Search the Community

Budget (including currency): 130K INR (about $1500 USD) Country: India Games, programs or workloads that it will be used for: Deeplearning, Training Neural Networks, Video rendering,3D rendering , Minecraft, CS:GO, and general stuff Other details My PC will run Windows 11 as it's OS. I chose some parts for my build. Please check this out at https://pcpartpicker.com/list/ZPTJ28 I chose Intel i7 for it was under my budget and a decent CPU for my workload. The main for my workload is GPU, so I chose GeForce RTX 3060 as it has some good amount of CUDA cores along with 12GB VRAM. I am going to use a 4k monitor though display resoulutions don't matter for me. Please suggest changes required.

Budget (including currency): $1,500 Country: USA Games, programs or workloads that it will be used for: AI/ML model training, general computer science related tasks, obviously gaming Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): I've never built a PC before, even though I have watched countless videos on how to, and I am looking to get something that will set me up for a while. For some context, I'm currently a second-year comp-sci student and will be specializing in Machine Learning and AI. I haven't had too many courses specifically related to those fields yet so I'm really unsure what kind of hardware works best for it. The only PC I have as of right now is a Dell XPS 13 with an i5-1135G7 and 16 GB of RAM. I have no complaints about this laptop, but as of yet, I haven't really used it for demanding tasks and when compiling C++ code or even trying to play Minecraft, it definitely struggles. As a little side note, I am very deep in the Apple ecosystem (iPhone, iPad, Apple Watch, Apple TV, Airpods, and Homepod Mini) and I'm also not opposed to getting something from Apple, with the acknowledgement that the gaming performance will suffer. I have a PS5 and as much as I'd love a PC to play much better games, it is not my biggest worry. Any advice would be much appreciated. I'm looking to buy this system at any point in the next year. Thanks!

Budget: ~$1800 total Country: USA (NY) Games, programs or workloads that it will be used for: Fortnite, BG3, Minecraft (hosting server), random AAA shooters, Hogwarts Legacy | NUTS simulations, model development and local training, local LLM development/deployment Other details: pc part picker list, getting an m.2 at some point on sale (gen 5 worth?), getting a 34in monitor, upgrading from ASUS Scar Strix II GL504GS (gtx 1070, i7-8750h, 32g ddr4), would like to play 1440p 144hz. Hello everyone, I am building my first real PC after living the gaming laptop life for 6-ish years. I included my pc-partpicker list above, but my only real concerns are with the 7800x3d. I know it's a fantastic gaming chip, and I know it's a massive upgrade from what I have (plus the AM5 lifespan), but would a 14700k or 7900x do me better here for the same price considering my use cases? The split will probably be 65/35 gaming to work, but that work is really intensive. I have heard things about RAM instability with AMD chips, will I have issues like that with my setup here? Also, I know the 8000 series is coming out soon, should that have any effect on my choices here? Any advice would be greatly appreciated. Thanks!

Budget (including currency): ±R30000 (±$1600) Country: South Africa Games, programs or workloads that it will be used for: (Linux) Development, Gaming if possible Other details: I want to buy a laptop I will dual boot with Windows and Ubuntu for remote work (without a monitor in some situations so bigger screen is better). Any recommendations will be greatly appreciated. I have always been a Mac user... I REQUIRE an NVIDIA graphics card for Machine Learning. Thanks!

macOS has a built-in AI Image Remover function on Apple Silicon powered Macs (right click on an image file -> Quick Actions -> Remove background). However, it does a pretty terrible job with hair details. On the other hand, we have the website remove.bg that does a fantastic job of removing backgrounds flawlessly while keeping hair details, but it’s insanely expensive when you have hundreds of images. Is anyone aware of an AI/Machine Learning background remover for Mac with batch processing that has a perpetual license and cost less than $100 USD? Alternatively an annual subscription ≈ 30 dollars? UPDATE: I tried “AI Batch Background Eraser” by Ivan Blayasoft from the Mac App Store but it won’t output an image for some reason?

