7 ms·
Intel's staunch refusal to support ECC RAM on non-Xeon chips seems incredibly chintzy at this price level. Who is the customer that's simultaneously willing to
by smacktoward 9y ago
Intel's staunch refusal to support ECC RAM on non-Xeon chips seems incredibly chintzy at this price level.
Who is the customer that's simultaneously willing to spend $2,000 on a CPU and so desperate to save a few bucks that they'd settle for non-ECC RAM? I want to meet that person.
- xchaotic 9y agoI'm running VMs in a LAB and I need multples of 6 virtual machines so these high core count CPUs are exactly what I'm after and I don't mind the occasional bit flip as it's for testing only, not production. But I agree supporting ECC RAM wouldn't add much to the cost of the CPU itself.
- detaro 9y agoI think these things are primarily aimed at gamers and overclocking enthusiasts. They don't care about ECC at all, or might even worry that gets in the way of overclocking (since at least for now, nobody optimizes ECC DIMMs for that) Those who care are supposed to spend even more on Xeons.
- snuxoll 9y agoNobody is doing enthusiast overclocking on these, they run hot enough at stock to need AT LEAST a 120mm closed loop cooler.
- Filligree 9y agoDon't underestimate the enthusiasts. Someone will overclock 'em.
- snuxoll 9y agoI mean, I guess they WILL - not many people invest in HEDT boards and CPUs to not milk every last bit of performance they can for their dollar, but even with a 360mm rad and a high flow pump you can’t push these very far, the thermal and electrical window just isn’t there.
- detaro 9y agoYou're right, most OC fans probably will look towards fewer cores they can push harder. Didn't really think about that.
- jonathanstrange 9y agoI've got a nagging question about this, since this topic comes up again and again and I'm planning to buy a new machine soon after having run an overclocked i7 920 for many years: How important is ECC RAM if I want 32 GB or more RAM? What are the consequences of not having ECC RAM? I'll probably go for AMD this time anyway, but was wondering in general how important that is. Are non-ECC machines known to crash? Wouldn't that be a factory design flaw that would need to be fixed, at least according to EU consumer protection laws?
- Filligree 9y agoTypically what happens is silent corruption. Rarely, very rarely you'll get a crash, but even if a bit flips inside of program code there's no guarantee that that code will ever run again -- and if it does, the malfunction might not be obvious. Most memory is non-program data, however, and most non-program data won't cause obvious errors if corrupted. A picture that renders incorrectly, a broken model in a game, a single burst of static in a video... It's insidious precisely because it's hard to notice, and corruption never gets better over time.
- blacksmith_tb 9y agoThat certainly makes sense for files, which are read into memory and written back to disk, potentially corrupt - but I wouldn't expect most compiled code to be written back to disk? (I am still on an old cheese-grater Mac Pro, so my 32GB are ECC which means I'm not worried personally, just curious).
- vbezhenar 9y agoAnother important advantage of ECC RAM is early detecting of malfunctioning memory. I've yet to hear from anyone that he's regularly running memtest. While cosmic rays might be rare event, bad memory stick is certainly possible and it'll produce much more errors than cosmic rays. Some people live with it for a long time, cursing their blue-screened OS or "access violation error" programs. It's hard to distinguish software bugs from hardware bugs. ECC will let you know about it immediately.
- RubberSoul 9y agoIs there any research or even anecdote from people running lots of machines that indicates how frequently corruption occurs? Suppose I'm doing scientific computing, is ECC solving a super rare problem, or something that might happen weekly or monthly?
- sireat 9y agoI believe there was data that with 64GB of RAM you can expect at least a single bit error or two a month. Thus ECC would be super helpful to those that expect that most of the 64GB will be filled with bits that shouldn't be changed. This article by a system integrator is a pretty decent data point: https://www.pugetsystems.com/labs/articles/Advantages-of-ECC-Memory-520/ https://www.pugetsystems.com/labs/articles/Advantages-of-ECC...
- smitherfield 9y ago>>>I believe there was data that with 64GB of RAM you can expect at least a single bit error or two a month. But this would be if you were using 64GB nonstop for a month, which is quite a bit rarer than just having 64GB. And certain filesystems (ZFS, Btrfs) and most databases have software checksumming functionality to detect bitrot. Not to say ECC isn't useful or multiple layers of protection aren't prudent, just that the problem it solves is something you're unlikely to encounter below datacenter scale.
