Hacker News Reader

Comment by sdeziel | Hacker News Reader

> the problem for the PC industry is that old PCs are good enough, hence people don't feel the need to upgrade.

Throwing away a working electronic device to replace it with a new one is really bad when you think of the ecological impact. I'm glad that some resist to the urge of buying shiny new stuff.

View on HN

I am running 5 - 16 year old PCs on my network and they play 99% of the games on steam.

Stuff I purchased in 2000 is still running like a CHAMP. ASUS Motherboards( half died in the first year, and were replaced ). Once they make it past 1 year they seem to run forever. I have upgraded components like the disk drive, a few of the processors from 2.3Ghz Core2 Duo, to a blaze 3.0 GHz Core2 Duo. And replace thed GPU with what I can as friends upgrade theirs and sell old ones for cheap. Now that they are all running Windows 10 64Bit, I can put up to 8 Gb of ram in them. My kids love it, they can have 4 friends over and play Killing Floor, Men of War, Terreria.

I mean for the average use case, a 5 year old pc generally has the compute capability.

Gaming requires updating (though 32nm cpus are still perfectly fine).

Scientific computing will move to the cloud soon enough, as you either need absurd capabilities, or a recent laptop is generally powerful enough for data science (depending on the usecase).

Hell, I dusted off a 10 year old core2duo HP laptop and it does most of what I need it to (browsing, jupyter notebooks, etc.) fine. The one common app that's painful to use it with is Facebook and other sites that became a sort of monstrosity (the ones with 40 NoScript dependencies you need to whitelist to see content)

I'm using a 2013 Macbook Air and the only website that makes it feel aged is Facebook.

Even the input boxes are poorly and awkwardly given these stupid resizing animations.

This.

I've had my HP Workstation for 5 years now. I found an xw8400 workstation off of EBAY for around $400. Oh sure the specs are a little lower than what you'd find on the newer PC's:

- Dual intel Xeon x5355 processors at 2.67GHz

- 16GB of DDR2 RAM

- Two 500GB hard drives in RAID config

- nVidia Quadro FX 4600 768mb video card

But I can expand the RAM out to 32GB and I've already upgraded the RAID drives to an SSD. I can also upgrade the video to 1.5gb or 2gb video card. I'm still impressed how stable its been and how even the 768mb card handles Adobe resources without any lags - I constantly have two or three Adobe programs running at a time which is normally a huge resource hog.

I just fell in love with the performance and the stability of the HP workstations. For a fraction of the cost, and its easily upgradeable so I don't feel like I need to go drop 1K on a new laptop or PC. Also, with such an abundance of these on the market, you can always find a really good deal on them.

Yep, and ECC RAM starts getting pretty important >8GB if you value stability. Workstations depreciate like bricks typically, so a last-gen system can be had for a song if many cores if what you're after. A short while back, some Xeon engineering samples were flooding eBay and were a better deal per-core than whatever the latest i7s were at the time. I forget the model number.

I'm running 32 GB Ram for 5 years without the siltiest stability issue. ECC is probably for really, really big machines 128 GB up or overall an overrated myth.

Yeah, me too. I think if one needs the RAM for running VMs (so, not doing ZFS or running Prod databases), paying extra for an ECC enabled architecture isn't the greatest investment.

Probably the Xeon E5-2630 v4 [1], discussed previously [2].

[1]: http://www.techspot.com/review/1218-affordable-40-thread-xeo...

[2]: https://news.ycombinator.com/item?id=12249317

The Colluseum is still standing with 2,000 year old bricks. They don't depreciate that fast.

Could you tell me what I gain from workstations compared to standard dev boxes?

- Does the number of cores affect general performance or only helps in multitasking and data science? - Is ECC memory worth it (in practice, not in theory)?

Server architecture: more cpu cores, more cache (resulting in better performance per clock period), detection of RAM errors (1-bit and 2-bit) and correction for those (1-bit), built for non-stop operation.

ECC arch. is worth the money wherever correctness of the computed results or integrity and stability of running programs is imperative - scientific and engineering calculations, database systems handling precious information (money), etc.

