DDR5 RAM prices crashed by 20% in May
tech4gamers.com
tech4gamers.com
So frustrating that max installable RAM usually isn't listed in the specs
Just like a phone I bought years ago only listed a maximum micro-sd capacity of 32GiB even though the right formatting would allow 64 or 128. albeit being 'more wastefull' as the adress space didn't increase.
I was glad for the information.
When you get the factory working well enough that you make a gross profit, you can scale up. The. You reduce waste over time by making little tweaks, and your margins go up a little. But then you need to pay off the equipment, and the research costs, and nobody wants to taper that off in a long tail because what’s the point in making a product you make no money on the entire time, or leaving yourself open for the competition introducing a new product you can’t compete with because you don’t have money saved up for R&D?
So you pay off your loans and refill your war chest and everything after that is gravy. If you’re the first in your industry to get there, you can turn the crews on the competition by lowering your prices first. Or don’t and get sued for price fixing. Dropping the price by 3% isn’t going to get anyone to take a chance on your product instead of someone else’s. You’ve got to do something more dramatic than that. And then you might find out you’re not the first, because everyone else happily matches or exceeds your price drop.
My next rig may be an 13th-gen Intel if DDR5 prices don't come down to DDR4 levels (given that DDR5 isn't any faster than DDR4.)
Today is not the fall. Don't buy memory today for a processor that is not yet available. 'Next Generation' memory prices tend to go through this cycle, and it seems like AM5 will be out right around when the pricing gets pretty sane. If it released in Q1 2023 with DDR4 support, it would be a waste of die space. AMD has tended to keep sockets alive longer than Intel, so a transitional memory choice would be a larger burden.
(Are there any business-oriented AMD desktop PCs with Energy Star or similar certifications, or are those regulations too loose to matter here?)
Unless you use one of their fairly popular APUs, I assume? I got something something like 15 W on idle at the plug for an OEM AMD box. Is that a lot?
I'm not sure if the Bx50 chipsets are any more efficient or if that needs the absolute low end of A series chipsets (that I believe are instead some ASMedia designed IO stuff).
What kind of systems are those? I’ll need something like this soon.
It would be more efficient to replace a computer on a 5 or 10 year cycle rather than a 1 or 2 year cycle, but then you'd have a high end computer for a much lower percentage of the time.
for example, 4790K + GTX 980 is a 8 year old computer. a high end PC from 6 years ago has a GTX 1080 in it, look up how much that costs still
you'd be crazy to throw something like that away instead of selling it for about 400$ or giving it away to friends or family
>It would be more efficient to replace a computer on a 5 or 10 year cycle
if you only use it to check emails and procrastinate on the internet, then yeah. but people who buy 2000$+ worth of hardware usually do something more productive or entertaining than that. and like I said, when they do buy it, whatever they had before is usually sold off to recoup part of the new hardware cost. that's how it worked with hardware for as long as I can remember
you know what does go into landfills? disposable garbage phones, designed to be resistant to repair and maintenance, and made unbearably slow by forced software updates when it's time for you to consume the new product. and I find it very ironic that these zealots we're talking about have a great affinity for the products from that particular company
Look up PUE for data centers. Because I work for Google, I recall, for example, that we build very efficient data centers:
https://www.google.com/about/datacenters/efficiency/
Ie. just 10% overhead.
The mid level 12th gen chips absolutely fly in gaming workloads, it's only if you max all cores for hours on end that the "Sell your grandmother for a MHz" approach Intel currently has makes a difference.
I found good article. https://www.techpowerup.com/289728/intel-z690-motherboard-co...
also in 2022 six hours of battery life is not something I could life with. while you're right that you don't need to sit 6 hours in a single sitting, it could easily be that you don't have an option to charge.
transatlantic flights, for example. (ok, no grass there :-))
With the Intel MBP I had to carry around a huge jackery-style portable battery
I took it into the mac store, they replaced the whole board/cpu unit. No change.
So either I have some other fluke, my software stack has something strange I can't debug (activity monitor only shows apple processes using all the CPU), or people just have different opinions of what fast means.
