Search the Community

Budget (including currency): ~1500USD (not including GPU). Don't have tight budget as hopefully (mostly) funded by university. Just really would like efficiency in performance/cost ratio. Looks are much less important. Country: United States Games, programs or workloads that it will be used for: Mixture of Academic deep learning via Jax/Pytorch, python scientific numerical simulations (some parallelizable, but some sequential only), most immense scale tasks pushed off to cloud (eventually will have to run 100s of neural netowrkss training in parallel but can be pushed to a computational cluster just testing for several parallel runs of a single 501x501x501 neural network.) Primarily will flash Linux on it--most likely Debian or Pop OS for maximum hardware capability. Currently I run these on my 6-core 2.5Ghz laptop, and it slows to a crawl when I try to work on anything else contemporaneously. It also takes ~4-14 hours for a single run, which is pretty long given the task isn't that intensive. 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): No existing parts list. Completely from scratch build. Have a single Ultrawide monitor 1440p Odyssey G7 32" a few years old, an existing mouse and keyboard (Leopold Keyboard FC980M and Logitech G502 Hero), and a thunderbolt docking station for existing 2018 MB Pro. The goal is to build a performant, but cost-efficient build that will be primarily used for testing and running code, primarily python numerical code and machine learning software like Jax and Pytorch. Size of hard-drive is less important and I have a personal laptop I can code on while code is running on desktop. I will probably do a mixture of running code in person and remoting into the desktop from my laptop. While I know GPUs are important, I am thinking about holding off on the purchase of a GPU separately until prices come down on the 4090 or 4080, hopefully over next month or two. I am deciding whether to try running everything from the CPU (as I do now on my laptop), or buy a cheaper hold-over GPU used like a Nvidia 2070S. For CPU I am primarily thinking of getting an Intel 13900K or 13900KF (not sure if integrated graphics is necessary if I want to wait on Nvidia GPU/generally good suggestion to display Desktop while Nvidia is running CUDA cores), but am open to alternatives. I do think achieving close to 6Ghz via OC if possible is appealing given the mixture of sequential and parallel tasks. For Hard-drive I am thinking of a Samsung 970 Evo. And for RAM I am still considering, but am currently thinking of Corsair Vengeance 16GBx2 DDR5 memory. Again, if it can be overclocked, that is even better. For CPU cooling I was thinking iCue h150i Elite Capellix, but if there is a better alternative, I would consider that. For power I was considering 1000W Gold 80+ EVGA power supply. I've heard that supposedly the 13900K seems to work best with Z790 Chipset boards and there is even supposed to be up to 350W overclocking. I don't know if heat output is manageable or if >6Ghz overclocking is truly possible, but if it is that would be great. I am not sure how much to invest in a board/what to consider here. I've thought about the MSI 790 Carbon, the Asus Maximus ROG 790 Hero, etc. but don't know how much the more expensive motherboards add in terms of thermal performance/OC potential and when you hit declining returns. For the case, I am not sure what would be best. I am likely to use the build over the next 2+ years and work in a pretty quiet academic setting as a grad student, writing papers etc as code runs. So ideally I would like a case that is both quiet and has good thermals. Ideally something, where when combined with noise cancelling headphones, I wouldn't hear it at all. I considered Bequiet case 500dx, but don't know if there are better alternatives. Space is not necessarily a constraint as I have room under my desk for a larger case. What would be nice, but is not necessary, is to have some USB-C ports as pretty much all my other tech is USB-C based. Any suggestions would be greatly appreciated. I am pretty comfortable building a PC, but haven't done a lot with things like thermal paste, etc. on the more-involved end. That being said, from lurking around LTT it doesn't seem that bad to do.

Hi there! I've been interested in machine learning for quite some time now, and have finally decided that I wanted to upgrade my hardware to include a GPU with more VRAM, as the models that I work with are usually quite large. For reference, I still want to continue using my framework 12th generation (i7-1260P, 32GB RAM) laptop, which is why U need to use an eGPU solution. Do you think that the P40/P100 are feasible cards for my requirements or should I use another card with less VRAM. My budget is around 500€ max. (For just the card). Thank you in advance.

