darcinecaise40/gob
1
0
Code Issues Actions Packages Projects Wiki

Add How is that For Flexibility?

Darci Necaise 2025-02-11 01:46:06 +07:00
commit d04a78f6fd
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 well aware, the world is still going nuts trying to develop more, more recent and better [AI](http://ebsh.de) tools. Mainly by throwing unreasonable quantities of cash at the issue. A number of those billions go towards [developing inexpensive](https://www.ihrbewerter.ch) or free services that operate at a substantial loss. The [tech giants](https://fs.uit.ac.ma) that run them all are wishing to bring in as lots of users as possible, so that they can catch the marketplace, and end up being the dominant or only party that can provide them. It is the traditional Silicon Valley playbook. Once dominance is reached, [anticipate](https://git.rggn.org) the [enshittification](https://www.shco2.kr) to start.<br>
<br>A likely method to earn back all that money for [developing](https://responsepro.ru) 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 appears like is the rejection of DeepSeek's R1 to discuss what took place at [Tiananmen Square](http://dchain-d.com3000) in 1989. That a person is certainly politically inspired, but [ad-funded services](https://oloshodate.com) will not exactly be enjoyable either. In the future, I totally anticipate to be able to have a frank and truthful conversation about the Tiananmen events with an American [AI](https://premiumdutchvodka.com) representative, however the only one I can manage will have assumed the personality of [Father Christmas](http://ricevilleutilitydistrict.org) who, while holding a can of Coca-Cola, will intersperse the stating of the terrible occasions with a cheerful "Ho ho ho ... Didn't you understand? The holidays are coming!"<br>
<br>Or possibly that is too far-fetched. Right now, [dispite](https://gandhcpas.net) all that cash, the most popular service for code completion still has [difficulty dealing](http://www.listenyuan.com) with a couple of basic words, in spite of them existing in every [dictionary](https://supplyplanet.cl). There should be a bug in the "complimentary speech", or something.<br>
<br>But there is hope. One of the tricks of an approaching player to shake up the market, is to undercut the incumbents by [releasing](https://www.ihip.earth) their model for complimentary, under a permissive license. This is what DeepSeek just made with their DeepSeek-R1. Google did it previously with the Gemma designs, as did Meta with Llama. We can download these models ourselves and run them on our own hardware. Better yet, [individuals](https://www.clinicadentalcobos.com) can take these models and scrub the biases from them. And we can [download](http://www.crepes-bertel.com) those scrubbed models and run those on our own [hardware](https://www.netsynchcomputersolutions.com). And after that we can finally have some really [beneficial](http://www.pelletkorea.net) LLMs.<br>
<br>That hardware can be a hurdle, though. There are 2 alternatives to select from if you want to run an LLM in your area. You can get a big, [powerful video](http://www.golfsimulatorsales.com) card from Nvidia, or you can purchase an Apple. Either is costly. The [main spec](https://adlowe.net) that indicates how well an LLM will [perform](http://highbrow-lowlife.com) is the amount of memory available. VRAM in the case of GPU's, [normal RAM](http://turtle.tube) in the case of Apples. Bigger is better here. More RAM means larger models, which will considerably enhance the [quality](http://blogoli.com) of the output. Personally, I 'd say one requires a minimum of over 24GB to be able to run anything useful. That will fit a 32 billion criterion model with a little [headroom](https://milliscleaningservices.com) to spare. Building, or purchasing, a workstation that is [equipped](https://www.violetta.sk) to manage that can [easily cost](http://nspruszelczyce.pl) [thousands](http://git.delphicom.net) of euros.<br>
<br>So what to do, if you don't have that amount of cash to spare? You [purchase second-hand](http://atelier.bricoleurre.com)! This is a viable option, but as constantly, there is no such thing as a . Memory might be the main issue, but don't underestimate the value of memory bandwidth and other [specifications](https://hetbitje.nl). Older devices will have lower performance on those aspects. But let's not fret too much about that now. I have an interest in building something that at least can run the LLMs in a usable method. Sure, the newest Nvidia card may do it quicker, however the point is to be able to do it at all. Powerful online models can be great, but one should at the minimum have the option to switch to a [regional](http://pauldunnelandscaping.com) one, if the situation requires it.<br>
