SK Hynix has finished the first DDR5 chip

waitfags win again

except they never win because they aren't even playing

Attached: arg linuc.png (512x512, 51K)

gooks trying to maintain their monopoly
fucking disgusting

Three companies maintain like 95% of the market share, and the other 5% is held by like 5 taiwanese companies.

chinks will soon take over.

...and crash the market. Can't trust any slants.

>DDR5
DDR3, and DDR4 hasn't even been available in speeds/densities that takes advantage of their capabilities.
Everyone will again be forced to upgrade for no reason.

cheap ECC DDR3 yay

Anything in the standard about rowhammer mitigation?

>just buy ECC ram (they don't have memesinks) and build a TR pc.
They can, the 32GB DDR3 LRDIMMs i bought all came with them.

Processors are so slow faster ram isnt going to speed anything up. Lower power is good for craptops though

Because the heat density will probably end up being the same

This, also overclockers gonna overclock.

We haven't needed heatsinks on RAM since DDR3. Even being generous at 4W a stick for 8GB DDR3 with 16 ICs your only talking 0.25w per IC if that. This is even less for DDR4 but yet we still see fuck huge heatsinks on anything remotely high clocked that do nothing but add cost and get in the way of air coolers

>We haven't needed heatsinks on RAM since DDR3.
LRDIMMs still need them you tard

Attached: Screen Shot 2018-11-15 at 10.03.17 PM.png (2454x1090, 326K)

price starts at $18999

Its not bad for being compliant with the finished JEDEC specifications, but Samsung half a year ago had DDR5 DIMMs that were 5500mhz at low voltage, and reached 6400mhz standard voltage.
The spec is going to end up scaling pretty damn high, and density should be quite high as well. ECC DIMMs for enterprise will be a huge improvement over the paltry DDR4 currently in use at least.

If there standard fucking height DIMMs there is nothing wrong with that. I'm just tired of monstrosities like this and just want RAM that's guaranteed to clock high but without being obtrusive

Attached: 51Xe-wA8wIL.jpg (500x500, 32K)

No no no. It means we never lose.

Attached: 1515375000970.png (123x136, 39K)

>5500mhz 6400mhz
How is memory able to operate at such high frequencies when cpu gets unstable? Isn't there operating frequency limit on silicon that can only be broken with graphene?

Its DDR, double data rate. It isn't actually truly operating at 6000mhz~, thats just the effective speed because its signaling twice every full cycle. The actual DRAM would be operating at 3000mhz, just as a 2400mhz DDR3 or DDR4 kit is actually clocked at 1200mhz.

oh, right. I forgotten about the ddr part. So is probable that we'll be seeing memory operating frequencies hitting a wall too on ddr5.

Its certainly possible. It should be relatively easy for DDR5 to blow right past 7000mhz effective clocks as it starts to mature. Maybe we'll see 8000mhz kits a couple years into the life cycle.
There are a lot of nuances to complex signaling though. DRAM being a much simpler type of circuit than logic lends it to being less limited, it could potentially have no problem passing consumer processor speeds.

Though I'd think future memory standards will probably evolve to use a much wider IO standard and move away from conventional DDR.

based gooks

Gamur retards that's why

HBM v DDR wars have been going on for a while. At this point I think the DDR standard might evolve in that direction but my opinion is that the industry working group has too much power to be supplanted by any other contender except in niche and pre-assembled solutions.

HBM still is DDR, its just a different PHY, interface, and packaging. By its designed it can't signal across long distances like its defunct competitor HMC could. Though a standard more like HMC could be a replacement to conventional DIMMs.
Where HBM really made a lot of headway is in how compact its PHY is for its insane width. Saving on die area and having lower leakage while enabling more bandwidth is what everyone wants.

Knew I still had this picture saved somewhere.
28nm Fiji compared to 28nm Tonga. GDDR5 PHY basically is DDR3 PHY, theres very little different between them besides how the memory itself behaves.
A full 1024bit HBM PHY is only slightly larger than a 64bit GDDR/DDR PHY.

Attached: 05065715663l.jpg (1358x789, 397K)

where's PCIe 5.0?

good thing i just got 16gb of ddr3 for $45. feels god man.

nice.
I might buy a few first gen DDr5 ram's 64GB
since i'm already on 32GB DDR4 for like 3 years now lol

Memory manufacturers can suck my dick.
Yes you heard me well.
Fucking cartel.