Does anyone have any info about benchmarks on ML workloads, such as widely used model (i'm mostly interested in LLMs) inference speed for different configurations? What would be CPU inference speed on something like Xeon-W 2400? Or threadripper? Assuming any of those has comparable amount of memory to possible maxed-out M2 ultra config. How it compares against 2x or 4x 3090s/4090s? How all above compares against maxed out M1/M2 mac studio? How all above will cope with loading 30 (or even 60-70) billion parameter models? How they will work with 4-int/8-int/16-float models? And diffent inference engines (huggingface transformers/vLLM/auto-GPTQ/llama-cpp/GGLM) Aren't M1/M2 macs are actual secret ML monsters with their huge unified memory? And taking into account price of 3090/4090, M1/M2 macs can actually be a cost-effective (!) measure for this task. Here are my calculations : If I get 4x 4090, assuming each cost 1600 bucks, it's 6400. I will get 96 gigs of v-ram. And it's gpu's alone, not even talking about rest of the system. And inference speed might suffer a bit if I have to run model in distributed mode (I gotta do it even if I load llama-2 70b in int-4, for instance) If I configure m2-ultra mac to have 76 GPU cores and 192 gigs of unified memory, it will cost me .... 6600 dollars. For a bit more than 2 times of gpu-accessible ram. I need benchmarks. Does anyone have any? Mac's looking more cost-effective than PC is peak clown-world. And the funniest part, it might be a reality.

Budget : £1000 Country: United Kingdom Games, programs or workloads that it will be used for: Valorant with some screen recording and light ML training and coding for personal projects( planning to dual boot with Ubuntu). Specs chosen: AMD Ryzen 5 7600x, Nvidia 4070 12GB ( MSI Ventus), MSI B650P pro WiFi motherboard, Corsair vengeance 32GB RAM (6000 Mhz CL36), Corsair RM 850, Corsair 4000D airflow Other details : I've a 240 Hz 1080p monitor, G pro Superlight and a decent mechanical keyboard. Doubts : is the 4070 an overkill? But downgrading my GPU makes sense if I downgrade my CPU too right? Also not sure about the PSU as well. I also need a cooler for the CPU, which would be best in a budget sense?

Hi All, I have a question for a dual GPU setup for a ubuntu machine learning setup. I have an RTX3090 and it is working great in my setup. I would like to add another gpu and I was thinking of the A6000 actually. I know you can't easily mix gpus but they both use the same driver when I checked (https://www.nvidia.com/Download/driverResults.aspx/209261/en-us/). Is there any reason not to try this?

Budget (including currency): 6500 USD / 70.000 NOK incl VAT Country: Norway Games, programs or workloads that it will be used for: Machine learning, simulations, robotics Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): Starting some early planning for a build that will happen either to the end of 2023 or q1/q2 2024. My research adviser has said to avoid workstation graphic cards from the Quadro lineup and use GTX 30 or 40 series (he is rocking a 3090) I dont´t think a Threadripper class of workstation is applicable here. Nor is the budget high enough. But I have been looking at 7950 X3D 4090 128 GB DDR5 6000MT/s 1200W Corsair PSU Any comments or other recommendations? If these parts were to be selected, what manufacturers would be preferable, now that EVGA no longer makes GPUs? MSI motherboard and GPU? ASUS? Gigabyte? What would you choose and why?

Been wanting to have a PC since ages and I think it's time finally. I need help with the parts though. Budget (including currency): Considering a budget of around £1500 (in the UK). Willing to consider going above the budget for better performance. Country: United Kingdom. Games, programs or workloads that it will be used for: Valorant ( might screen in the future?) Machine Learning models for personal projects and freelancing stuffs. I do have access to a powerful server for complex ML model training and do not require high VRAM. Other details : I've a 1080p 144Hz monitor which I plan to upgrade or use as my second monitor soon. ( Don't know about higher resolutions as I only play Valorant) I've been thinking about the Nvidia 4070 or should I look out for used 3090s instead? Don't want to go AMD for GPU as the stability and support for my ML modules are relatively less and support for Nvidia is available throughout the community. For the CPU, the majority of my work needs high single core performance ( bonus for gaming too IG). So Intel is the way to go then? I need a new Headset and mouse though (wireless maybe? Fairly happy with my keyboard even though it's wired.). I've a MacBook air which I'd ideally like to connect and work on some casual stuffs. (not a necessity though) Currently using a my 8 Year old GTX 1050 laptop which is on its last leg.

So I got a new GPU and since I seem to run into power limitations pretty often, I will also have to get a new PSU ig. I still got my old GPU and I was thinking, if it is possible to take both GPUs for machine learning tasks? I wouldnt need to use both for display output or gaming or anything like that, just ML would be nice to be run on both. for reference, the GPUs are a 2060 and a 4070. My motherboard should be able to handle the bandwith, as its datasheet shows 2 x16 PCIe slots (for reference its this https://www.msi.com/Motherboard/MPG-X570-GAMING-PRO-CARBON-WIFI/Specification). Now if it works, the question is what PSU to get. Since the GPUs both have around 200W, my cpu needs at most 65W but lets say also 100W, and the rest of the pc takes another idk, 50-100W, I would say a normal 650W PSU as per specs for the 4070 would be enough. However, I am very unsure about that. The question is how high the GPUs spike. Right now I have a 550W PSU (cannot recommend, I permanently see it hitting a voltage or power limit with HWMonitor and never get 100% GPU power, rn the max is 95%) and that needs to change. So my question basically is: 650W PSU, 750W PSU or "futureproof" and get even more?