In practice, it seems ECC in a personal computer is not necessary for most people, but in 2009, a study of DRAM errors in datacenters, ECC was found to be very benefitial:

"Across the entire fleet, 8.2% of all DIMMs are affected by correctable errors and an average DIMM experiences nearly 4000 correctable errors per year"

http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf

Outside of the rare socket 2011 i7's (the only real i7 IMHO) you get more PCIe lanes and significantly more memory bandwidth with xeons too. The extra PCIe lanes can matter for SLI setups..

But this is where Intel irritates me. I want a socket 2011 CPU with both an unlocked (or low core high clock rate) multiplier as well as ECC RAM. For some reason, intel refuses to make such a product.

I was looking for something similar and indeed there isn't much choice in that department. I eventually switched to the high-core lots-of-performance/per-buck E5-2670 especially considering its low price. But if you want high single thread performance, take a look at E5-1650 or, if you have unlimited budget, E5-1680 v2 (12% faster).

The latest v4 parts have turbo frequencies again peaking at 4ghz. Which when compared with the i7-7700, that is a 12% frequency deficit. If you count in a couple percent between haswell & kaby lake, its probably at least a 15% single thread advantage to the part that costs far less.

Hence the unlocked 2011 parts which people routinely run up in the ~4.5Ghz range (myself included) with little issues. The only point of instability seems to be the larger cache, which can have its multiplier independently limited, at which point the stock voltages suffice for a significant clock rate increases. Why intel couldn't bin and sell a 4.2Ghz 2011 part like devils canyon is a mystery.

Thanks a lot. Does it make sense to have more cores vs faster cores if I only plan on using IDEs and doing general development? No video editing or doubling my machine as an application server to serve my personal website.

The DIMM vs ECC point and data you bring up is very interesting. Thanks for that link.

Take this with a grain of salt since this is mostly the marketing and what I've gotten in my own experiences. Of course, this can differ wildly from box to box so take that into account as well.

>> Does the number of cores affect general performance or only helps in multitasking and data science

More cores means more data can be processed at the same time. For a gamer, even if you have 20 cores, it doesn't help you because they're after speed. For myself, it's being able to run several Adobe products doing video editing, photo editing, and graphic design, on all the same machine, at the same time. Think of it as workload vs. speed. Workstations are more powerful and built for heavy workloads, whereas gaming PC's are mainly built for speed.

And yes, the main application is in crunching big data sets, CAD design, or anything heavy on the graphical side of things. I definitely saw a big swing in stability when I switched to a workstation with the myriad of Adobe products I use.

>> Is ECC memory worth it (in practice, not in theory)

So in theory this is the advantage of ECC:

ECC (which stands for Error Correction Code) RAM is very popular in servers or other systems with high-value data as it protects against data corruption by automatically detecting and correcting memory errors. Standard RAM uses banks of eight memory chips in which data is stored and provided to the CPU on demand. ECC RAM is different as it has an additional memory chip which acts as both error detection and correction for the other eight RAM chips.

In practice, I can say it for sure helps your system to be a lot more reliable and stable. I've been running my workstation for several years without a power cycle and its been rock solid, no issues, no hiccups, no nothing. As such, this is the main advantage of workstations is their reliability. I also like the idea in some models of stuffing up to 128GB of RAM or running dual video cards.

In the end, with so many people opting for laptops and gaming systems, you can get your hands on a hell of a workforce PC for under $500. The processors and video card are still dirt cheap (around $100). The only mildly expensive stuff is the RAM. I've really good luck with the HP workstations, but that's just me. I'd say dig around and see what you can find. It's also pretty easy to build one of these from scratch since the parts are so widely available.

Thanks a lot. This was a very detailed experience you provided. Also, is there something I might miss from a workstation focused motherboard compared to a workstation one? Like dedicated audio cards, SLI etc.

I do plan to research before spending but want to get a general feel before going in.

That sounds like a lovely workstation. I had a dream machine of overclocking a dual x5670 xeon machine, but motherboards that let you do that now are too rare and expensive.

What about power usage?

One area where CPUs made significant progress was energy efficiency, especially when idle.

It comes down to how you generate the electricity to run your hardware. Manufacturing processes take a lot of juice and materials.

Reply on news.ycombinator.com