The fuck are you on about, processors work much, much faster than memory can be accessed, anything to reduce memory access times is a huge win for performance. Hence why there's not so much attention on muh ghz these days and much more on cache sizes and prefetching, the processor can only work as fast as you can feed it data.

>need to get a new board now
>no fucking money
>server still runs DDR3
>server STILL has 2 borked RAM ports
fuck this technology is gay

Decreased voltage or decreased actual power?

Power = Voltage*Voltage*dynamic capacitance* clock speed

DDR5 7000MHz for SUPER FAST GAYMEN

CL45 at least

>and crash the market
I'll take anything for the korean ram mafia being fucked sideways and to have affordable memory

they will crash the prices of RAM/SSD which is great.

then you went out of your way to get the gayest EEC ram

I skipped DDR3 entirely. Actually if you count server, my server could be going from 2 to 5

>I skipped DDR3 entirely
so none of your devices use DDR3?

No ur wrong

When is the general availability?

Would you say that DDR4 has had a good run over all?

Never it will always be in a (((shortage)))

>DDR3, and DDR4 hasn't even been available in speeds/densities that takes advantage of their capabilities.
>Everyone will again be forced to upgrade for no reason.
DDR5 fixes this.

WAITFAGS ALWAYS WINS BABY
SEE YOU IN 2020

Yeah, Heat density goes up every process shrink. the "lower power consumption" stick that they show off everytime is just a result of having to keep heat density in check for servers.
You think that is a heatsink? that thing keeps the entire stick equally hot. it doesn't cool the thing.
Fuck those things.

weren't we already at 4 data pumps per clock cycle?

>DDR5 fixes this.
Unironically yes, it might help alleviate the RAM crisis.

I got 32GB of DDR3 for $200, the price for an equivalent DDR4 set would be $600, so fuck DDR4. Memory prices are the main reason I've had this system for 6 years.

>muh CL
No matter how shit the CL is for DDR5, it will consistently BTFO previous DDR revisions

DDR3 went as low as 8ns real latency.
Best DDR4 goes as low as 8ns latency.

It's all bullshit barring for lower voltage with each generation.

>Your dumb MHz alone don't mean shit
>Your dumb CL alone doesn't mean shit

>Have tons of free recycled DDR3 4+ GB ram
>Have a few of free recycled Gen3 CPUs 3.2+ GHz (i3, i5)
Got my gaming PC with a i5-3570 running at 4.0GHz OC water cooled with with 16 GB of DDR3 ram. I am more than happy with this setup.
DDR4 was a scam and DDR5 better go cheap otherwise I am waiting for DDR6.

>16Gb
Does this mean 16(8x2) will be the norm now, like 8(4x2) was? I doubt we're going straight to 32.

RAM never aimed for better performance or better latency. Only additional channels and shorter path traces could do that. They aim for more and more storage at the same characteristics. for us consumers that means lower wattage on similar dimms, for servers that means larger capacities. Alot of workloads out there, like caching servers, can't rely on HDD or SSD storage.

Ram prices are like ma cock in highschool, keeps going up and down motha fooka.

fuck you all and suck my benis baby.

Believe us, when it's the Chinks' turn things will get awful for everyone.

DDR4 has come down a bit in the last 2 months

Attached: trend.ram.288dimm.ddr4_2133.4x4096.d40e3550f2988cdf754b39aee2f2bc80.png (1134x348, 27K)

>Still waiting for $100 16GiB era

Crypto's and other high density chips seem to be on the decline. Hopefully the server market doesn't crash the prices now.

>meme keyboards
dude i literally need at least 7k rollover for osu mania

>unironically being this fucking retarded

can you not read?
>mass production in 2020

4000mhz ram is no faster than 3400mhz ram 99% of the time. You cant refute this

>4000mhz ram is no faster than 3400mhz ram
speed is not everything you retard. latency matters.

Here is the fun thing. When you have a huge jump from DDR1 to DDR2 it is great. When you have a huge jump from DDR2 to DDR3 it is great.
When you have a jump from from DDR3 to DDR4 and its meh. Now is not the time to go price fixing for profit. Not a huge support, no where close to DDR 2 or 3.
DD3/4 to DDR5 is still.. ehhhh.. level. Prices still high. This hurts everyone but the customers. This only hurts business bottom lines. This is how you shoot yourself in the foot without a gun. Complain no one upgrades and try to blame everyone but your greed.