Budget (including currency): ~1000 Euro without GPU Country: germany Games, programs or workloads that it will be used for: Gaming (Overwatch rn, but might also be stuff like forza horizon in the future) & Machine Learning Other details (existing parts lists, whether any peripherals are needed, what you're upgrading from, when you're going to buy, what resolution and refresh rate you want to play at, etc): So I am already more or less set on a few components, I am just wondering if there are any inputs on mainly the motherboard and power supply. I looked it up and it seems like right now the 13600K is the best value (integrated graphics because those 30 bucks extra are worth the hours of debugging it can save...) and id like to upgrade to DDR5 to be able to still game modern games in a few years time without losing much speed. The Motherboard has a few requirements tho. FIrstly, it needs two full on GPU slots. I wont fill them right now, but I want the option to do so once I train bigger models. Secondly, I probably need at least 2 NVME slots - more wouldnt hurt. ML needs a lot of space. I could also use SSDs, but honestly I dont really see much of a price advantage at this point (do correct me if I am wrong tho lol). For now I only need a new 2TB drive. Ram... idk? Probably 64GB? Which ones tho? Power supply: I was thinking a 1000W titanium power supply. Mainly because with ML stuff runs quite long and once you run stuff over night, you do not want to lose 20% of the energy. I was thinking Seasonic, but maybe there are more bang for the buck ones? I dont want low quality ones tho. The case should be the Corsair 4000D Airflow. However, I was wondering if that one also exists without glass panels but with some metal stuff? and are fans always included there? If yes, id use those. If not, id use the old ones from my pc rn. As for monitors, any recommendations for 1440p monitors that can be used in a 3 monitor configuration? They are not included in the budget, I will probably get them over time, but they should still not be too expensive and have at least 144Hz... So, I hope this was not too much of a tall order, if needed idm being over budget by a little, do argue as to why tho. I only saw one other Machine learning build and that was purely for ML and on an older technology, id like to at least have the option for DDR5. If you have any youtube videos that target this combination, feel free to share them! I have really not seen many videos that target such a build (or say why its a bad idea). Thanks in advance!

Hello everyone, I am looking for a new laptop to buy around the 1400 euro range (+-200) which I will use mainly for work (as a data scientist) but also want the possibility to game on it (not games with the highest requirements ofc). Do you have any suggestions? I have been looking at things like the Inspiron 16 Plus Laptop (Intel® Core™ i7-11800H, 32GB RAM, GeForce RTX™ 3060), or the ASUS Vivobook Pro 16X OLED (AMD Ryzen™ 7 5800H, 16GB RAM, GeForce® RTX 3050 Ti), however I am out of the loop as to which CPU's/GPU's are better in what context so feel free to suggest any other combinations (such as AMD GPUs) or laptops that you would think are best. Thank you very much for your suggestions PS. I live in Germany so some prices may differ from the US market.