Hi good day, My team is working on some neural network modeling for use in the medical sector. However some of the datasets we are working with are between 10 - 30 GB of data. We are running into some resource complications, having to truncate our data. Google Colab isn't a solution as we are outside the US and can't use Colab Pro. At the moment we can use a machine provided by a professor at our university, however this can only be used for some hours on some days. (i9-9900k, 64GB RAM, Nvidia RTX Titan). Even with this setup some data processing jobs takes 6-10 hours and even more. We are interested in building a Server to run JupyterHub for our use. I would like some recommendations on what would you all recommend for the CPU and the amount of RAM as we regularly hit the caps on both. As well as, some advice on how to ask companies if they are willing to sponsor some components to help further our contributions to the research field. Thanking you in advance for your time and help.

Budget (including currency): 50,000 HKD Country: Hong Kong Games, programs or workloads that it will be used for: Machine Learning, CUDA 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): Hey fellas, I am a student in a computer science lab and we want to build a pc for machine learning tasks. We don’t have much experience in setting up a heavy processing pc, hence I would like your help. A shop recommended us a build-up (attached in the post) The shopkeeper also recommended us to go with a higher power supply (mentioned in the end as optional) so that we may upgrade it to 4 GPU(s) in the future. However, we are skeptical if the current setup can fit 4 GPU(s). Are we right? We would like to have the option to add more GPU(s) in the future, given it doesn’t exceed the budget too much. Can you please suggest any required changes to achieve the same?

Hello Fellow Scam Baiters! Okay, so that title is a little bit confusing. Here's the thing. I need a lot of sample data. I'm trying to train an Open AI GPT-2 model to try generate natural language texts and try to waste more time. Training these models takes a lot of data. Like, a lot of data. Like tens or hundreds of gigabytes of plaintext data. A lot of data. But, to get started, what better place to get started collecting said data than the Linus Tech Tips forum! SO, here's what I need from you. You can post as many replies as you want to this topic. For each correspondence (scammer contacts you, you respond), just copy what they said and how you responded. Drop as many replies as you want! For context on how your data will be used: I'll be manually moving data out of the forum onto a private database. One training point will be one correspondence, rather than an entire conversation. Think about it this way: when the scammer tries to get contact information, a lot of scam baiters will respond in a similar way. The machine model will try to pick up on some pattern there: ask for contact information means that I should respond by sending a fake name. So, if you want to make my job a little bit easier, only drop one correspondence in one reply. We do not anticipate any risks associated with collection of this information. Your participation is completely voluntary, and there are no penalties if you choose not to participate or skip any correspondences. By participating, you are providing important, real-life information that can be used to train the next generation of neural networks for natural language processing. You can remove your contact information and any email headers, as well as reply or forwarding headers. If you don't, they'll be removed in the data cleansing process. The datasets generated will not be made public unless by valid request (i.e., you're a student at an accredited institution and you're doing research). If you have questions about this policy, please send a direct message.