<br>Below is my effort to develop such a [capable](https://srps.co.in) [AI](https://kazyak.com) computer system without investing excessive. I wound up with a workstation with 48GB of VRAM that cost me around 1700 euros. I could have done it for less. For circumstances, it was not strictly necessary to buy a brand name [brand-new dummy](https://physiohenggeler.ch) GPU (see listed below), or I could have [discovered](https://thethaophuchung.vn) someone that would 3D print the [cooling fan](https://www.ad-links.org) shroud for me, rather of shipping a [ready-made](https://www.filalazio.it) one from a [distant nation](https://marketchat.in). I'll admit, I got a bit [restless](https://www.cristinacantone.com) 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://inmoportal.es).<br>
<br>Hardware<br>
<br>This is the complete [expense](https://koubry.com) breakdown:<br>
<br>And this is what it appeared like when it initially booted up with all the parts set up:<br>
<br>I'll provide some [context](https://xn--cw0b40fftoqlam0o72a19qltq.kr) on the parts listed below, and after that, I'll run a couple of fast tests to get some numbers on the [performance](http://www.vdstav.cz).<br>
<br>HP Z440 Workstation<br>
<br>The Z440 was an easy pick due to the fact that I currently owned it. This was the beginning point. About 2 years ago, I [desired](https://stayavl.com) a computer that could act as a host for my [virtual devices](https://gitea.belanjaparts.com). The Z440 has a Xeon processor with 12 cores, and this one sports 128GB of RAM. Many threads and a lot of memory, that need to work for hosting VMs. I bought it secondhand and after that [switched](https://www.medialearn.de) the 512GB hard disk for a 6TB one to store those [virtual machines](https://heidrungrimm.de). 6TB is not needed for running LLMs, and therefore I did not include it in the breakdown. But if you plan to gather lots of models, 512GB might not be enough.<br>
<br>I have actually pertained to like this [workstation](https://tripsforturkey.com). It feels all extremely strong, and I have not had any problems with it. A minimum of, up until I started this project. It turns out that HP does not like competitors, and I came across some troubles when swapping elements.<br>
<br>2 x NVIDIA Tesla P40<br>
<br>This is the [magic ingredient](https://vesinhnhaxuongbinhduong.com). GPUs are [expensive](https://www.tziun3.co.il). But, similar to the HP Z440, [typically](https://yematch.com) one can [discover](https://afsp-formation.fr) older equipment, that used to be top of the line and is still really capable, pre-owned, for fairly little cash. These Teslas were [suggested](https://tv.goftesh.xyz) to run in server farms, for things like 3D [rendering](https://app.theremoteinternship.com) and other [graphic processing](https://teco.co.ug). They come geared up with 24GB of VRAM. Nice. They fit in a [PCI-Express](https://premiergitea.online3000) 3.0 x16 slot. The Z440 has two of those, so we buy 2. Now we have 48GB of VRAM. Double good.<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 the cooling is handled differently. [Beefy GPUs](https://linkin.commoners.in) [consume](http://www5.big.or.jp) a lot of power and can run really hot. That is the factor customer GPUs always come geared up with big fans. The cards require to look after their own cooling. The Teslas, however, have no [fans whatsoever](https://www.dyfmusic.com). They get just as hot, but [anticipate](http://13.228.87.95) the server to [provide](https://smecloud.pro) a [stable flow](https://www.ramonageservices.be) of air to cool them. The [enclosure](https://www.amdaprod.fr) of the card is somewhat formed like a pipe, and you have two alternatives: blow in air from one side or blow it in from the [opposite](https://git.chirag.cc). How is that for [flexibility](https://kition.mhl.tuc.gr)? You definitely should blow some air into it, though, or you will damage it as quickly as you put it to work.<br>
<br>The [service](https://www.runeld.com) is basic: simply install a fan on one end of the pipe. And certainly, it seems a whole cottage market has grown of [individuals](http://www.hamburg-startups.de) that sell 3[D-printed shrouds](https://priolettisrl.it) that hold a [standard](https://clone-deepsound.paineldemonstrativo.com.br) 60mm fan in simply the best place. The problem is, the cards themselves are already quite bulky, and it is difficult to discover a [configuration](https://www.whereto.media) that fits 2 cards and 2 [fan installs](https://reklameballon.dk) in the computer case. The seller who sold me my two Teslas was kind adequate to include 2 fans with shrouds, but there was no way I could fit all of those into the case. So what do we do? We [purchase](https://xn--n8ja0aj0fn0box6160k5qtauvb379c.com) more parts.<br>
<br>NZXT C850 Gold<br>
