- Joined
- Dec 12, 2022
I'm guessing the RGB sells better, so isn't a shelf warmer, maybe.
GPUs & CPUs & Enthusiast hardware: Questions, Discussion and fanboy slap-fights - Nvidia & AMD & Intel - Separe but Equal. Intel rides in the back of the bus.
- Thread starter Smaug's Smokey Hole
- Start date
- Joined
- Nov 15, 2021
Maybe one of you nerds can help me.
I have a program that is writing thousands of files to a shitty commodity SSD I have in the server node. Don't blame me, I didn't write this gay shit. It is what it is, the way this program handles its output is known to be objectively bad practice, but the programmers didn't care because they're fuckwits with a known reputation for being fuckwits. I raised the file limit to something very large over the 1024 limit, and now, instead of running out file handles, it freezes up at, I don't legitimately know, maybe around 5000 files open.
This is because I am trying to run a program that spawns 1024 instances, and each instance is trying to write several tiny files. Making it spawn fewer instances is not an option, I have to test this config this way for completely valid, non-retarded reasons.
Do I need a less shitty SSD, or is there some magical switch I can flip in Linux to fix this?
I have a program that is writing thousands of files to a shitty commodity SSD I have in the server node. Don't blame me, I didn't write this gay shit. It is what it is, the way this program handles its output is known to be objectively bad practice, but the programmers didn't care because they're fuckwits with a known reputation for being fuckwits. I raised the file limit to something very large over the 1024 limit, and now, instead of running out file handles, it freezes up at, I don't legitimately know, maybe around 5000 files open.
This is because I am trying to run a program that spawns 1024 instances, and each instance is trying to write several tiny files. Making it spawn fewer instances is not an option, I have to test this config this way for completely valid, non-retarded reasons.
Do I need a less shitty SSD, or is there some magical switch I can flip in Linux to fix this?
- Joined
- Feb 28, 2015
- Joined
- May 9, 2017
No, those are just regular old cage nuts.
That's pretty close, yeah.
- Joined
- Mar 30, 2023
- Joined
- May 9, 2017
I'm not sure what size that is, but the design is EXTREMELY close. Mine don't have that extra corner hole, but based on that page description I'm guessing that's for drainage.
Unfortunately it doesn't look like they have a store front page... I'll see if I can find somewhere that sells them.
- Joined
- Dec 19, 2022
That has to be a software issue, no? SSD shouldn’t be aware of active file handles, both SATA and NVME are just block devices to the OS, you’re not telling them individual files, just blocks to read/write/scrub. I guess if you’ve already raised file handles for the Linux VFS you could check if there is another such switch, or a hardcoded limit, for the file system used by the actual disk.
- Joined
- Nov 21, 2020
Is the output small enough to fit in ram? Can you try using /dev/shm or mount a tmpfs somewhere else, then copy the files once it completes?
Also, you'll want some metrics when it hangs, vmstat/top, load average, free memory and see what's actually going on. Probably also check "dmesg" to see if the OOM killer is being invoked or there are other kernel messages.
- Joined
- Nov 15, 2021
Yeah, I can just make a 512GB...dev...shim...thing and run there. I'll figure it out.
- Joined
- Jul 3, 2021
Can someone explain why something like an Intel 300 is $100? That thing has two cores. Is it just priced so absurdly to keep anyone from actually buying it? I imagine it’s sold at volume to OEMs for much less.
- Joined
- Jul 18, 2019
At a certain point, you run into the fixed-cost limit for how cheap CPUs can get in terms of just the raw materials that go into making them. AMD is still selling new 2-core Athlons for like $80 last I checked.
- Joined
- Aug 28, 2019
If it has leaked out onto the market through third-party sellers from OEM machines, pricing could be strange. But Intel does have an RCP of $82-92, and it appears to be a genuine product in a retail box sold on Newegg.
Intel Processor 300 is a refresh of the Pentium Gold G7400, based on a 6+0 Alder Lake die (with less L2 cache per core than true Raptor Lake chips). 200 MHz faster, no turbo.
They shouldn't be disabling 2/3 of cores and 1/2 of the iGPU and selling it directly to people, but if they wanted to sweeten the deal they would put it at $50, like the AMD Athlon 3000G used to be. There's a floor for how cheap a CPU can be but I think they're just not making very many of these, thank Pat.
- Joined
- May 5, 2022
Change FS to XFS for higher parallel workload? Check to see if using swap space due to you running oom?
Also like someone said, tmpfs but use f2fs with high i/o capability (supposedly)?
- Joined
- Jan 25, 2024
My RAM is being EXTREMELY gay.
When examined in Task Manager, it's supposedly operating at 5500 MHz:
However, CPU-Z reports that it's operating at only 685 MHz:
Geekbench reports only 1370 MT/s:
Typically, MHz should be half of MT/s, so the CPU-Z and Geekbench numbers are consistent, while Task Manager is the odd one out.