Background: I'm a machine learning developer who runs a chatbot service on discord, and does research with reformer models (a variant of the transformer model). I recently decided that I would really like to upgrade my service backends from entirely CPU compute to GPU compute, which (as you might imagine) is extremely expensive on hosting providers. AWS charges $657 per month to use half of a K80, and GCP charges $328.5 per month for the same service. (GCP charges extra for the virtual machine it runs in, so prices may vary). For someone like me, who doesn't make much off of my service, either of those prices is completely unacceptable. So of course, the solution is to build your own rig. But woo-hoo! It's covid time and GPU prices have gone completely mental. After a bit of research, I noticed that the answer to my GPU woes was sitting in front of me the entire time. Tesla K80s were selling on ebay for $130 a pop (at their lowest early-covid. You'd be hard-pressed to find one for $180 nowadays) and that's for 2 GPUs on a single card That's right, cloud providers. I can build a rig just as good or better, for less than half your prices. Since this card is passively cooled, built for servers, and I figured I could make use of the remote mangement tools as well, I focused on building my new rig around a server chasis. Turns out both these choices were quite important in driving costs down. Here's my part list: HP ML350 gen 8 (Specs below) - $175 Pair of E5-2620 6c @2.0ghz 32gb DDR3 @1333 6 hard drive caddies (btw you can sell these for $6-$9 each lol) 2x 460w PSU(s) Tesla K80 - $180 https://www.ebay.com/sch/i.html?_from=R40&_nkw=tesla+k80&_sacat=0&_sop=15 the stupid cables for the two above items - $26 <- if you use ebay insead of amazon you can probablly find cheaper cables, I was in a hurry, and these worked. https://www.amazon.com/dp/B07SHT579T?psc=1&ref=ppx_yo2_dt_b_product_details https://www.amazon.com/dp/B0796NVNY7?psc=1&ref=ppx_yo2_dt_b_product_details (optional, but reccomended): 1 or 2 HP 750w PSU(s) - ~$15 (optional, it works just fine without them but reccomended if you plan to game on it) Pair of E5-2680 10c @2.7ghz- ~$40 (optional, some servers come preinstalled with storage) hard drives: - ~$12.5 <- 500gb drive (if you don't have one already) an 8gb+ SD card For the reccomended spec (assuming you have an SD card and 2.5" drive handy), that's $392 (before taxes) Before you go building this yourself, please realize that this is: 1. A server. most, if not all things in the server are proprietary and really lack the features of a workstation/gaming pc 2. Loud. At a (pretty cold) inlet temp of 65 degrees F, or 18 degrees C, this server will run its fans at 30% or higher, which will be ~45db of constant noise. 3. Going to need a lot of work. I won't take responsibiliy if you fry your server, since this will need some cable modding, and nothing in the HP servers is standard. Now that that the disclaimers are over with, let's go over some hardware pain points I first ran into. The HP ML350 Gen 8 is capable of handling 3 Gen 3 16x high power cards, but only has two power outputs for them on its power supply daugher board (the daughter board is hidden under the fan cage). The user manual lies about the pinout on the daughter board, with the pinout appearing to be a 10 pin, but actually, it's a 8-pin PCIe female connector (SEE FIG.1). BUT HOLD YOUR HORSES. That's also a lie. It's keyed like a PCIe 8-pin female, but the power it outputs is equivalent to a EPS 8-pin (e.g. 12v across the top and ground across the bottom) No one makes a cable to this pinout, not even HP. Not even if you wanted to pay $100 for it on amazon. The Tesla K80 actually wants this pin layout, but unfortunately, unless you know how to make your own cables, this isn't really an option. There is a really old thread about this cable debacle here, but they don't end up with a working solution, so I'll detail the possible solutions to this below. Fig. 1 - Connector on ML350p Gen 8 My solution was to buy a standard PCIe 8-pin male to 2x PCIe 8-pin male, and the standard Tesla K80 2x PCIe 8-pin female to 8-pin EPS male cable. Depending on the GPU you want to run, you may only need to buy and mod the first cable. If you do this configuration, DO NOT plug the two adapters together into the motherboard without modding them, as you WILL send 12 volts to ground, shorting and potentially kill your power supply(s). You'll want to cut the cheaper of the two cables, in case you end up giving up on this or use a different solution. If you cut the PCIe 8-pin male to 2x PCIe 8-pin male, cut cable 4, according to Fig. 2. If you cut the 2x PCIe 8-pin female to 8-pin EPS male, cut the two small loops coming out of the female connectors. (see Fig. 3) MAKE SURE TO PUT ELECTRICAL TAPE/HOT GLUE AROUND THE CUT LOOSE ENDS The alternative solution is to buy a cable like this: https://www.amazon.com/gp/product/B08R8821VZ/ref=ppx_yo_dt_b_asin_title_o00?ie=UTF8&psc=1. You won't need to mod it, but I highly advise against it. the last two ground pins on a PCIe 8-pin do provide power, roughly 75w each. Your card might not work at full capacity without those connected PSU-side. Fig. 2 - PCIe 8 pin standard Fig. 3 - Telsa K80 cable loops If you plug the cable(s) into the server now, it should boot and not short, without the card plugged in. ALWAYS CHECK THE POWER OUTPUTS OF YOUR CABLES BEFORE PLUGGING THEM INTO YOUR CARD. The outputs should match the EPS 8-pin standard (for the Tesla k80) or the standard defined by your graphics card manufacturer. Only once you've verified the output, should you plug it into your card and expect no magic smoke. I plugged the card into Lane 3 (the only X16 slot for CPU 1), and it worked just fine. You'll now want to use the GPU immedietely, I presume! But not so fast. This GPU has an eye-watering 24 GB of vram onboard, and systems aren't really designed for that. Luckily, there's a setting in HP's bios that can fix that ;3 On most motherboards, this would be a setting called "Above 4g Decoding", but for HP, this is not the case. Instead, it's in a hidden menu in the BIOS. You'll want to follow these steps: Boot up the server and press F9 when prompted to boot into BIOS Press CTRL+A on your keyboard