• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Arrow Lake build for programming?

Joined
Oct 3, 2024
Messages
105 (0.48/day)
Location
Belgium
System Name N/A
Processor i7 3770
Motherboard Asus Z77-A
Cooling Cooler Master Hyper 212 EVO
Memory Patriot Signature 2x8GB DDR3-1600 CL11
Video Card(s) none (iGPU)
Storage Crucial MX500 500GB + Micron 1300 256GB+Patriot Burst Elite 120GB+Samsung 830 128GB
Display(s) Dell P2317H
Case N/A
Audio Device(s) Sennheiser SP20
Power Supply Corsair CX430
Mouse Logitech M90
Keyboard Belkin K100
VR HMD N/A
Software Fedora 42 beta/Ubuntu 12.04 (for android development)/Windows 11 Pro
Benchmark Scores coming soon!
I am going to upgrade my current system soon, as it really is getting quite old (I built it in 2020 with used parts, I had a very small budget as I was fucking 10). My budget is ~700 euros as I'm still a high school student, and one thing that caught my eye was how on Phoronix testing, the 245k just murders both the 14600k AND the 9700X on programming (my primary workload aside from reading tech and space news and posting on forums). Am I mistaken that this would make a good machine for my usecase?


Full specifications: https://de.pcpartpicker.com/list/vm8m3w (buying from Amazon Germany as it's much cheaper then from my native Belgium, also don't mind the lack of a case as my current machine has been running fine for the last 5 years without it and I don't feel like wasting at the minimum around 10% of my budget on a decent case)
 
If you look at TPU reviews and the tasks you are doing out competes the rest, buy what's best for your budget and applicational use.
 
I just took a peek why have you selected DDR5-5600 when that generation uses DDR5 6400 as there native spec
 
If you look at TPU reviews and the tasks you are doing out competes the rest, buy what's best for your budget and applicational use.
TPU appears to come to the same conclusion
1737907036523.png

I must also clarify that another part of the reason that I'm considering this is the technology behind Arrow Lake is very cool. Especially Foveros and the fact that the compute tile is N3B. It may not matter to me as a end user, but it's cool nonetheless.

I just took a peek why have you selected DDR5-5600 when that generation uses DDR5 6400 as there native spec
Because I don't really need fancy memory. You're right, of course, but for me even my current DDR3-1600 is overkill so I didn't see the point in spending more
 
Very soon we will have 64GB DIMMs. 2x64GB 5600 and 6400 just around the corner:)
Aren't those already a thing in the server space?
 
Programming can mean many different things. But it often means working with big data sets and having to run virtual machines. Without knowing any details, i'd just say that you probably need a lot of RAM. Give some thought to that.

Also, as you won't be overclocking anyway, why not wait for Arrow Lake non-K?
 
Last edited:
Oddly enough, Arrow Lake is great for pro and semi pro workloads and bad for gaming (okay maybe not bad but just not what was expected!). Intel have just released an update for the Core 200 series which seems to have made it even better for your type of workloads (though still try to track down anyone with your unique use cases) but still not great on the gaming side. Check out Kitguru who only today released an update on the new patches for Arrow Lake and finally thinks they are ready, which should help you. On the memory, buy the cheapest but fastest DDR5 you can as it will only help!

 
Last edited:
Programming can mean many different things. But it often means working with big data sets and having to run virtual machines. Without knowing any details, i'd just say that you probably need a lot of RAM. Give some thought to that.

Also, as you won't be overclocking anyway, why not wait for Arrow Lake non-K?
Yep, that sounds like a good idea. It is already released but ludicrously overpriced, so I guess I'll just wait a bit more.
 
Having cool technology can give you fuzzy feelings, but please, don’t get distracted by that which doesn’t actually matter for it to work well and keep working well for your use. I have some cool technology myself, it’s collecting dust in an unused room of our house. Or, I’m even using some unusual tech right now, my processor’s apparently a limited edition (I’ve read on Wikipedia hours ago, after using this computer for years), that Intel had recalled shortly after release and superseded with bug-fixed models (no, not that fix). But, it’s not great, the modern web (especially incredulously awful YouTube) has just become too much of a resource drain on this small machine.