Hi all - First time post & first time PC build. I have been a MacBook user in my personal life for 8 years, and use a PC for development at work. I am building a PC as my existing MacBook cannot support 4 screens, and its performance for machine learning is lackluster relative to what I am looking for. PCPartPicker Part List CPU: AMD Ryzen 9 3900X 3.8 GHz 12-Core Processor ($432.41 @ Amazon) Motherboard: Asus TUF GAMING X570-PLUS (WI-FI) ATX AM4 Motherboard ($189.99 @ Best Buy) Memory: Corsair Vengeance RGB Pro 32 GB (2 x 16 GB) DDR4-3200 Memory ($164.99 @ Best Buy) Storage: ADATA XPG SX8200 Pro 1 TB M.2-2280 NVME Solid State Drive ($149.97 @ Amazon) Video Card: MSI GeForce RTX 2080 SUPER 8 GB GAMING X TRIO Video Card ($818.00 @ Amazon) Case: Cooler Master MasterCase H500M ATX Mid Tower Case ($239.99 @ Amazon) Power Supply: Corsair RMx (2018) 750 W 80+ Gold Certified Fully Modular ATX Power Supply ($155.98 @ Newegg) Operating System: Microsoft Windows 10 Pro OEM 64-bit ($139.99 @ Other World Computing) External Storage: Seagate Expansion 8 TB External Hard Drive ($131.99 @ Adorama) Total: $2423.31 Prices include shipping, taxes, and discounts when available Generated by PCPartPicker 2020-05-01 01:14 EDT-0400 1. Budget & location $2600 United States. 2. Aim. I have two requirements: A. PC should be able to support 4 monitors for day trading. B. PC should support Machine learning workload. 3. Monitors. I do have the 4 27-inch monitors, and they have HDMI & DVI ports. ASUS VE278 with 1920×1080 resolution. I would like a recommendation for a 4-arm stand if someone has one. 4. Peripherals: I plan on using 64 bit Windows 10 Pro. Is there any other recommended software to buy on a new PC (e.g. for virus protection?) Already have mouse & keyboard. The biggest cost choices came down to CPU, GPU, and motherboard. While I originally was leaning with the Ryzen 3700X, I had extra budget and decided to go with 3900X since the 3800X seems like a little improvement. CPU for Machine Learning helps with some of the pre-processing of data prior to training, so I do see incremental value here. I decided to go with the Asus TUF GAMING X570-PLUS (WI-FI) ATX AM4 Motherboard as it will pair with my Ryzen 3900X and seems to have good reviews overall. Have had good experience with ASUS before. I have been torn between the 2080 Super vs 2080 Ti. I decided to go with the 2080 Super because for machine learning purposes, I would like to see some return on investment in regards to the work I expect to do on the machine. If I see the return on investment in my Machine Learning work, I may just spend more at a later time & get a second GPU to support my workload. One area that is particularly confusing for me is cases. I do care about the double-tempered glass with RGB light aesthetic, but with so many options and lots of variance in reviews, its hard to understand who delivers quality here. Some of the brands that seem good are Corsair, ThermalTake, and Cooler Master. Any recommendations here? The $240 for the case does seem pricey, but I think it should have good air flow. Welcome any feedback, thank you so much. This is a large investment for me.

Hi, I haven't seen it being covered in this section of the forum, and it qualifies as news to me: As PCgamers's article states, Valve is toying with machine learning as a way to overcome cheating in Counter Strike. Their strong feelings about this approach being the best one stem from the unavoidable realization that fighting cheats the old fashioned way is equivalent to a never ending arms race, fixing anti cheat to detect cheats which will be made relevant again when programmers of said cheats finally find what have changed. Cheating in online multi-player has always been a concern, especially in competitive games such as counter-strike. It is always welcomed to find new smart ways to fight against it, and it is a feature many competitive players ask for continuously. Will this however be enough? One can argue that rising the game price to throw off cheaters from the perspective to rebuy the game would be enough. Other can argue that machine learning algorithms could be fooled as well by someone knowing how they work. From what I learned about the subject, the objection that it can be fooled is valid. It however remains that it is a complex enough algorithm, which may only be messed with by really talented programmers with strong mathematics background, which are less common than only really talented programmers. In addition, if the learning is unsupervised (as with neural networks or some other methods), it becomes harder to understand how it is choosing to flag players and therefore to fool it. What place can it have in the future though? Indeed, it is noteworthy to point out that those algorithms alone could ban legit players all the more easily. Should it be the only way to ban players? As of now it seems it will only be a support to the peer reviewed system. It is a pioneer approach, which has its drawback for Valve at least. Their spokesperson pointed out that it will need huge computing power to account for enough matches to train it upon, and then on the ability to watch every game to flag cheaters. Enough to need additional datacenters, since the servers cannot cater to this need. Should they try to do the calculation on the client side as well to relieve them from part of those calculations? I don't know if this fits the format, but I tried to speak of everything they said in the article and to add from my perspective to ask you guys what you think about this!

