AMD Ryzen 3900X

The Intel i9-12900k has a TDP of something like 240 watts (I don't remember the exact number). The AMD 5950x has a TDP of 105 watts. The 5950x does roughly 10-15% worse single core and 10% worse all-core. Intel is far behind AMD. Overclock the 5950 a bit and you're back to where the i9 is stock, but still with less power. (I haven't actually done any research into that last sentence, but that seems plausible) The AMD customer experience is also much better. I can upgrade - and soon will be upgrading - my 2 year old 3900x system to a 5950x. I don't have to buy a new motherboard like I would have had to with Intel. Plus my system has had PCIe 4.0 for a few years (which I have been utilizing with a super fast SSD) and also supports ECC RAM - Intel will price gouge you for those by making you go to their enterprise targeted CPUs/motherboards (to be fair, they do now support PCIe 4 on the i9). Plus Intel withholding ECC from their consumer CPUs - even the high end ones - is just a major dick move and hurts all of the people like us on HN who want to have good hardware. And considering all the scummy anticompetitive stuff Intel has done over the last few decades, I'd say it's better to go with AMD even if their chips had a slight perf/$ disadvantage compared to Intel (but they don't). Of course, I'm not saying AMD is amazing and wouldn't do evil stuff if they got the opportunity, like Intel had, but AMD is still the underdog for just a bit longer, and in duopolies it's generall…

> There are simply far fewer differences between CPUs made in 2010 and today vs CPUs made in 2000 to 2010. I have stopped replacing machines (and smartphones) because they became outdated: the vast majority of compile tasks is finished in a fraction of a second, applications basically load instantly from SSD, and I never run out of RAM. The main limiting factor in my day-to-day use is network latency - and nothing's going to solve that. My main machine is a Ryzen 9 3900X with 32GB of RAM and a 1TB SSD. And honestly? It's probably overkill. It's on the replacement list due to physical issues - not because I believe I'll significantly benefit from the performance improvements of a current-gen replacement. I'm hoping it'll last until AM6 comes around! Every task is either "basically instantly", "finishes in a sip of coffee", or "slow enough for a pee break / email response / lunch break". Computers aren't improving enough to make my tasks upgrade to a faster category, so why bother?

I've been using a CS381 for years which sadly now is EoL, don't know if they're replaced it, but it's an excellent case. Previously I was using a Fractal Node 304 but that was cumbersome for drive maintenance. It has 8 hot-swappable SAS bays (also SATA compatible) and I run a Ryzen 9 3900X in 65W eco mode on an AsRock Rack X470 board which has another 8-12 SATA ports (can't remember the exact number, not used because I use an HBA for the bays), so connectivity for storage is high. There's 2 spaces for SATA SSDs on top of the drive bays and you could fit more in various spots if you tried, and 2 NVMe slots on the motherboard. Also got a single-slot nvidia GPU in there and a 4-port Gb NIC to supplement the 3 existing Gb ports on the board itself (one is dual-purpose for IPMI), some models of the AsRock rack have dual 10G ports too. It runs most of the time at around 90W which I think is exceptionally low for the performance available, and can go to about double that when the GPU is in use, still very reasonable.

The Marginalia search engine ran for over a year on a AMD Ryzen 9 3900X with 128 GB non-ECC RAM and a mix of NAS drives and SSDs of various types ranging from consumer crap, through enterprise drives, and even an Optane. 16 TB mechanical storage, 4 TB SSD. All on domestic broadband. It survived the HN front page and even bigger traffic spikes. Did blip out when Elon Musk tweeted a link to one of my blog posts, but only momentarily. Now that server's been relegated to run a test environment and to perform various odd jobs. I do think everyone interested in programming should have some sort of server in the house. Being able to run processing jobs for a few days really does radically expand what you're able to do.

I have played a little bit and I'm a bit confused about all the negative feedback regarding the performance. I have an AMD Ryzen 9 3900X and NVIDIA GeForce RTX 3090, so not too bad, but also not the best. And the game runs totally fine for me. I just reduced the screen resolution (I think 1920x1024 or so) and left the other settings as they were (I think all set to High). And I get very constant 45 FPS without hiccups. It feels very smooth and playable. Maybe it will get a bit worse when my city gets bigger, but so far, I don't see any problems. This is even before the first patch, which was released today. This patch is supposed to optimize the performance more. I will try it later. Note, this is with Linux. I run it on Linux with Proton 8. So maybe it actually runs better on Linux than on Windows?

I like multiple monitors too! I loved the two 32” 4K LG 32UD99-W monitors I used to have. Those were perfect too. Only switched to the 55” OLED TV because I lost those monitors—One had to to back to the company that provided it for working at home… aaaaand I broke the other. With big top-heavy monitors I have learned the lesson that all surfaces must be grippy towards each other under all circumstances. Also when and if the drawer-stand put under the monitor scooches 1mm out of the desktop area and off its grip-pads. Actually the lesson is: Always just get adjustable gas-pump monitor arms for everything. Less clutter, more desk space, more flexibility, better ergonomics. It’s even more worth it than I suspected it would be. They cost less than I thought and make a bigger difference. I have a nice little Ryzen 3900X machine. Reasonably lucky silicon - it runs at 4.2GHz at a slightly reduced voltage. 2 × 16GB DDR4-3200 CL14 sticks which are happy at 3400MHz CL14. I don’t recall the rank – it’s Crucial BallistiX. The sticks were too expensive for what they do! – I bought them as an experiment to see what the difference was if I got the fastest memory I could find without going to the scary steep part of the price-performance curve. There is a difference!, but not strictly worth it. GPUs: I have Radeons – 5700 XT and 6800 XT. Those are overkill for development-only but very nice. The 6800 XT is smoother and snappier at driving a 4K desktop! I’m curious how the 6600 XT is. (I…