So next, I went into my BIOS/UEFI settings, but there is no way to access any information about the RAM or edit any timings, via either the dashboard or the advanced options. Clicking the "Memory" card on the BIOS screen simply displays the amount of memory with no details about speed and no ability to change any timings or other settings:
Google says to use "AI Tweaker," but there is no "AI Tweaker" option available anywhere within my BIOS. Hitting F7 for "Advanced" options also does not yield anything about the RAM other than the amount of memory:
This computer is ASUS - Vivobook Pro 15.6" Laptop FHD - AMD Ryzen 9 7940HS with 32GB RAM - NVIDIA Geforce RTX 4060 - 1TB SSD
When examined in Task Manager, it's supposedly operating at 5500 MHz:
However, CPU-Z reports that it's operating at only 685 MHz:
Geekbench reports only 1370 MT/s:
Typically, MHz should be half of MT/s, so the CPU-Z and Geekbench numbers are consistent, while Task Manager is the odd one out.
So next, I went into my BIOS/UEFI settings, but there is no way to access any information about the RAM or edit any timings, via either the dashboard or the advanced options. Clicking the "Memory" card on the BIOS screen simply displays the amount of memory with no details about speed and no ability to change any timings or other settings:
Google says to use "AI Tweaker," but there is no "AI Tweaker" option available anywhere within my BIOS. Hitting F7 for "Advanced" options also does not yield anything about the RAM other than the amount of memory:
This computer is ASUS - Vivobook Pro 15.6" Laptop FHD - AMD Ryzen 9 7940HS with 32GB RAM - NVIDIA Geforce RTX 4060 - 1TB SSD
Attachments
- Joined
- Dec 20, 2019
This may be a stretch but I had the same thing happen not with memory but the cpu on a Dell laptop. It was locked at the lowest possible scaling frequency and would not scale up. The only solution was to remove the battery fully and put it back. Maybe look to see if people have tried the same with memory for the laptop model and see if it fixes it?
- Joined
- Jul 18, 2019
Check your power profiles in Windows. That laptop looks like it's using LPDDR5 which does a lot of dynamic clock shenanigans for lower power operations. You might also want to try hwinfo64 which might give you more fine-grained breakdowns of frequencies.
- Joined
- Jul 3, 2021
Is it just me or is Incel’s success in spite of its efforts?
Turbo Boost has been a big steaming pile of confusion with people removing power limits and becoming surprised that the system throttles. But OEMs do the same thing, especially with laptops that fry your thighs.
The P-core/E-core optimization for Windows 11 literally did nothing and in some cases Windows 10 is actually better (coupled with Windows 10 actually being a better operating system). I’m not sure if it’s Incel’s or Microcock’s fuck up so I’ll laugh at both.
APO is a lolcow technology. Originally it was only going to support 14th gen. Now it supports 12th and 13th, but it still doesn’t actually improve anything! Imagine being an engineer and spending months optimizing games for autistic speds only for your project to not even work.
Incel seems to be a marketing company with a fab. Their design strategy is to throw lots of power and make up an acronym to make their technology sound like it works better than it actually does. There isn’t even any thing special about E-cores. They’re just smaller cores.
Still won’t touch GayMD with a twenty-nine and an half foot pole.
Turbo Boost has been a big steaming pile of confusion with people removing power limits and becoming surprised that the system throttles. But OEMs do the same thing, especially with laptops that fry your thighs.
The P-core/E-core optimization for Windows 11 literally did nothing and in some cases Windows 10 is actually better (coupled with Windows 10 actually being a better operating system). I’m not sure if it’s Incel’s or Microcock’s fuck up so I’ll laugh at both.
APO is a lolcow technology. Originally it was only going to support 14th gen. Now it supports 12th and 13th, but it still doesn’t actually improve anything! Imagine being an engineer and spending months optimizing games for autistic speds only for your project to not even work.
Incel seems to be a marketing company with a fab. Their design strategy is to throw lots of power and make up an acronym to make their technology sound like it works better than it actually does. There isn’t even any thing special about E-cores. They’re just smaller cores.
Still won’t touch GayMD with a twenty-nine and an half foot pole.
- Joined
- May 5, 2022
Longsoon for the win?
- Joined
- Mar 10, 2024
E-cores exist primarily for market segmentation and selling OEMs new CPU SKUs when they have no new silicon. Intel loves market segmentation. Historically, they used to use enabling or disabling Hyperthreading for this purpose, but once core counts on CPUs started to rise, it became a problem. It's obvious that a 4 core 4 thread CPU is better than a 2 core 4 thread CPU, but what about a 4 core 8 thread CPU vs a 6 core 6 thread CPU, or ultimately with Comet Lake where they had to stop, an 8 core 16 thread CPU vs. a potential 10 core 10 thread CPU? (IE the 10700K vs the theoretical 10800K that never came out).
E-cores was their solution to this problem. They can sell a seemingly limitless number of different parts by mixing and matching P-cores and E-cores. That's the entire reason for the "14th gen", to sell Dell and HP and Lenovo "new" parts for their specs sheets.
I don't think Intel isn't "trying", they just have an extremely mature business with lots of very old partnerships and making the biggest splashes with their product launches isn't all that important to them.