Original Article: https://futurism.com/googles-new-ai-is-better-at-creating-ai-than-the-companys-engineers/ Essentially, Google have created an AI powerful enough that it can create better versions of itself, which in turn can create better versions of itself etc. This is essentially the first steps towards the singularity, an AI more powerful than humans that could out-think us and destroy us should it wish. Currently, it exists only as Artificial Narrow Intelligence (ANI), but due to the layering, this could soon progress to AGI (artificial general intelligence) and ASI (artificial super intelligence), as each AI becomes progressively better at coding AI. EDIT: Here's a good video that explains some of the consequences of ASI:

In a step forward for the human-machine interface (or in a scary step to losing your privacy) a study has been published on a new approach for a Neural Network using Machine Learning to interpret functional Magnetic Resonance Images (fMRI) of the visual cortex into actual images. From the MIT Technology Review: https://www.technologyreview.com/s/604332/mind-reading-algorithms-reconstruct-what-youre-seeing-using-brain-scan-data/?set=604340 Looking at the DGMM images, they definitely appear to be the closest to the original presented images... Just hope that this doesn't turn into an Orwellian situation. https://www.engadget.com/2017/05/09/neural-network-recreates-brain-images/

Figured I would make this as a single thread as opposed to two separate threads since the topics are closely related concerning machine and robotic learning... First up, Microsoft Maluuba, which is a research group that Microsoft acquired back in January, has released a paper outlining a method of using recurrent networks to allow robots to generate questions from a body of text. https://techcrunch.com/2017/05/10/microsoft-maluuba-is-teaching-machines-to-ask-questions/ Essentially, this means that if you tell your personal computer assistant to perform a task, it could potentially ask you for additional details in order to better accomplish that task, such as if you were to schedule an appointment, it could ask for details on location and other people involved with that appointment. On the second topic concerning machine learning, MIT's CSAIL has developed a new technique for robots to learn to perform tasks and pass that information on to different robots. Basically, instead of having a programmer write code for all the possible ways a task could be performed or having the robot learn via demonstration of a task, the new method called C-LEARN is a hybrid of both techniques. https://www.csail.mit.edu/teaching_robots_to_teach_other_robots Kind of interesting with this approach is after some key constraints have been programmed in and the task has been demonstrated to the robot, the robot can then suggest separate motion plans to accomplish the task that can be approved or edited as need be. When the robot is left to it's own decisions, it was able to complete the task about 87.5% of the time, however after it's plans were reviewed and edited, it could complete the task 100% of the time. The other interesting part of this is that the robots were able to transfer their 'knowledge' on how to complete a task to other robots to use. As for non-'Skynet' applications of this, the article points out: So I guess we can look forward to our robot overlords questioning our reason for existing when they teach other robots how to take over. https://www.engadget.com/2017/05/10/mits-programming-system-lets-machines-learn-from-each-other/

Originally read this article on Hot Hardware then followed up with the Science Mag article. Doctors at Stanford University in Palo Alto have published the results from a study they conducted where they tested four separate Machine Learning Algorithms against the American College of Cardiology/American Heart Association (ACC/AHA) guidelines for predicting a heart attack. The outcome of this is that the AI was actually more accurate than the ACC/AHA guidelines by a margin. According to the Science Mag Article here is how the testing was performed: http://www.sciencemag.org/news/2017/04/self-taught-artificial-intelligence-beats-doctors-predicting-heart-attacks As for the results, the article point out: This is definitely a positive application of Machine Learning AI and hopefully this in combination of a skilled doctors expertise can improve the accuracy rate of predicting coronary issues and other health problems even further. The big thing that should be stressed though is that I don't think it would be a good idea to remove the human element from reviewing the data, but just supplement a doctors expertise with the AI's ability to model complex datasets. http://hothardware.com/news/machine-learning-ai-algorithms-predict-heart-attacks-better-than-doctors

I want to do a machine learning build in the future. I know that machine learning algorithms are better run on GPU, but I know that the CPU is important as well. I plan on getting Dual NVidia GPUs running in an SLI config for that, but I'm debating on a CPU. Would something like a Coffee Lake i5 or i7 do it for me? Would it be better to run a Ryzen 7 1800X? Or should I go balls to the wall and get a pantshitter threadripper? What do you think about it?