Fortunately enough, Arrow Lake’s 245K (or non-K of that realm) does seem like a good choice for what you have planned—maybe.
i’ll echo Wirko’s sentiment, you’d need to look a bit deeper than that. It’s a shame that media and many reviewers as well as commentators often cloud the waters even further with seemingly helpful simplifications. It’s something I’ve struggled with for years when I’ve started out and I think I’ve only resolved it once I got gud enough so I basically know better than those people anyways. ;) (And frankly, I often instead defer and say, I don’t know, cause, that’s my next point.)

So yeah, programming can mean different things. Rust, C, C++, Python, Perl, JS? As others have mentioned, it often involves container stuff these days—which I guess is true, I just wouldn’t have had this on my mind, since I’ve never programmed like this.
Some languages have made a name of themselves for how awfully long they take to compile (and this can not always be sped up though parallelization), other stuff profits from more cores, cache and memory bandwith very much.
Some five years ago (maybe more, maybe someone will date it) when I last read up on it, the heaviest programs (Firefox, Chrome) in full LTO-mode, would take more than 16GiB of your RAM to be put together at the end and swapping was OK but not great (from what I remember). This amount of RAM was still somewhat big for consumers back then. I don’t think requirements have gone up infinitely after that; in fact, Thin-LTO had pushed them down quite a bit without forfeiting at least some of the benefits. Fortunately enough, 32GiB is considered kinda the standard “premium” loadout by now, it feels, with many who actually want to program saying they’ve gone 64- or even 128-. (And this was maybe a year ago, or so.)
I can’t stress enough that I have no clue whereto they’re even moving all their memory. It’s incomprehensible for me. VMs? For what? I guess if you need to deploy multi-platform, you could have all your tests run in parallel and have that go even faster when you do have enough cores? That’s the thinking when time is money, in which case you’d probably go high end.
Now, since I do at least somewhat keep up with my personal interests in programming, but am very focused in my preferences on leanness (while being feature rich and comfortable to use) end efficiency, and since I use utterly weak machines myself (your thing has twice as many cores as mine and they’re also faster each), I’m wary of serving you dumb advice, even though, I’ve been in the general vicinity of programming for almost twenty years now.
You already have a computer. Where are your actual bottlenecks? What’s keeping you from making progress, having fun, whatever? Is there any technology you want to use but can’t? (I couldn’t realistically compile any modern browser on this half-tablet, though I’ve compiled a couple of kernels for my own use and I just had to wait.)
Like, there is no single “good for programming build” I believe (that’d be wild). Are you lacking cores, speed-per-core (mostly frequency), memory bandwidth, memory quantity?
When you don’t use heavy tools—yes, this might sound very trivial—programming can be lightweight. Even if you attempt to throw hardware at sluggishness, you might find that switching editors does yield better results.
So, sorry for not deciding much of anything for you … my posts sometimes grow large like this without intent (I haven’t planned it out in advance)—but I think, you might either be sweating it too much or not enough. Find out where you’re hurting today, if you’re hurting at all. The 245K will serve you well as a default, but, if you’re not really pushing any boundaries today, then you have to say, so might anything even cheaper you can get your hands on. (Maybe some trusted person to inherit it from, instead of the wide net?)
Or, if your junk still works, maybe don’t buy anything at all? Do you want to game? If you can game on it, you can do non-money-earning, not utterly serious programming on it, I reckon.
I’ve been hurt more often and more strongly by a lack of RAM (4GiB in here, 8GiB in the newer device, with another 8GiB in a drawer ready to be installed after warranty’s end) than by a lack of grunt, though again, kernel compilation on a 1P4E Alder Lake Pentium did take surprisingly long, that was a bummer for sure.
Yet, I have to insist, if you want to learn programming right now, it’s either you go a traditional route, where HW requirements will be modest enough that your old rig should get it, or you’ll go down the AI route, in which case … doesn’t most of that stuff run on someone else’s computers, so you’re mostly fighting ridiculously abominatious misgivings of programs whose code is so awful that it just takes (a literal) ten times more power than it otherwise would, though apart from that, you don’t need much?
Yeah, I guess ARL can manage that as good (and bad) as anything else.
For running AI stuff locally, you should probably rather look towards spending money on your accelerator (graphics card). You might not even need to update your system for that, go check out how much your slow system memory and PCIe3 bus actually limit your GPU of choice, it’s possibly not much at all. (Or maybe it is, feel free to let me know.)
 