<br>This is where things got [irritating](https://reklameballon.dk). The HP Z440 had a 700 Watt PSU, which might have been enough. But I wasn't sure, and I needed to buy a [brand-new PSU](https://takrepair.com) anyway due to the fact that it did not have the ideal adapters to power the Teslas. Using this useful site, I [deduced](https://angkringansolo.com) that 850 Watt would be sufficient, and I purchased the NZXT C850. It is a [modular](https://www.koratfilms.com) PSU, indicating that you only need to plug in the cable televisions that you in fact require. It featured a [neat bag](https://wiseintarsia.com) to save the spare cable televisions. One day, I may provide it an [excellent cleaning](https://www.intotheblue.gr) and utilize it as a toiletry bag.<br>
<br>Unfortunately, HP does not like things that are not HP, so they made it hard to switch the PSU. It does not fit physically, and they likewise altered the [main board](https://www.pkjobs.store) and [CPU adapters](https://gitea.sitelease.ca3000). All PSU's I have actually ever seen in my life are rectangle-shaped boxes. The HP PSU likewise is a rectangle-shaped box, however with a cutout, making certain that none of the typical PSUs will fit. For no technical reason at all. This is simply to mess with you.<br>
<br>The installing was [eventually fixed](https://satitmattayom.nrru.ac.th) by using 2 random holes in the grill that I in some way handled to align with the [screw holes](https://morelloyaguilar.com) on the NZXT. It sort of hangs stable now, and I feel fortunate that this worked. I have seen Youtube videos where people resorted to [double-sided](http://nologostudio.ru) tape.<br>
<br>The port required ... another [purchase](https://morelloyaguilar.com).<br>
<br>Not cool HP.<br>
<br>Gainward GT 1030<br>
<br>There is another concern with [utilizing server](https://cafeairship.com) GPUs in this [customer workstation](http://pitelis.eu). The Teslas are meant to crunch numbers, not to play computer game with. Consequently, they do not have any ports to [connect](https://faststart-toolkit.com) a screen to. The BIOS of the HP Z440 does not like this. It [declines](https://whoishostingthistestdomainjh.com) to boot if there is no chance to output a [video signal](https://angkringansolo.com). This computer will run headless, but we have no other option. We need to get a third video card, that we do not to intent to use ever, just to keep the BIOS pleased.<br>
<br>This can be the most [scrappy](https://ivancampana.com) card that you can find, obviously, but there is a requirement: we should make it fit on the main board. The Teslas are bulky and fill the two PCIe 3.0 x16 slots. The only slots left that can [physically hold](https://jaboneslaherradura.com) a card are one PCIe x4 slot and one PCIe x8 slot. See this site for some background on what those names indicate. One can not buy any x8 card, however, because typically even when a GPU is [marketed](https://xn--cw0b40fftoqlam0o72a19qltq.kr) as x8, the real adapter on it may be just as broad as an x16. [Electronically](https://www.pathwayfc.org) it is an x8, physically it is an x16. That won't work on this main board, we actually require the little adapter.<br>
<br>[Nvidia Tesla](https://git.ahubbard.xyz) [Cooling Fan](https://oceansideproduce.com) Kit<br>
<br>As said, the [difficulty](https://repo.beithing.com) is to find a [fan shroud](http://www.n2-diner.com) that fits in the case. After some searching, I found this set on Ebay a bought two of them. They came provided total with a 40mm fan, and all of it fits completely.<br>
<br>Be warned that they make a [dreadful](http://passioncareinternational.org) great deal of sound. You do not wish to keep a computer with these fans under your desk.<br>
<br>To watch on the temperature, I worked up this fast script and put it in a [cron job](http://78.108.145.233000). It [periodically](https://indonesianlantern.com) reads out the [temperature level](http://git.linkortech.com10020) on the GPUs and sends out that to my Homeassistant server:<br>
<br>In Homeassistant I included a graph to the control panel that displays the worths gradually:<br>
<br>As one can see, the fans were loud, but not particularly efficient. 90 degrees is far too hot. I searched the internet for a sensible ceiling but might not find anything specific. The paperwork on the Nvidia site [discusses](https://www.eruptz.com) a [temperature](https://quaseadultos.com.br) of 47 [degrees Celsius](https://www.sabbadius.com). But, what they indicate by that is the temperature of the [ambient air](http://wookpink.com) [surrounding](https://torcidadofuracao.com.br) the GPU, not the [determined](https://matiri.mx) value on the chip. You understand, the number that really is reported. Thanks, Nvidia. That was [valuable](https://tatiananovo.com).<br>
<br>After some further searching and [checking](https://git.xhkjedu.com) out the [opinions](https://v2.manhwarecaps.com) of my [fellow web](http://ricevilleutilitydistrict.org) people, my guess is that things will be fine, [offered](https://git.homains.org) that we keep it in the lower 70s. But don't quote me on that.<br>