Source: NVIDIA via CNET I can't upload the actual paper because of the attachment limitation to 20 MB of LTT so just click on the NVIDIA link above to read the actual paper which is 28.7 MB. I think the likes of TMZ or other famous celebrity gossip website would love this. From the paper: Just so you know, those pictures above aren't pictures of real people They're just computer generated. Although in Figure 5, one of the pictures kinda looks like Beyonce. The researchers concluded that it still needs improvement. To be honest I only understood a few of it but from the looks of it, it can be used by the likes of Apple to train Face ID in their neural networks in order to spot impostors or fake news websites can use it to create mass hysteria.

Source: PLOS One via Phys.org journal.pone.0185123.pdf Nasir, M., Baucom, B. R., Georgiou, P., & Narayanan, S. (2017, September 21). Predicting couple therapy outcomes based on speech acoustic features. (I. McLoughin, Ed.) PLOS One, 23. doi:https://doi.org/10.1371/journal.pone.0185123 I think Taylor Swift needs this as she makes songs based on the long list of her ex-boyfriends while writing songs about them and probably she'll stop playing victim. I can see two things happening here if this AI, rising divorce rates because an AI says they are incompatible or marriage counselors out of their jobs because couples put their trust on an Ai kinda like how a lot of people self-diagnose and treat their diseases on the internet. I am not a relationship or a dating expert and I have no plans of becoming like Taylor Swift who dated at least a dozen people. But therapists can use them as tool to help them during couples counselling. I can see app developers like Tinder, Ok Cupid and others jumping in to do self-help apps like placing the phone in the middle of a table while talking and the app will analyze if the other person is honest or truly in love or not. But I'm not yet convinced that analyzing pitch and tone alone of someone's voice while talking to their partner is a best way to determine if a marriage is doomed to fail or it is still worth saving. I still think that a relationship especially a marriage is based on communication and trust. I think a lot of relationships fail either because the other person is a philandering piece of shit or both the people involved didn't explored and tested the waters enough to determine if they're actually compatible or one or both people in the relationship change and either one of them is not a fan of it.

I've been watching a streamer on Twitch for the past couple of months who is building a game agent framework that allows you to play any game (Steam and non-Steam) automatically through various machine learning and "artificial intelligence" algorithms. It's been heavily developed and I wanted to share it with you guys as the developer released the first Beta today. Link to the GitHub repository: SerpentAI Personally, I'm a primarily C++ programmer and this framework got me into Python as well as exploring machine learning. Perhaps this framework triggers the creative minds of some of you. - d0p3t

Sources: Cornell University Library, Science Mag, and Threat post The article however didn't mentioned password managers so I guess it's safe to use a reliable password manager with two factor authentication. I have a feeling that as AI, machine learning and neural engines become more powerful, we might see cyber attacks much more serious. At the moment, it predicts what passwords are the easiest to guess to give companies chance to change their weak passwords into a more secure one. But as far as I'm concerned, most websites don't read passwords as plain text like "I<3myhotboss", websites read it hashed like this one "eed4b508e6f5acda3178c880bc490546" and I think there's already an online database containing hashed passwords that are used by hackers to brute force. But then, I can see this being used by legit password managers and they'll notify the user if the password they're using is easy to guess or has been used somewhere else so that they'll notify the user to change for a more secure password. So I'm all for this and I hope this will be implemented to current password managers.