Yes, Windows 2000 on today's hardware runs like a champ. Even Windows 2000 on contemporary high-end hardware ran very well. However, I do remember using installations of Windows 2000 on contemporary hardware that weren't snappy. Windows 2000 and XP needed a lot of RAM and a fast processor to be happy; running these operating systems on a machine that met or only slightly exceeded the minimum requirements was an unpleasant experience. Likewise, my experience with Windows 10 is that it's snappy on high-end hardware, but it's very sluggish on low-end hardware, especially if it's running on a HDD instead of an SSD. Six years ago I bought a refurbished ThinkPad T430. It had an HDD and originally shipped with 4GB RAM. Windows 10 was almost unusable with that configuration. Upgrading to 16GB of RAM helped, but it was still slow. What dramatically improved the system's performance was replacing Windows 10 with Linux Mint running Cinnamon. Suddenly my ThinkPad became pleasant to use. By comparison, I have absolutely no issues with Windows 10 running on my 12-core Ryzen 3900X with 64GB RAM and a PCIe 4.0 SSD. But that's high-end hardware.

ECC is working just fine on my Ryzen-based machine. PS C:\> wmic CPU get Name Name AMD Ryzen 9 3900X 12-Core Processor PS C:\> wmic Baseboard get Product,Manufacturer Manufacturer Product ASRock B450M Pro4 PS C:\> wmic MemoryChip get Manufacturer,PartNumber,BankLabel,Capacity,Speed BankLabel Capacity Manufacturer PartNumber Speed P0 CHANNEL A 17179869184 Kingston 9965745-002.A00G 2667 P0 CHANNEL A 17179869184 Kingston 9965745-002.A00G 2667 P0 CHANNEL B 17179869184 Kingston 9965745-002.A00G 2667 P0 CHANNEL B 17179869184 Kingston 9965745-002.A00G 2667 PS C:\> wmic MemPhysical get MemoryErrorCorrection MemoryErrorCorrection 6 # "6" is Multi-bit ECC

No contest! – The thing is mostly that I can mess with the Ryzen box, and run a bunch of different OS-es at full speed, and use various add-in hardware like video cards.

I have honestly, earnestly tried. I have tried to find a laptop that is for my needs and purposes truly equivalent to an Apple Macbook Pro. Or better. I always end up going with an Apple machine. There has always been a significant part of the assembly that just isn’t as good. It’s often disk speed or display quality. Build quality too. I want to underline that this is true for what I need out of a laptop. The opposite has been true in desktops. I have a Ryzen 3900X box and there still isn’t anything from Apple that I would replace it with. Not even to run macOS on it (which I do on the AMD box, using GPU passthrough).

The opposite has been true in desktops. I have a Ryzen 3900X box and there still isn’t anything from Apple that I would replace it with. Not even to run macOS on it (which I do on the AMD box, using GPU passthrough). My MacBook Pro 14" is pretty much on-par with my Ryzen 5900X CPU-wise (the M1 Ultra would surpass it by a wide margin). The GPU of the M1 Max is nothing to sneeze at, but I'd love it if they'd bring eGPU support to the M1 line. (And if one can dream, if NVIDIA would also make CUDA available.)

Agreed. Eigen can be relatively slow: in our Tiny Differentiable Simulator, with C++ templatized code, we generate C code using CppAdCodegen (by tracing a sim step) which is easily 5 times faster than the original Eigen code. And together with openmp it runs very fast: Ant OpenAI gym simulation at 2 million steps per second on an AMD Ryzen 3900x at 12 cores/24 threads). See the (unreadable) generated code from Eigen+CppAd Codegen here: https://github.com/google-research/tiny-differentiable-simul...

Game performance, and cooling in general outside of a very specific use case, isn't affected by what the marketing department of Intel says the 10 core AVX wattage of the 12600k is being compared to what the marketing department of AMD says the 6 core AVX wattage of the 5600X is. When a single core is maxed out the real world measured difference is roughly 12 watts between the two and remember the 12600k is doing slightly more per second than the 5600X. If your gaming device of choice is a laptop that'd be interesting but both are desktop gaming CPUs. If 24/7 all core workload power efficiency is your concern I'd steer you away from either gaming optimized option. I really like what AMD has done, my last 2 desktop CPUs have been a 3900X and a 5950X, but if I were to buy a entry level gaming CPU right now the 12600k would be very interesting regardless what the all core AVX thermals look like on a stock cooler.

I can confirm that ECC DRAM worked as expected on my current AMD Ryzen 9 3900X, an AMD Phenom II x6 1090T that I had in a previous build, and an AMD Opteron 185 that I had in a build before that. The operating system detects the error-correction capability in the memory controller, and errors are properly detected (tested by reducing the memory timings slightly below stable values). I don't think there's a mass-market laptop with ECC support, but AMD's desktop CPUs have supported ECC for as long as I can remember. That being said, it's been an optional feature that is up to the motherboard maker to implement, and not all do.