For context, I would say my 2018 Chromebook is generally faster for most things than my M1.
Rant over. I assume I have some dud computer it's just frustrating.
would that be 8GB M1 and swap process by any chance?
Speaking of chargers, a 45W tiny GaN charger is enough for an M1 MacBook (maybe not for all use cases but I’ve yet to have an issue). I use this one which isn’t much bigger than a usb-a iPhone charger. https://us.anker.com/products/a2664
[1]: https://www.apple.com/shop/product/MK122LL/A/power-adapter-e...
https://www.apple.com/uk/shop/product/MD837ZM/A/apple-world-...
My favorite part of the new 16” MBP is actually the screen. It’s so incredible I struggle to describe it. When moving the mouse it feels more instantaneous, everything is smoother. So don’t FOMO on the CPU, it’s good but not the killer feature.
EDIT: your music link sharing site is neat, be my friend
Perhaps that’s why…?
Also I don't really care about the battery, the amazing thing about the M1 chips is that they just don't get hot.
But the damn thing burns my lap.
My point being that ram prices have been high for the latest years before prices started re adjusting in the last 12 months and that DDR5 currently aren't that expensive.
just because players like f5 chose to ignore the patches and their Linux kernel patch shipped disabled by default doesn't mean they're out of the woods
Most of this is blatantly egregiously false.
And this is also blatantly false. Zen 3 CPUs have a bigger overhead mitigating Spectre than Zen 2 and modern Intel uArchs.
Check Phoronix for performance reviews.
https://www.usenix.org/system/files/sec22summer_lipp.pdf
The first two attack primitives have been patched, sorta, by making those measurements require root access (and making them inaccessible to VM guests). So they're not fixed, but in a properly configured system you won't be able to get at the primitives needed for the attack.
The third measurement, however, remains unpatched. It can be mitigated by turning on KPTI, but AMD doesn't change the default because of the performance consequences. KPTI works by constantly flushing the TLB cache every time you hop to kernel-level code (ie every time you make a syscall), thus creating a "cache barrier" between the two privilege levels - the attack then cannot leak something that is on the higher privilege level, because it's never in cache. This is the same way Intel mitigated meltdown, and it has a large performance impact - every time you talk to the network, or disk, or anything else, you have to flush your TLB cache.
AMD's official advice is "this doesn't work if you enforce address-space boundaries, follow all security best-practices", the subtext being you need to enable KPTI, but they won't recommend the option be set by default, and they've done their level best to shove it under the rug. Which is understandable, for some reason it's entered the public consciousness that "AMD doesn't have vulnerabilities" and it would be very bad for their image if that notion were dispelled.
https://www.amd.com/en/corporate/product-security/bulletin/a...
Briefly discussed and measured in Phoronix here: https://www.phoronix.com/scan.php?page=article&item=if-amd-k...
So yeah, AMD doesn't pay the same price for mitigating side-channel attacks... because they don't mitigate them.
As phoronix discusses, this gets back to the question of whether everyone needs to be mitigating these, we're not all cloud hyperscalers and there's never been a consumer attack using smeltdown-style attacks (it's far easier to just drop some malware and steal the password vault rather than to randomly scan through memory looking for it - you don't have to be the fastest animal in the herd... just not the slowest) but AMD has kinda chosen insecure-by-default here. Sure, that's faster, and you can turn off mitigations on Intel too. Should you?
The odds of an exploit are slight, but that didn't prevent people from taking heavy precautions when this category of exploits was (re-)discovered.
It is better to either wait for the price of ECC DDR5 UDIMM to fall or to buy now a workstation motherboard like the Gigabyte MW34-SP0, which supports the cheaper ECC DDR4 UDIMM.
For a workstation, it is usually better to have at least 64 GB DRAM, than to have less memory, even if it is faster. For example, compiling a large software project in a RAM disk may save much more time than having DDR5 instead of DDR4 would save, so the price difference should be less than 50%, preferably much less, for DDR5 to become attractive.