1. Budget & Location: $1.5K - $2.5K. Can receive parts in Austin, TX or my residence in Mexico 2. Aim: Deep Learning, scientific number crunching. I want a top-end GPU like the Tritan Xp and 64GB+ of RAM to do Data Science projects. I just finished waiting 17 hours while my current desktop trained a language classifier. That kind of sluggish turn-around makes development very difficult. I can rent GPUs on AWS, but in addition to costing money it is an awkward way to work and I prefer to have my own hardware. But I need something portable, because I may be traveling back and forth between Texas and Mexico and want to take my GPU with me. 3. Monitors: don't care. Big. 4. Peripherals: I need USB 3 ports for external drives and an Ethernet port I viewed the marvelous overviews Linus did of the Alienware Graphics Amplifier and the Razer Core, but the external GPU approach has big downsides: Weight of laptop: I don't want a heavy laptop Rapid depletion of battery: battery has to be recharged after 45 minutes of use Drivers and software: for Machine Learning I need Linux installed, not Windows 10 Home. Ideally my system would be a dual boot with both Linux and Windows 10 Professional. I fear that I might plop down lots of $$$ for these proprietary systems only to discover that they are not adaptable to Machine Learning applications. Linus has persuaded me that a compact computer in a Dan case might be a great alternative. However, the parts Linus put into the system he built with a Dan case are just too expensive for me, especially the Intel Xeon E5-2699 processor! I need a high-end GPU and *lots* of memory, but I don't require a top-end CPU. I am not a gamer and care nothing whatsoever about graphics. I just need to crunch huge amounts of data. QUESTION: Where can I get guidance building a system in a Dan case similar to the one Linus described in the video, but maybe with a better GPU (Nvidia Triton Xp) and a much less expensive CPU. Parts Linus used: Crunchyroll link: http://crunchyroll.com/linus Intel Skull Canyon NUC link: https://linustechtips.com/main/topic/... Buy GTX 1080 on Amazon: http://geni.us/ilxNCb0 Buy Xeon E5-2699 v4 on Amazon: http://geni.us/bpWoPYA Buy DAN Cases A4-SFX: http://bit.ly/2aNOyBX Buy Silverstone SFX-L 700W: http://geni.us/cABATzn

Hi, I've been been working in the field of Machine Learning for over 7 years now and have been recently handed over a significant chunk of funding to aid my adventures. I require a work horse machine capable of handling immense work loads with lots of computing. i value the suggestion of the LTT community and would like to put my plan forward for your review and criticism. my max budget is about 7K$ ( is flexible) and I've come up with a PC part picker list. i of course will have to find a source to buy/ ship it to India my base of operations. unfortunately it's coming 1k over budget and i would your suggestions in come cost cutting. https://pcpartpicker.com/list/xLLDPs here is the list. the mains here lots of RAM, beefy GPU and enough cores to last a long time with 24x7 operation handling hundreds of gigabytes of workload.

Hi what a first post on the forum, but I just recently was given 2 dell poweredge R710s with 48 gbs of ram each and no hard drives I'm currently using a program called knime on my desktop quad core with 8gbs of ram to do a research project for one of my classes in school which involves image classification. I was wondering is it possible to take a 250 gb ssd with windows 10 and just stick it in one of the power edges to take advantage of the processing power and if so how can I go about doing that because at the moment my quad core can not handle the load I'm putting on it. All help is appreciated

Sources: New York Times, Author's Note (via Google Docs), OSF (pre-print) wang_kosinski.pdf Comments could get nasty here. I need your moderating powers @iamdarkyoshi for anyone with comments against the CS (political/religious rants) or derails the topic. It resembles a lot of a topic I posted before about an AI predicting the outcome of a relationship. I won't lie that it got me concerned for a bit but reading the actual paper and the FAQ (Google Docs) the authors provided made me think it's indeed a way to warn people about the repercussions of artificial intelligence and machine learning. It reminds me when Elon Musk said that Mark Zuckerberg doesn't know the dangers of AI if left unregulated. I'm with Elon Musk on this one. On the lighter side of things, this technology can be used by dating apps targeted to gay individuals but then, I'd rather see the person face to face rather than letting an AI do it for me. Looking at the comments of other experts, it's obvious that the algorithm is very limited on their sample size and subjects since they only used gay white men and women. The thesis authors said in their Google document that: Just imagine you're a tourist in a country and a facial scanner at the airport identifies someone as gay or lesbian, they can immediately put you in a database. Just imagine oppressive countries who have staunch religious doctrines using this to identify closeted gay people and have them thrown from top of buildings or just be shamed by society into committing suicide. Or better yet, just imagine North Korea developing an AI who determines if someone has doubts on the regime or is planning to defect to South Korea. Knowledge is power and just like a knife that can be used to stab someone or cut that delicious medium rare steak, this algorithm is also a double edged sword. At the moment, it's not ready for prime time as it has limited sample diversity and the authors admit that. In fact as stated in their Google Doc that they want to be wrong and they are terrified of the results. More refinement is definitely needed and just like any scientific theory, it needs to be replicated. At the moment, all humanity knows about the potentials and risks of AI and machine learning is just at the tip of the iceberg. So there's no need to be concerned or alarmed just yet but we also shouldn't dismiss a scientific finding just because we disagree with it or it could pose harm to us and in fact, I would like for this paper to be published soon in the Journal of Personality and Social Psychology. I'm not a Psychology major but I encourage people with that degree to replicate this thesis but with much diversified subjects. Also, it should be kept in mind that "peer-review" is not always cranked up to be as a method of validating or dismissing a theory. Peer-review is important in academiai but most of the time, it's just validating research methodology, checking grammar and look for suspicious and interpolated/forged results. Peer-review doesn't determine if a thesis' conclusion is right or wrong. In fact, there are a lot of bogus theses that are "peer-reviewed". The only way to determine if a thesis' conclusion is right or wrong is if it's replicated and check the results to see if they match up or not. [See NCBI, Nature, CBC, WSJ) Again, I highly encourage everyone reading this thread to read this Google Doc as it clarifies a lot of concerns on privacy, human rights, etc. https://docs.google.com/document/d/11oGZ1Ke3wK9E3BtOFfGfUQuuaSMR8AO2WfWH3aVke6U/edit?usp=sharing