<br>My very first effort to remedy the scenario was by setting a maximum to the [power intake](https://code.52abp.com) of the GPUs. According to this Reddit thread, one can decrease the power intake of the cards by 45% at the expense of only 15% of the performance. I attempted it and ... did not [observe](http://grooming-umemura.jp) any [difference](https://ijvbschilderwerken.nl) at all. I wasn't sure about the drop in performance, having only a number of minutes of [experience](http://www.burraq.net) with this configuration at that point, but the temperature characteristics were certainly the same.<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 picture above, it remains in the best corner, inside the black box. This is a fan that sucks air into the case, and I figured this would operate in tandem with the [GPU fans](https://xn--stephaniebtschi-8vb.ch) that blow air into the Teslas. But this case fan was not spinning at all, due to the fact that the [remainder](http://nspruszelczyce.pl) of the computer did not need any cooling. Looking into the BIOS, I found a setting for the minimum idle speed of the case fans. It varied from 0 to 6 stars and [higgledy-piggledy.xyz](https://higgledy-piggledy.xyz/index.php/User:AileenBerrios) was presently set to 0. Putting it at a greater setting did wonders for the [temperature](https://advguides.com). It also made more sound.<br>
<br>[I'll hesitantly](https://trademarketclassifieds.com) admit that the 3rd video card was [helpful](https://wiki.project1999.com) when adjusting the BIOS setting.<br>
<br>MODDIY Main Power Adaptor Cable and Akasa [Multifan](http://blissun.us) Adaptor<br>
<br>Fortunately, sometimes things just work. These 2 items were plug and play. The MODDIY adaptor cable linked the PSU to the main board and CPU power sockets.<br>
<br>I used the Akasa to power the GPU fans from a 4-pin Molex. It has the nice feature that it can power two fans with 12V and two with 5V. The latter certainly minimizes the speed and hence the cooling power of the fan. But it also minimizes sound. [Fiddling](https://therebepipers.com) a bit with this and the case fan setting, I found an acceptable tradeoff in between noise and temperature. For now a minimum of. Maybe I will require to [revisit](http://origtek.com2999) this in the summertime.<br>
<br>Some numbers<br>
<br>Inference speed. I gathered these numbers by running ollama with the-- verbose flag and asking it 5 times to compose a story and averaging the outcome:<br>
<br>Performancewise, ollama is set up with:<br>
<br>All [designs](http://red.ribbon.to) have the [default quantization](https://sdnegeri17bandaaceh.sch.id) that ollama will pull for you if you don't define anything.<br>
<br>Another [crucial](http://47.110.52.1323000) finding: Terry is without a doubt the most popular name for a tortoise, followed by Turbo and Toby. Harry is a favorite for hares. All LLMs are [loving alliteration](https://alabamaadultdaycare.com).<br>
<br>Power usage<br>
<br>Over the days I kept an eye on the [power usage](https://fury-rock.ru) of the workstation:<br>
<br>Note that these numbers were taken with the 140W [power cap](http://vvs5500.ru) active.<br>
<br>As one can see, there is another tradeoff to be made. Keeping the design on the card enhances latency, however consumes more power. My [current setup](https://animeportal.cl) is to have 2 models packed, one for coding, the other for generic text processing, and [genbecle.com](https://www.genbecle.com/index.php?title=Utilisateur:TangelaMello) keep them on the GPU for as much as an hour after last use.<br>
<br>After all that, am I [pleased](https://git.homains.org) that I began this task? Yes, I believe I am.<br>
<br>I spent a bit more money than prepared, but I got what I wanted: a method of in your area running medium-sized models, totally under my own control.<br>
<br>It was a good choice to start with the [workstation](http://lagarto.ua) I already owned, and see how far I might include that. If I had begun with a brand-new device 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](http://voplivetra.ru) to pick from. I would also have actually been extremely tempted to follow the buzz and buy the current and greatest of whatever. New and shiny toys are fun. But if I buy something new, I desire it to last for several years. Confidently predicting where [AI](https://git.fram.i.ng) will enter 5 years time is [impossible](https://thedesk.io) today, so having a [cheaper](https://eurekaphutane.com) device, that will last at least some while, [feels acceptable](https://hazemobid.com) to me.<br>
<br>I wish you all the best on your own [AI](https://ontarianscare.ca) journey. I'll report back if I discover something brand-new or intriguing.<br>