When JEDEC was standardizing DDR5 they mandated on-die ECC which checks for errors that happen while the data rests on RAM. This was done to increase reliability because of high memory density. This is good and every DDR5 module has it.
However there’s another ECC: the one where errors are checked each time data is transferred between CPU and RAM. This requires the support of the CPU, the RAM, and the mainboard. This is important for data integrity and every computer that runs for multi-day periods must have it. I’ve found by reading papers about it that 1 bit error per 4 GB RAM per 3 days is guaranteed to happen.
If RAM is marked DDR5 it has on-die ECC. If RAM is marked DDR5 ECC it has on-die and transfer ECC. That is what I want and it is not optional at all in my opinion.
This estimate is several orders of magnitude too high. We’d be seeing weird bit flips and file corruption all over the place if this was even close to being true.
I have ECC systems that will report the number of errors detected and corrected. Still waiting to see any errors on my main machine with 64GB of RAM.
At server farm scale, the majority of memory errors come from a very small number of faulty memory sticks. It’s not an even distribution of errors across all memory.
So for my 64GB of RAM I get 2 bytes worth of errors / day. And a few hundred for 2 or 3 months since I didn't restart my PC.
Memory can be assigned to all sorts of things so the impact of a single bit change will depend entirely on what that memory is being used for at that moment.
For example, I load a program into memory, but I'm not running all of the code in that program all the time. Let's say the program includes code for exporting a list to excel, but that's a feature I never use.
Or it has a procedure which uses a local variable as a loop counter. Outside the confines of that loop the variable is meaningless.
Plus its unlikely you are even nominally using all 64g at the same time. The errors might all be happening in unused ram.
Even if it flips actual data, it may end up being no more than a typo in a document.
So yes, a single bit in 4gb might flip from time to time. But the probability of the flip being "meaningful" is less than 1.
There are even worse potential outcomes:
* A single bit error in a stream of compressed or encrypted data stream that renders the stream unreadable.
* A permissions check being yes instead of no.
* Backups being silently corrupted
Saying that say 1 bit per 4Gb will flip every day is a long way from saying that a _significant_ (or even detectable) bit flip will happen every day (per 4Gb). From a simple feel for "what memory is used for", coupled with "how much of the ram is actually used" makes me suspect that it's a very low probability of it being an issue.
Which makes the parent comment, about 1600 bits of data being flipped since the machine last rebooted [1] being both "true" and likely "meaningless".
[1] I'm not sure what rebooting has to do with anything though.
Like isn't this the kind of thing that Turing and Zuse had to deal with?
And if it isn't, when is it a problem that we will finally and absolutely excise?
Is it more of a 2030 thing? 2040?
That is very, very dubious indeed. I can't believe there's no ECC between ram and cpu, it's not credible for anything but a cheap home machine. Any sources?
Most paper characterizing errors are very old (so they used memory chips with large feature sizes), and they scaled their error rates by memory amount (because they didn't know better). This results in massively overinflated error bit rates. It's possible to prove that this is wrong very easily by just writing a program that allocates a large pool of ram, and constantly reads it and writes back to it over a period of time, and then check results after a while. I once did this for an argument on the internet about this -- on my old Intel Sandy bridge system with 16GB non-ecc ram, I allocated a pool of 8GB, wrote ascending integers from 0 to 2^31 to it, and then iterated over it, just reading it in, checking if it's still correct and terminating if it's not, and writing it back out to same address. This program ran for a week with no errors before I had to end it because I needed the machine. 1 bit error per 1.333GB*days is completely not credible.
> . I can't believe there's no ECC between ram and cpu, it's not credible for anything but a cheap home machine.
it is absolutely not wrong: there is no ECC between CPU and RAM for most consumer and desktop computers, and even tons of embedded systems, and that's basically the only bus lacking even basic integrity checks, so there really should be some ECC also there, nearly everywhere.
As for embedded, I'd be surprised if there's ECC at all, never mind on the bus (with the exception of high reliability stuff like spacecraft and factory controllers. Clocks, cookers, microwaves, etc don't count).