Last edited:
Having cool technology can give you fuzzy feelings, but please, don’t get distracted by that which doesn’t actually matter for it to work well and keep working well for your use. I have some cool technology myself, it’s collecting dust in an unused room of our house. Or, I’m even using some unusual tech right now, my processor’s apparently a limited edition (I’ve read on Wikipedia hours ago, after using this computer for years), that Intel had recalled shortly after release and superseded with bug-fixed models (no, not that fix). But, it’s not great, the modern web (especially incredulously awful YouTube) has just become too much of a resource drain on this small machine.

Fortunately enough, Arrow Lake’s 245K (or non-K of that realm) does seem like a good choice for what you have planned—maybe.
i’ll echo Wirko’s sentiment, you’d need to look a bit deeper than that. It’s a shame that media and many reviewers as well as commentators often cloud the waters even further with seemingly helpful simplifications. It’s something I’ve struggled with for years when I’ve started out and I think I’ve only resolved it once I got gud enough so I basically know better than those people anyways. ;) (And frankly, I often instead defer and say, I don’t know, cause, that’s my next point.)

So yeah, programming can mean different things. Rust, C, C++, Python, Perl, JS? As others have mentioned, it often involves container stuff these days—which I guess is true, I just wouldn’t have had this on my mind, since I’ve never programmed like this.
Some languages have made a name of themselves for how awfully long they take to compile (and this can not always be sped up though parallelization), other stuff profits from more cores, cache and memory bandwith very much.
Some five years ago (maybe more, maybe someone will date it) when I last read up on it, the heaviest programs (Firefox, Chrome) in full LTO-mode, would take more than 16GiB of your RAM to be put together at the end and swapping was OK but not great (from what I remember). This amount of RAM was still somewhat big for consumers back then. I don’t think requirements have gone up infinitely after that; in fact, Thin-LTO had pushed them down quite a bit without forfeiting at least some of the benefits. Fortunately enough, 32GiB is considered kinda the standard “premium” loadout by now, it feels, with many who actually want to program saying they’ve gone 64- or even 128-. (And this was maybe a year ago, or so.)
I can’t stress enough that I have no clue whereto they’re even moving all their memory. It’s incomprehensible for me. VMs? For what? I guess if you need to deploy multi-platform, you could have all your tests run in parallel and have that go even faster when you do have enough cores? That’s the thinking when time is money, in which case you’d probably go high end.
Now, since I do at least somewhat keep up with my personal interests in programming, but am very focused in my preferences on leanness (while being feature rich and comfortable to use) end efficiency, and since I use utterly weak machines myself (your thing has twice as many cores as mine and they’re also faster each), I’m wary of serving you dumb advice, even though, I’ve been in the general vicinity of programming for almost twenty years now.
You already have a computer. Where are your actual bottlenecks? What’s keeping you from making progress, having fun, whatever? Is there any technology you want to use but can’t? (I couldn’t realistically compile any modern browser on this half-tablet, though I’ve compiled a couple of kernels for my own use and I just had to wait.)
Like, there is no single “good for programming build” I believe (that’d be wild). Are you lacking cores, speed-per-core (mostly frequency), memory bandwidth, memory quantity?
When you don’t use heavy tools—yes, this might sound very trivial—programming can be lightweight. Even if you attempt to throw hardware at sluggishness, you might find that switching editors does yield better results.
So, sorry for not deciding much of anything for you … my posts sometimes grow large like this without intent (I haven’t planned it out in advance)—but I think, you might either be sweating it too much or not enough. Find out where you’re hurting today, if you’re hurting at all. The 245K will serve you well as a default, but, if you’re not really pushing any boundaries today, then you have to say, so might anything even cheaper you can get your hands on. (Maybe some trusted person to inherit it from, instead of the wide net?)
Or, if your junk still works, maybe don’t buy anything at all? Do you want to game? If you can game on it, you can do non-money-earning, not utterly serious programming on it, I reckon.
I’ve been hurt more often and more strongly by a lack of RAM (4GiB in here, 8GiB in the newer device, with another 8GiB in a drawer ready to be installed after warranty’s end) than by a lack of grunt, though again, kernel compilation on a 1P4E Alder Lake Pentium did take surprisingly long, that was a bummer for sure.
Yet, I have to insist, if you want to learn programming right now, it’s either you go a traditional route, where HW requirements will be modest enough that your old rig should get it, or you’ll go down the AI route, in which case … doesn’t most of that stuff run on someone else’s computers, so you’re mostly fighting ridiculously abominatious misgivings of programs whose code is so awful that it just takes (a literal) ten times more power than it otherwise would, though apart from that, you don’t need much?
Yeah, I guess ARL can manage that as good (and bad) as anything else.
For running AI stuff locally, you should probably rather look towards spending money on your accelerator (graphics card). You might not even need to update your system for that, go check out how much your slow system memory and PCIe3 bus actually limit your GPU of choice, it’s possibly not much at all. (Or maybe it is, feel free to let me know.)
Interesting read. I must confess my current rig IS working totally fine (I've been toying with AI which obviously does not run fine and I have to defer to Colab for that), but I want an upgrade (again, I don't need it, I just want it) and might as well speed up compile times in the meantime. I don't have the budget for a GPU that can run anything serious. I'm not a gamer, which is why I get by with the utterly useless Intel HD 2500 graphics in here. My programming consists mainly of Python data analysis, Flutter programming, and the occasional C++.