attached to your server. A new menu called "SERVICE OPTIONS" will pop up Navigate to the new menu and go into it. Find the entry "PCI Express 64-BIT BAR Support" and enable it Exit the BIOS The K80 should now work. Credit where credit is due, and big thanks to RiDDO and his post on the HP forums here: https://community.hpe.com/t5/ProLiant-Servers-ML-DL-SL/ML350p-Gen8-amp-nVidia-Tesla-K40m/td-p/7045883#.YPERc-j0nrc If you have ILO setup on the server, you can find the card labelled as UNKNOWN under System Information>Device Inventory>PCI-E Slot 3, unless you have an HP-branded K80. ----------------------------------------------- Hardware setup over (aka half the battle) If you've made it this far in you're probably pretty intrugued about how in the world I plan on making this setup work as a gaming server, or want to build it yourself. After all, this entire time, I haven't really explained how I intend on using this... without any video outputs... (there isn't even a framebuffer on this graphics card.........) I my case, I use ESXi 7.0 (you can get a free trial here) to run virtual machines, one for a linux machine that runs hosting stuff with half the GPU, and a windows dev/gaming enviornment for my second half of the GPU. To install it, you can use rufus to make the ISO image a bootable USB, and install ESXi to an SD card the usual way (you can find the server's onboard SD card slot inside the server under the fans). You'll probably want to change some configuration options on ESXi, such as setting a static IP address. Once ESXi is installed, you'll want to toggle passthrough for these GPUs and reboot the server. For each virtual machine you create, make sure to edit their configuration settings with these flags: pciPassthru.use64bitMMIO=TRUE hypervisor.cpuid.v0=FALSE pciPassthru0.msiEnabled=FALSE An explanation: pciPassthru.use64bitMMIO is basically "Above 4g decoding" for your VM, hypervisor.cpuid.v0 makes your vm think it's running on physical hardware, and pciPassthru0.msiEnabled is a sanity thing that I added (idk if it's needed) The machine learning VM is my own code and enviorment, and so it's configuration isn't important to this post. I'm sure you all want to know how to game on windows on the server. Well lucky for me, someone else already created a tutorial for setting the GPU into WDDM mode, enabling you to game on the GPU. Follow the guide here if you're interested in that: https://www.reddit.com/r/pcmods/comments/nhfwh7/guide_using_an_nvidia_tesla_k80_datacenter_gpu/ As for me, I'll detail how to fix the bigger issue with this setup. There's no video output on this server, meaning that most low-latency video/input/audio streaming services won't work. You could use the output vmware gives you or windows remote desktop, but frankly, they're quite terrible in terms of latency and image quality. There aren't iGPUs in the CPUs either, since these are Xeons. Well, don't fear! You can use that K80 in WDDM mode to make parsec work, and for free (yep, don't pull out your hard-earned money to buy parsec teams). Essentially, what you need is a virtual display driver (just the driver) that fools windows into thinking there's a second display attached, when there actually isn't. After many hours of research and testing, only one worked for me. Follow this guide and install the display driver: https://www.amyuni.com/forum/viewtopic.php?t=3030. In the installer exes' directory, run "deviceinstaller enableidd 1" Once you're done, it should show up in windows' device manager as a USB mobile moitor (See Fig 4). Go into windows' display settings, and add parsec as a High-Power GPU application MAKE SURE TO HAVE PARSEC INSTALLED AND READY TO GO FIRST Set the virtual machine to only display on monitor 2. Log into your vm from the parsec client If you mess up and can't see your monitor anymore, reboot the vm from ESXi and run the comand "deviceinstaller enableidd 1" from the installer exes' directory. I reccomend not automating this process in case it causes issues or parsec happens to not work. Refer to Fig. 5 for a working example of it running. There are drawbacks to this method, of course. The driver only supports 1920x1080 output, you can't change the scaling or the resolution, even in parsec. If you can find a solution to this, PLEASE tell me Fig. 4 Fig. 5 ----------------------------------------------- Conclusion Overall, despite all the hurdles, this was quite the fun build to make, and it ended up being very worthwhile in the end. It's worth noting that there's a few drawbacks to the setup: You can't run too much stuff at the same time, since HP doesn't understand the Tesla K80's temperature The front half of the Tesla K80 is 15-20 degrees C cooler than the back half (and if either reaches 90 degrees C it will crash the entire server) - I game on the cooler GPU and underclock the card to keep a stable 66-70 C on it (Iif you know how to make HP's ILO run the fan curve with the GPU that would be great) Power usage is quite high. While gaming and machine learning (hosting), the server draws 410w-460w, and I estimate that the server can easily draw 650w if everything is running. (to be fair, the power costs are still lower than shilling out money to cloud providers) You might run into weird PCIe Passthrough issues. For me, one of the two GPU cores hates it if I shut down the VM it's attached to. It will purple screen of death my ESXi host and idk how to even begin fixing it. Not all games work. Some games are programmed to grab the first GPU that it sees and then switch to high power later (in this case, it would be vmware's "igpu" and that one doesn't support 3D applications while PCIe passthrough is enabled. You can see my timespy (demo) benchmark scores below, with the GPU power limited to 76% (and connected through parsec). It is very poor, but in more optimized games, you might find it ok, if you've been running an integrated GPU for games before, like me. (it seems the CPU is the major limiting factor here, which I will be upgrading in the future) Update: after a bit of tinkering, a CPU upgrade to the E5-2680, and overclocking, I've increased the score by 1100 points: Definitely can improve thr overclock, but I'll probably need to apply better thermal paste to the GPU