Hello All, This is my first post to the forum, so please take it easy on me. I recently bought a 2006 Mac Pro (bought it for a $1). The unit has a bad motherboard in it, so most of the components are either too old or not worth reusing. It did however have 2 Xeon processors inside, I plan on using for a later project. The parts I plan in reusing are the case itself, the power supply 980 Watts, and the fans. The rest is junk Here's my idea: I want to build a Machine Learning Box that I can use for my Master program. I want to use it to train models, be used in some neural net applications (training/testing), things like that. I have been looking all over the internet, for tips and guides on what makes the best bang for your buck machine. I have used some gaming builds as a starting point, then tweaking to my use case. The machine will run Ubuntu 14.04 or 15.04 primarily with possible dual boot support for Windows. ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ Build Requirements: Intel Chip: i7 or Xeon (Based on the total budget for the project I believe i7 is the only option) NVIDIA GPU(s): For the initial build phase I will buy 1 GPU, with the intention of buying a second down the road. Thus SLI is important. Plus the developer support with DIGITS, CUDA cores, etc are important. Supporting Motherboard: Ideally having a motherboard that would drop into the Mac Pro case would be AWESOME! But I am almost certain some mods to the case will have to be made. Also, having fast I/O, PCI expandability, and M.2 support are feature requests. Somethings I found that should be important for this build include: PCI Express Lanes -> I believe that 40 lanes is a requirement because I plan on having 2 GPU's (Not sure yet which NVIDIA card I want yet), and I would like the lanes to be divided equally between the 2 GPU's (16x16). That way I avoid any possible bottlenecks. RAM -> From what I have read for this use case, it is best to have enough RAM (memory) to hold entire datasets. My plan is to start with 32 GB initially, then move up to 64 GB and a step 2, then possible to 128 GB as a step 3. So expanding RAM is important. At the time of writing this I believe that quad-chanel Memory is the way to go, the benefit being speed. What do you guys think? M.2 -> I want the system to be extremely fast. I would like to use a M.2 solution for holding the OS (Ubuntu), and OS related file, and applications. All the datasets will be held in Secondary Storage (Western Digital Red Hard Drive) Question I have: 1) When deciding in a CPU, should I go with more cores and a slower clock? Or fewer cores and a higher clock? (Based on this use case) 2) Does my assumptions for PCI Express Lanes hold true? Would there be a bottleneck if 1 GPU is getting 16 lanes and the other is getting 8? 3) Is there any guides that you guys know of that I could follow? 4) What NVIDIA GPU's have the best SLI support? Or which GPU have been found to work best in this use case? (I know Titan X is great, but please keep the budget in mind) 5) Do you guys have any suggestions on the build? Summary I have made 2 different PC Part Picker lists that can be found at: https://pcpartpicker.com/user/sprichard/saved/#view=8TXRBm They are listed under the "Mac Pro $1000 Build" and "Mac Pro $1500 Build" As you can see from the link I have build for $1000 budget and $1500 budget. I am working on a $2000 budget build. I would say $2000 is the limit for this build. Let me know what you guys think! I am open to all suggestions as i have not purchased anything, and I am still trying to get my head wrapped around all that I need to know to make a badass Mac Pro ML Box.