Out of curiousty, what ARE your specs?
 
You don't need a very powerful computer to do programming.

Your system is pretty old though. I would say just get a cheap 12600K and 32 GB DDR5-6000 just to get on something more modern.
 
the 245k just murders both the 14600k AND the 9700X on programming (my primary workload aside from reading tech and space news and posting on forums). Am I mistaken that this would make a good machine for my usecase?
I don't know what you mean by programming but 14600k and the 9700X are categorially better for things like compile times, as is everything with more cores in general :
1737921921087.png
 
I don't know what you mean by programming but 14600k and the 9700X are categorially better for things like compile times, as is everything with more cores in general : View attachment 381840
I'm assuming these tests were done on Windows? The numbers I were quoting were from Linux testing, which is what I use, and ARL is quite a bit better on Linux.
 
Interesting read. I must confess my current rig IS working totally fine (I've been toying with AI which obviously does not run fine and I have to defer to Colab for that), but I want an upgrade (again, I don't need it, I just want it) and might as well speed up compile times in the meantime. I don't have the budget for a GPU that can run anything serious. I'm not a gamer, which is why I get by with the utterly useless Intel HD 2500 graphics in here. My programming consists mainly of Python data analysis, Flutter programming, and the occasional C++.

Out of curiousty, what ARE your specs?
If what you're mostly doing is Python and AI stuff, then wouldn't a 9700x do way better?
Just take a look at the numpy and tensorflow benchmarks on phoronix and Zen 5 totally crushes the core ultra.
Otherwise, if you're mostly doing large C/C++ compile jobs, then the 245k gets the upper hand due to the extra cores.
 
If what you're mostly doing is Python and AI stuff, then wouldn't a 9700x do way better?
Just take a look at the numpy and tensorflow benchmarks on phoronix and Zen 5 totally crushes the core ultra.
Otherwise, if you're mostly doing large C/C++ compile jobs, then the 245k gets the upper hand due to the extra cores.
Hmm... I'd say my workload is a pretty equal split. The choice is, then, whether to go for Core Ultra and better technologies like CUDIMM and a not-total-shit IMC, or Zen 5 and a much better upgrade path
 
AMD said the AM5 will support the next Ryzen CPUs, but that is the end of the socket. I would not be surprised if the Intel 1851 Socket will not get another generation. The Ultra series IMC is a beast, but suffers from high latency. Won't matter for anything but gaming.
 