Hi, I know that the title may not be the best description of what I'm trying to accomplish but hear me out. Me and few friends decided that we are going to have a competition in building the best autonomous racing robot. Our goal is to build a large obstacle course, set a general route through it and let our robots make the most efficient and fastest path while avoiding obstacles. All of this is supposed to happend outdoors in area which size equates to around 100 m2. We are not complete noobies in making autonomous stuff and all have experience in machine learning, small electronics , motors etc. But one thing we disagree about is the requirements for our outside navigation systems. We are having trouble choosing between normal cameras,stereo cams, lidar, gps, ultrasonic sensors, all of those?? Even tho our project is complicated we are still trying to keep our costs down. So I'm looking for an advice from someone with more experience. What combinations or a singular part should we choose as our navigation+obstacle detection system for our project? I'm really looking forward to your answers. Thank you.

Hi, I would like to replace my MacBook Air 2015 with some ARM processor and at least 32 GB of RAM. I would like to do photo and 4K editing. Furthermore, I would like to get into machine learning. I found Jetson AGX Xavier Developer Kit ( https://developer.nvidia.com/embedded/jetson-agx-xavier-developer-kit ).However, I could not find any case for it. Is there any better SBC for the above tasks? I appreciate any help you can provide. Mic

Budget (including currency): $4000-$5000 (R60000 - R70000) Country: South Africa Games, programs or workloads that it will be used for: Machine Learning (Training deep learning models), Gaming is not essential, but ML and gaming PC parts go hand-in-hand. The RTX 3090 (if I can eventually get one around MSRP) is essential for training large models, due to its 24gb of memory. So unless there are better GPUs with CUDA support (basically just NVIDIA GPU's) and large memory (above 16gb) at around the same price as the 3090, the 3090 is the core component. Looking for advice on the best parts to complement the RTX 3090 without bottlenecking it or overspending on unnecessary parts. Don't really know too much about overclocking yet, but it would be nice to future proof this ability to speed up certain components if necessary. P.S. I chose the 1000W Superflower PSU since it is cheaper than the other branded 750-850W PSU's (corsair, cooler master, etc...) in my country. Is it still overkill to have a 1000W PSU for this build? I read something about running a PSU at less than 60% of its max wattage is not great for the PSU, looking for some confirmation regarding this statement. Also, any advice on if all the components will fit well in side the case and what fan configuration I should use would be appreciated. Thanks in advance.

Summary In a method fundamentally different than DLSS or anything of the sort intel used a AI trained on a real life data set to enhance GTA V at playable frame rates. The video posted focuses mainly on the cityscape not the characters. Quotes As the researchers put it in the abstract of their article. My thoughts At a glance that their paper it is not clear just what type of hardware they did this on. Did it take a modern workstation computer perhaps with a quadro GPU or one of Intels own GPU's for running all of this AI or did it take a super computer? In my area of expertise if we run a simulation then we specify the computer, the OS the whole software stack so that other researchers can try to replicate our work. Intel may have reason to not want to let that kind of information out into the world. If Intel has a desktop GPU that can, perhaps with help from servers while online , that can do this to any old game it would make everything else obsolete. Your 3090 would be instant garbage. Something tells me that isn't the case though. Imagine a GTA 7 where instead of textures created by artist to be more realistic and approximate car paint, trees, or human skin ... the textures lighting etc are based on samples from the real world. Combined with greatly enhanced deep learning and ray tracing. It has been my personal opinion that we are nearing the point of diminishing returns when it comes to visual fidelity in video games and interactive graphics. A point where having the latest graphics card, higher resolution, and higher frame rate will give marginal practical improvements in fidelity. This shows the path to that point lies in a fundamentally different sort of graphics processing than has been done before. That it is not just a matter of more and different types of cores on the GPU but also using them in a smarter way. Using reality to create realistic games. Which leads to another issue. Who really wants their game about Modern Warfare or criminality to be super realistic? Consider Yvonne's reaction to a VR game with Zombies. I myself once had would could be called a flashback due to having played GTA San Andreas and fought to exterminate the Ballas gang. For a spit second I saw some youths dressed just like them and thought I was in the game. Is too much realism really a good thing? Imagine a photorealstic Mortal Kombat Fatality? Sources https://www.theverge.com/2021/5/12/22432945/intel-gta-v-realistic-machine-learning-cityscapes-dataset http://vladlen.info/papers/EPE.pdf -- formal but not peer reviewed paper.