Kind of, that's why I was I was saying it is not wrong. IMO though, there should be ECC on everything but the really cheapest computers and if the application even allows it, meaning ECC for business desktop computers, and home computers marketed as using state of the art tech, like Core processors. Maybe you can avoid it in game consoles, or computers marketed for 99% gaming (except they risk being expensive and using good quality processors, so what's the point), but that should be it.
IIRC, the error rate is more like 1 bit error per 4 GB RAM per 30 days (in the relatively recent studies made by Google for their servers).
Nevertheless, that is still not negligible. For workstations and servers with 64 GB DRAM or more, the errors can become frequent. Also, at high altitude the error rate is higher.
Moreover, besides the errors due to radiation, there are errors caused by electrical noise, which can become more frequent in old computers with cheap DIMM sockets, whose contacts have oxidized.
(I actually had this problem in some HP laptops, which had been stored for a long time without being used; being cold made the SODIMM sockets more vulnerable to humidity; after identifying the cause and scrubbing the sockets, the frequent memory errors disappeared.)
I don't know if internal ECC errors ever get reported.
Holy crap the thing flies and the cooler barely spins on idle, if at all.
AMD has announced that their new generation of desktop processors launching in the fall will use DDR5 exclusively, while Intel's current desktop processors can use DDR5 or DDR4 depending on the motherboard. So memory manufacturers should be preparing for a sharp increase in DDR5 demand instead of it just being a premium option.
Furthermore RAM prices generally may have been kept higher in the last two years due to demand from companies and individuals buying or upgrading kit for home working. Even without the migration of many workers back to offices that extra demand would be rapidly cooling off anyway, as those upgrades and new purchases are³ not part of a regular pattern, so this may be an expected correction.
On top of that the tech is no longer as new: production costs will be dropping as processes are refined and output numbers increase. Though I suspect the current drop is more a demand-side thing for several reasons including the above.
Also: early pioneers have been scalped already, the next tranche of buyers may not bite until there is this sort of price drop.
[1] why bother if other parts of your system won't keep up well enough for you to see the benefit?
[2] fuel, food, interest rates rising in some countries affects those with a mortgage, et cacas: fancy tech is a luxury, people may put off such purchases until the costs of essentials settle a bit
[3] hopefully!
I also saw a rumor that the power management IC used by DDR5 DIMMs was in short supply.
If I can, sure. Glad for the middle ground option though. e.g. I'd rather have 48gb slightly faster kit than 64gb for example.
Other thing is that I've got a lot offloaded onto a home server already so less need for GBs on desktop
Of which, what fraction are counterfeit, floor sweepings, rejects, or stolen? That's too many resellers.
Good for this year's laptops and whoever decides to go with Zen 4
BL and DFE are shortly on https://en.m.wikipedia.org/wiki/DDR5_SDRAM
Now, about DRAM latency... perhaps DDR6 can have an I WANTZ IT NAO signal.
B) outside of crypto assets, 20% downwards in a month is a crash in every asset class
When a product in demand moves from prototype into mass production, the price drops rapidly. That isn't a crash. The expected price is to get close to DDR4.
for example, many financial publications will write the exact same thing for a 2% intraday price movement
there is nothing to explain or rationalize or correct, its hyperbole
I think its interesting, to know about prototyping to mass production, but is or isn't a crash? okay. shrug. its a waste of time to split hairs over that, its hyperbole.
It's fine that you don't care about the difference, but for people that aren't particularly in touch with the computer components market it's very useful to point out the difference between a stable price plummeting below the norm and a brand new price plummeting toward the norm.
In the context of semiconductors, isnt everything inflated and would plummet toward a historical norm with little regard to newness? for now
A few people will buy the 2022 Ford-Model-X for a premium at launch, but once they've bought it, Ford are just selling to the people with a 2010 Ford who just want to upgrade to what's in the new catalogue.
I'm sure some of those upgraders will pick up a bargain on that onsold 2020 stock - but Ford aren't devoting resources to keep making those old models.
If oranges drop from $5/kg to $4/kg I am not calling it a crash.