Hmm... I'd say my workload is a pretty equal split. The choice is, then, whether to go for Core Ultra and better technologies like CUDIMM and a not-total-shit IMC, or Zen 5 and a much better upgrade path
It doesn't seem like you're going for neither CUDIMMs, nor high-speed DIMMs, nor even a high-density setup, so does the IMC even matter for you?
OTOH, do you actually fathom doing an upgrade down the lane to a probable Zen 6 CPU that may drop on AM5?

The best argument I can think of is that the i5 is 20 eur cheaper, in case you're on a really tight budget.
 
What about 8700G?
 
What about 8700G?
That's a monolithic Zen 4 part, originally meant for mobile that was repurposed for AM5.
A 7700x ought to be faster than it, CPU-wise. The 8700g is only relevant if you're interested on the integrated 780m it has.
 
That's a monolithic Zen 4 part, originally meant for mobile that was repurposed for AM5.
A 7700x ought to be faster than it, CPU-wise. The 8700g is only relevant if you're interested on the integrated 780m it has
I'm well aware, but it seems to be the cheapest AM5 8 core part
 
I'm well aware, but it seems to be the cheapest AM5 8 core part
A 9700x would still fit within your budget, wouldn't it?
If all you're aiming for is the cheapest product, an alder lake i5 will probably give you the best bang for the buck, and be a hefty upgrade over your current setup.
 
Programming can mean many different things. But it often means working with big data sets and having to run virtual machines. Without knowing any details, i'd just say that you probably need a lot of RAM. Give some thought to that.

Also, as you won't be overclocking anyway, why not wait for Arrow Lake non-K?
Software engineer over here agrees. IDEs will eat a lot of RAM (and no, VSCode is not an IDE), you will have a bunch of browser tabs open, pointing at various doc sources... If you have to save now, at least plan to get 32GB soon(ish).
Also a 32" 4k monitor is a huge help. Nothing like giving your IDE as much real-estate as it deserves.
Neither of the recommendations are musts. But they are the what will improve your experience a lot if you have room for improving the build you posted.
 
A 9700x would still fit within your budget, wouldn't it?
If all you're aiming for is the cheapest product, an alder lake i5 will probably give you the best bang for the buck, and be a hefty upgrade over your current setup.
Software engineer over here agrees. IDEs will eat a lot of RAM (and no, VSCode is not an IDE), you will have a bunch of browser tabs open, pointing at various doc sources... If you have to save now, at least plan to get 32GB soon(ish).
Also a 32" 4k monitor is a huge help. Nothing like giving your IDE as much real-estate as it deserves.
Neither of the recommendations are musts. But they are the what will improve your experience a lot if you have room for improving the build you posted.
How does this look?:


PCPartPicker Part List: https://de.pcpartpicker.com/list/LqFQGJ

CPU: AMD Ryzen 7 9700X 3.8 GHz 8-Core Processor (€341.58 @ Galaxus)
CPU Cooler: Thermalright Assassin X 120 Refined SE 66.17 CFM CPU Cooler (€18.00)
Motherboard: *Gigabyte A620M DS3H Micro ATX AM5 Motherboard (€98.37 @ Galaxus)
Memory: *Crucial CT2K16G52C42U5 32 GB (2 x 16 GB) DDR5-5200 CL42 Memory (€79.98 @ Amazon Deutschland)
Storage: Kingston NV3 1 TB M.2-2280 PCIe 4.0 X4 NVME Solid State Drive (€54.90 @ Amazon Deutschland)
Power Supply: Corsair CX (2023) 650 W 80+ Bronze Certified ATX Power Supply (€61.90 @ Amazon Deutschland)
Total: €654.73
Prices include shipping, taxes, and discounts when available
*Lowest price parts chosen from parametric criteria
Generated by PCPartPicker 2025-01-29 18:31 CET+0100
 
Back
Top