Hi, I wanna get an eGPU (probably a 2060 or maybe a 3060/70 if those are in my budget) My laptop is a Surface Pro 2017 / 5 which doesnt have usb-c, thunderbolt or access to mPCIe slots. Any ideas? Is there a way to connect it over the usb 3.1 gen 1 port? (5gbps) I know it would be slow, but i dont think that would really matter if im not using it for graphic applications. I will be building a pc some time in the future (a year or 2 away tho) once I have enough money, so I will be buying a graphics card and PSU anyway. thanks heaps, jensen

I vaguely recall from one of the recent LTT Labs videos (or maybe it was the $1M computer series) that Linus was talking about wanting to do more use-case specific hardware testing, especially for machine learning applications. Is that true? I believe that right now there are far more hobbyist ML practitioners/researchers out there than there are professional ones, and some of those are forming communities that are doing some really impressive work. For example, the RLGym community uses reinforcement learning and the OpenAI Gym specification to build Rocket League bots that compete in the RLBot tournaments. For most of these communities, including RLGym, access to compute by far the most limiting factor for progress. Admittedly Rocket League might not sound like all that worthy of a cause, but the community is actually advancing the state of the art in multi-agent learning as well as hybrid imitation learning/reinforcement learning techniques, and these advancements can have major implications for other fields such as advanced robotics control systems. With that in mind, if LTT/LTT Labs is considering doing more use-case specific testing for ML applications, have you considered whether it's possible to perform that testing in such a way that supports community-lead AI/ML projects like RLGym? If it's something that you'd like to do but aren't quite sure how, I'd be very happy to volunteer some time to help figure it out. Full disclosure: I'm an active participate in the RLGym community, but I'm not an RLGym project maintainer, and I don't own or receive revenue from the RLGym twitch channel or anything else affiliated with RLGym. I'm just a hobbyist who loves hacking on distributed systems. I focus most of my community time/effort on improving the efficiency of the learning systems we use so that we can squeeze every last bit of impact out of any/all donated compute.

Hi, I found a basic notebook online to write code to implement CNN from scratch. Can some machine learning specialist help to fill the notebook with the relevant bits and pieces. cnn.ipynb

While working out the details for my next build, I came across some Amazon and Newegg reviews of X670E motherboards. There seem to be a lot of complaints with these boards (e.g., memory issues). Do you know if there are widespread problems with these boards? Could they have been rushed to market to meet holiday marketing goals? Should I wait until these issues are resolved, or consider current alternatives? Target build: CPU: AMD Ryzen 9 7950X GPU: Nvidia GeForce RTX 3080 Motherboard: ??? Cooling: Corsair iCUE H150i ELITE LCD Display Liquid CPU Cooler (AIO closed loop). RAM: DDR5 5200 MHz. Target budget: $3500 USD or less. Country: United States My goal is to configure and buy a prebuilt to support my AI/machine learning research. This activity involves investigating new architectures. The data sets can be large, but they’re not the massively huge sets used by the “deep learning” crowd. I rejected the AMD Threadripper chips due to cost and their suboptimal performance on single threaded benchmarks. I believe the AMD Ryzen 9 7950X CPU is a good fit for me, because it has enough cores/threads for multithreaded programs, and its single threaded performance makes it excel at gaming. Also, its power consumption is far less than equivalent CPUs from Intel, which fosters longevity. In addition to answering the questions above, please let me know if I’m on the right track.