- ljcn 9y agoSome research yes. To answer your final question directly - more like hourly. Google did a big study in 2009 [0] and found error rates between 2.4E-11 and 6.7E-11 errors/bit-hour, or, in more familiar units: about 1 error per 2-GByte of memory per hour. Obviously most errors go unnoticed, but for anything where random bit flips are unacceptable, ECC is necessary. Note too that error rates go up with memory density, not down, so things might be worse with more recent chips. [0] http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf http://www.cs.toronto.edu/~bianca/papers/sigmetrics09.pdf
- 9y ago
- technofiend 9y agoThat and refusing to up the number of directly connected PCI lanes. The latest chips support 40 lanes but some of them are outboard of the CPU and connected via a controller. High speed cards like a 10 gigabit Mellanox can use 8 PCI lanes all by itself. Throw in some M.2 devices and you're out of directly connected lanes.
- CyberDildonics 9y ago10 gigabits is about 1.2 gigabytes. PCIe bandwidth is about 1 gigabyte per lane. A 10gb card does not need 8 PCIe lanes.
- kogepathic 9y ago> PCIe bandwidth is about 1 gigabyte per lane Per lane (unidirectional) bandwidth by PCIe revision [0]: PCIe 1.0: 250MB/s PCIe 2.0: 500MB/s PCIe 3.0: 985MB/s PCIe 4.0: 1.969GB/s PCIe 5.0: 3.9GB/s > A 10gb card does not need 8 PCIe lanes. It depends on the PCIe revision(s) the card supports. If the card needs to support 10GBit on a PCIe 1.0 link, then it will need an x8. The bandwidth of PCIe 1.0 is ~250MB/s per x1 link, but there is protocol overhead and 4 lanes doesn't provide enough bandwidth, so you'd have to use an x8 (since x6 keyed cards are not a thing). The most popular cards with enthusiasts are refurb Mellanox cards, which are PCIe 2.0 but support PCIe 1.1, so they need to utilize PCIe x8 to provide full speed 10GBit: PCI Express 2.0 x8 (1.1 compatible) [1] [0] https://en.wikipedia.org/wiki/PCI-Express https://en.wikipedia.org/wiki/PCI-Express [1] http://www.mellanox.com/page/products_dyn?product_family=20&mtag=connectx_en http://www.mellanox.com/page/products_dyn?product_family=20&...
- CyberDildonics 9y agoPCIe 3.0 was finalized a decade ago. Equating old cards to a technical limitation is ridiculous.
- technofiend 9y agoFunnily enough it was the mellanox I had in mind.
- bsenftner 9y agoLook directly at any accountant trying to shave expenses, and then at your manager or your manager's manager being pressured by that accountant.
- eloff 9y agoI did that, more or less. I only dropped $700 on a 10 core Xeon CPU for my workstation, which does support ECC, but I went with 128GB of non-ECC RAM because the price for the ECC stuff was ridiculous. I wish I'd known about AMDs chips at the time though. I suspect my next machine will be an AMD, which hasn't been true in a decade.
- ComputerGuru 9y agoThere is barely a premium on ECC RAM these days. And it's ironic you bought so much RAM since the more you have the more important ECC gets.
- eloff 9y agoI got 128GB DDR4 for $600. You can't even get 128GB ECC DDR for double that a year later. I was a bit concerned about memory errors as well, but my workstation has been rock solid.
- ComputerGuru 9y agoI'm not sure how many slots you have, but the MSRP for 2x16GB DDR4 is $250, though I've seen (and got) them lower.
- eloff 9y agoUsing 8x16gb sticks. But even at $250 for a pair, that's $1000. The choice was between 64GB ECC and 128GB non-ECC and I decided to live dangerously. So far, no regrets.
- drzaiusapelord 9y agoIts not even ECC its also multiple processors. If I want two i7's, too bad, I also need to buy Xeon. I think this is a way AMD can differentiate itself. I don't think anyone is selling dual socket ryzen boards either. Last I heard you'll need to pay for Naples for that. These are odd moves to make during the long tail of Moore's law. Not sure if protecting the server market from a small amount of overly cheap buyers putting workstations into server use is worth hobbling everyone else and forcing them to pay the Xeon and Naples premium. No idea who is spending $2k on this either, especially when the 1800X is $449 right now. Whale gamers I suppose.