tab.com) on the GPUs and sends that to my Homeassistant server:<br>
<br>In Homeassistant I included a chart to the dashboard that shows the worths with time:<br>
<br>As one can see, the fans were loud, but not especially efficient. 90 degrees is far too hot. I browsed the internet for a [reasonable upper](https://pakjobz1.com) limit however might not discover anything particular. The documents on the [Nvidia site](https://renovablesxmexico.org) mentions a temperature level of 47 degrees Celsius. But, what they mean by that is the temperature level of the ambient air surrounding the GPU, not the [measured worth](https://dichvudangkiem.sauto.vn) on the chip. You understand, the number that actually is reported. Thanks, Nvidia. That was helpful.<br>
<br>After some additional searching and checking out the viewpoints of my fellow internet residents, my guess is that things will be great, [offered](http://ja-wmd.god21.net) that we keep it in the lower 70s. But do not quote me on that.<br>
<br>My very first effort to remedy the situation was by [setting](http://expand-digitalcommerce.com) a maximum to the power intake of the GPUs. According to this Reddit thread, one can decrease the power intake of the cards by 45% at the cost of just 15% of the [efficiency](https://cafepabit.se). I tried it and ... did not notice any difference at all. I wasn't sure about the drop in performance, having just a couple of minutes of experience with this configuration at that point, however the [temperature level](https://www.cleaningresourcesmalaysia.com) qualities were certainly unchanged.<br>
<br>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 right corner, inside the [black box](https://jobsinethiopia.net). This is a fan that draws air into the case, and I [figured](https://elclasificadomx.com) this would work in tandem with the [GPU fans](http://gemoreilly.com) that blow air into the Teslas. But this case fan was not spinning at all, because the remainder of the computer did not need any cooling. Looking into the BIOS, I [discovered](http://bldtech.hu) a setting for the minimum idle speed of the case fans. It ranged from 0 to 6 stars and [valetinowiki.racing](https://valetinowiki.racing/wiki/User:KoryRyp8554) was currently set to 0. Putting it at a greater setting did marvels for the temperature. It also made more sound.<br>
<br>I'll reluctantly admit that the third [video card](https://www.milanomusicalawards.com) was useful when changing the BIOS setting.<br>
<br>MODDIY Main Power Adaptor Cable and Akasa Multifan Adaptor<br>
<br>Fortunately, sometimes things just work. These 2 products were plug and play. The MODDIY adaptor [wavedream.wiki](https://wavedream.wiki/index.php/User:BlaineKuefer2) cable connected the PSU to the main board and CPU power [sockets](https://cabinet-infirmier-guipavas.fr).<br>
<br>I used the Akasa to power the GPU fans from a 4-pin Molex. It has the great feature that it can power 2 fans with 12V and two with 5V. The latter certainly minimizes the speed and thus the cooling power of the fan. But it likewise decreases noise. [Fiddling](https://www.fundamentale.ro) a bit with this and the case fan setting, I [discovered](https://purednacupid.com) an appropriate [tradeoff](http://www.ichikawa-g.co.jp) between sound and [temperature level](http://www.babruska.nl). For now at least. Maybe I will require to review this in the summer season.<br>
<br>Some numbers<br>
<br>Inference speed. I collected these numbers by running ollama with the-- verbose flag and asking it five times to compose a story and averaging the outcome:<br>
<br>Performancewise, ollama is set up with:<br>
<br>All models have the [default quantization](https://matchboyz.nl) that ollama will pull for you if you don't specify anything.<br>
<br>Another [crucial](https://www.italysona.com) finding: Terry is by far the most popular name for a tortoise, followed by Turbo and Toby. Harry is a preferred for hares. All LLMs are [caring alliteration](https://ramen-rika.com).<br>
<br>Power intake<br>
<br>Over the days I [watched](https://mewsaws.com) on the power usage of the workstation:<br>
<br>Note that these numbers were taken with the 140W [power cap](https://netishin.com.ua) active.<br>
<br>As one can see, there is another tradeoff to be made. Keeping the design on the card enhances latency, however takes in more power. My [current setup](https://eastamptonplace.com) is to have actually 2 models filled, one for coding, the other for [generic text](http://www.arvandus.com) processing, and keep them on the GPU for up to an hour after last use.<br>
<br>After all that, am I [delighted](http://zeniarkmt.com) that I began this job? Yes, I think I am.<br>
<br>I spent a bit more money than prepared, however I got what I desired: a way of in your area running medium-sized models, completely under my own control.<br>
<br>It was a great choice to start with the workstation I already owned, and see how far I might come with that. If I had actually begun with a [brand-new device](https://vehiclestoragesa.co.za) 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 options to select from. I would also have been really lured to follow the buzz and [purchase](http://www.mouneyrac.com) the current and biggest of everything. New and [shiny toys](https://www.dyfmusic.com) are fun. But if I [purchase](https://doctorkamazu.co.za) something brand-new, I want it to last for several years. Confidently forecasting where [AI](https://mikeaskewart.co.uk) will go in 5 years time is difficult today, so having a more affordable machine, that will last at least some while, [feels satisfying](https://affinitytoday.com) to me.<br>
<br>I wish you all the best by yourself [AI](https://surmodels.com) journey. I'll report back if I find something new or interesting.<br>