Which again doesnt matter except on poozen because its shit

nah, it matters if you use anything that is fast enough to use it. poorfag.

There was 1066mhz DDR2.
DDR3 launched and 1333mhz were standard fair for years, we only really saw availability of 2400mhz a year or so before DDR4 went into production.
There are now low voltage DDR4 4266 chips in production, being used in devices.
Even at launch of DDR4 there were 3000mhz+ kits available at sane prices, the only hiccup at all with DDR4 was sloppy baseline JEDEC standard 2400 kits.

Funny how retards seem to not remember things correctly.

And yet prices went up. Seems how retards tend to forget how things were priced.
crucial.com/usa/en/memory-performance-speed-latency

Name an intel process that benefits from low latency fast ram beyond cl 15 3200mhz ram amd poorfag retard

anything that is not made for manchild gaymers like you

you know, actual computing tasks for actual work.

Name one retard youre talking out your ass

>gaymer manchild needs to be spoonfed about the most basic things
yikes!

Thanks your proving you dont have a single fact to back up your false arguements.

cry me a river, /v/edditor.

No one mentioned games, project more retard. Compiling, rendering, encrypting, encoding, transpiling, simulating, compressing, decompressing, none of these tasks give a fuck about ram speed beyond cl15 3200mhz

>foaming at the mouth gaymer

>gaymur
>rather buy LED cased RAM than ECC
Why are they so dumb. Don't they know that fast collision mitigation is better than cased LED RAM?

>.. a jump from from DDR3 to DDR4 and its meh
>.. DD3/4 to DDR5 is still.. ehhhh.. level.
Performance based analysis. I'd argue against. I'm sure there are systems with ddr3 and ddr4 support and you would guaranteed see way better performance when you buy high end ddr4 versus ddr3. If bought pre-market crash probaly around the same price.
>Prices still high
Yes, a different issue. new nodes have always become cheaper for the same performance; eventually.
>This hurts everyone but the customers.
What?
>Complain no one upgrades and try to blame everyone but your greed.
everyone "upgrades", that got us into this situation. If the Dram market isn't pricefixing, which is entirely plausible, they will use the volatility to secure a higher profit margin and build new fabs. This is not pricefixing and you commies dont understand that a multi-billion dollar investment needs proper projection of profits. if the crypto crashes hasn't taught you anything, nothing will.

This. new ram technology reached the market exacty right each time. deploying new tech too soon usually sees regression in performance, too late and you get huge boosts that you could have brought earlier. RAM is more lenient in that matter anyway, right bulldozer?

>This one metric proves me right
Your own metric shows ddr4 being better than ddr3 in every way.

even your moved goalpost of ddr4 costs versus ddr3 doesn't make sense because ddr3 has also increased in price.

>TFW, your generation had to adjust for RAM inflation.

I haven't been able to find anything about this and rowhammer type vulnerabilities.

I remember back when rowhammer was a major concern, DDR4 was already coming out and was not a fix for it. Also ECC wasn't a complete fix either (ECC made it less likely, but it was still possible).

I found a Defcon talk where the guy was able to map out the physical layout of the RAM modules, and using rowhammer-like repeated writing overwrote some memory in the host machine from inside a VM.

Quite a serious break, and something that might require a revision to the standards, but distressingly not much discussion of this recently.

Rowhammer only affects availability and integrity. I know AMD has an option to encrypt RAM. dont know wether intel has one per default.

The fix is also more complicated than it first looks. it would require complex logic or expensive ddr modules to prevent. OS level mitigation is probaly going to be implemented somewhere sometime.

The cheapest (2x8GB) kits are about 100 bucks right now while only 2 months ago they were about 135~

newegg.com/Product/Product.aspx?Item=N82E16820158229&Description=16GB RAM DDR4&cm_re=16GB_RAM_DDR4-_-20-158-229-_-Product

what about latency? Isn't that the big disadvantage to DDR4 compared to 3?

Market crashes lead to subpar shit at high prices. China sympathizers can GTFO.