Hi, Been a long time LTT follower, but just joined the forum and this is my first post. The community here looks really nice - and thank you for being supportive I am looking to buy a laptop for doing serious software engineering work, including training ML models, etc. - which I need to learn/practice moving ahead. My OS of choice is Ubuntu (I am just used to Debian packages, and know where everything is, and feel very comfortable here). As this will be my only machine - it has to be very stable - and hence I am a little scared of getting systems like Razor Blade 15 or Asus ROG Z - where users have installed Linux distros like Ubuntu but with compromises and with some things not working (connecting to projectors, wifi issues, sleep-wake issues, etc) - and hence I am inspired to get a System76 laptop. I did consider a ThinkPad with Ubuntu (which is stable), but they don't seem to have a good graphics card anywhere which I will likely need to use with TensorFlow, etc. I am considering either an Oryx Pro which maxes out with an i7-9750H and RTX2080 (16.x 1080p 144Hz screen), or an Adder WS which maxes out with an i9-9980HK and RTX2070 (4K OLED screen - which I don't really need but I'm not complaining about either). I don't know which of these combinations should I go for considering that this is a laptop (thermals), I need it mostly for coding, ML training, etc. and also - maybe once in a while - running a few games on Steam (but that is not a priority in any way). Can someone please help me out here? I very much appreciate your advise! Thank you again Oryx Pro: https://system76.com/cart/configure/oryp5 Adder WS: https://system76.com/cart/configure/addw1

I am a huge fan of LTT and I happen to work on machine learning (in academia). Later this year, Dec 8th - Dec 14th, one of the biggest academic conference in machine learning, NeurIPS 2019 will be held in Vancouver. https://neurips.cc/ It might be interesting to make a video about it. Although the conference has an academic nature, all the hardware (Nvidia, Intel... ) or the software compaies (Microsoft, Google...) will be there and will demonstrate their latest technologies. Plus, I think it may also be interesting to interview some academics and ask them how the advances of PC hardwares has changed the machine leanring research. Just an idea.

I've been meaning to build a Deep Learning machine for quite sometime now and because it is my first pc build since 2012 I wanted to get some advice on two things. First, I'd like to know your opinion on my build as it is, before Computex. Second, since Computex is coming up, do you think I should wait for the new Ryzen CPUs or go for the one I choose? Oh, and in case you ask, currently I live in Greece and my budget is 1100 € no more. Also, as a monitor, I have a 4K smart TV, which I know it is not ideal but it'll do for now, because I mainly use my laptop for work. Keyboad HP Pavilion 600 32.49 Mouse Logitech M171 9 CPU AMD Ryzen 5 2600 Box 144.79 SSD Crucial P1 500GB 67.45 GPU Nvidia GeForce RTX 2070 8GB 500 RAM G.Skill RipjawsV 16GB DDR4-3200MHz (F4-3200C16D-16GVKB) 91.49 PSU Seasonic M12II-620 Evo 76.82 Case CoolerMaster MasterCase H500 100 MOBO Asrock B450 Pro4 74.35 Final Price: 1,096.39 €

Hello Guys, I need your help, your suggestions. Recently I was having a very hard time training my models on a non GPU Laptop I have to upgrade and I am not able to find a single laptop with Rtx 2070 or above available in India under budget. Training on cloud is also too expensive for me and I have got no choices left. I don't know what should I do building a pc will not be an efficient option as I am a student and I spend most of my year on campus. Even If I get a laptop with a thunderbolt3 port I will not be able to get a gpu enclosure for it (none available here). I don't know what I should do. Please guys let me know what I should do. Thank You.