Budget (including currency): €1500 (euro) Country: Ireland Workloads that it will be used for: Machine Learning/AI and general programming/software development I'm a 4th year comp-sci student doing my final year project on Deep Reinforcement Learning/AI but I've been having some issues because my only computer is a laptop with 8GB RAM and an insufficient GPU. I considered getting an eGPU but decided my best option would be to put some time into doing my own build as I think it'll stand to me going forward. The problem I'm having is there's so many options and I don't know where to start/if what I'm choosing is advisable. Any feedback from anyone who's done something similar would be greatly appreciated. My budget is 1500 euro, plus or minus 100 euro. What I'm thinking is: - 10 core CPU, i9 possibly but don't know if Intel or AMD is preferable. - For GPU, I'm not sure so would appreciate some help. Ideally one with tensor cores - 32 GB DDR4 RAM, but open to alternatives - For storage, possibly, Samsung (MZ-V7E1T0BW 970 EVO SSD 1TB - M.2 NVMe Interface Internal Solid State Drive with V-NAND Technology, Black/Red - For Motherboard, again I'm not too sure. - Water cooling would be nice but only if there's room in the budget.

Confused between --> ASUS TUF Gaming F15 2022, 15.6"(39.62 cm) WQHD 165Hz, Intel Core i7-12700H 12th Gen, RTX 3060 6GB Graphics, Gaming Laptop (16GB/1TB SSD/90WHrs Battery/Win 11/Office 2021/Gray/2.2 Kg), FX577ZM-HQ067WS https://amzn.eu/d/h0tirBw And This --> ASUS TUF Gaming F15 (2022), 15.6"(39.62 cms) FHD 300Hz, Intel Core i7-12700H 12th Gen, RTX 3060 6GB Graphics, Gaming Laptop (16GB/1TB SSD/90WHr Battery/Win 11/Office 2021/Gray/2.2 Kg) FX507ZM-HF068WS https://amzn.eu/d/hoJAbH8 Planning on buying it in march 2023, so might as well consider a better upcoming laptop if there is one under the pricepoint.

Budget (including currency): US $ Wanted to shoot for about $1,600. Ended up near $1,800. Country: USA Games, programs or workloads that it will be used for: Machine Learning. Possible gaming (I assume). ML isn't a strong suit within my knowledge base for builds. So, after doing a bit of research, I wanted to see if my results were valid among those who know what I don't. Other details: Fresh Build from scratch. I've done some PC builds in the past, but it's been a min. My friend is currently pursuing a degree in CS with concentration in Machine Learning. Therefore, not a huge budget to work with as a college student. Friend wants something that is more modern than their current rig (much in this case), yet still has some relevance down the road for possible entry-level career work. What I have so far: https://pcpartpicker.com/list/LVPFyK My thought process on certain items are below. If my understanding is incorrect or maybe a different perspective would be useful, feel free to let me know. CPU: Intel i5 13600K To my recent understanding, ML would normally want to utilize more cores/threads. However, with 13th Gen release, I thought the i5 13600K became somewhat comparable to the i9 12900K(F) for $200 less with a sacrifice of 2 cores/4 threads. CPU Cooler: EVGA CLC 280 113.5 CFM Liquid CPU Cooler Friend wanted liquid cooling. With 13th Gen power draw, I didn't disagree. I understand an additional mounting may (?) be required. Mainly concerned with case dimension specs with this cooler and whether or not problems may arise. Motherboard: MSI MAG Z690 TOMAHAWK WIFI ATX LGA1700 A difficult process to narrow down to this board. A bit on the pricey end, but I landed on this for various reasons including DDR5, WiFi, I/O, VRM design, etc. I thought this was a solid board that could live a while amongst the new rollouts we've been seeing. RAM: G.Skill Ripjaws S5 64 GB (2 x 32 GB) DDR5-5600 CL36 With ML taking up a fair amount of GPU resources, having 64 GB among two DIMMs seemed like the way to go in order to avoid CPU paging out to disk as often while the GPU was doing work. Chose 5600 MT/s for its slight increase in speed, XMP 3.0 (unsure if XMP version usage also depends on Mobo BIOS capabilty and if it does, does the Mobo I selected support it), and price point being fairly reasonable for DDR5. GPU: MSI RTX 3060 Ventus 3X 12G OC GeForce RTX 3060 12 GB Probably the biggest points of contention with this build. I tried comparing many different GPUs for their price/performance and memory mostly. To my understanding of ML tasks... the more GPU memory, the better. Wanted to avoid LHR models in order to not limit future (potential) uses. For a college student/entry-level work, thought this was about the best choice. Power Supply: SeaSonic FOCUS Plus 750 Gold 750 W 80+ Gold Certified Fully Modular ATX I felt like this was a solid choice and adequately handled the workloads well, but there could be something I'm not considering.