DRAM prices to keep increasing through 2017

[cbn]

http://ein.iconnect007.com/index.ph...rice-trajectory-through-2017/103962/?skin=ein

^^^ I'm surprised to find out this is what is expected to happen.

P.S. Looking at Newegg prices right now I see 2 x 4GB DDR3 1600 starting at $50 and 2 x 4GB DDR4 2133 starting at $55.

 

[nopainnogain]

I was expecting this.

Thus, I will stick (no pun intended) with DDR3 until 2018 or beyond.

 

[Elixer]

Thing is, the 'yield issues' will be fixed, then, they will have a flood of chips flying out.
When that happens is the question--and it shouldn't take more than 6 months to fix, unless, there are some really serious issues.

 

[w3rd]

collusion...

 

[sandorski]

Because of this, I will likely not upgrade until next year. Not gonna pay $150(cdn) for 16gb 2133 DDR4 when 3200 should be about $100.

 

[harobikes333]

What's a guy supposed to do if they want to build a new AMD rig when VEGA hits...?

 

[Bouowmx]

Get the DRAM and SSD (flash memory is also affected, to a smaller degree) you want now.

 

[harobikes333]

Get the DRAM and SSD (flash memory is also affected, to a smaller degree) you want now.

Easier said than done...? AMD AM4 motherboards are supposedly finicky when it comes to which RAM they like?
I'll dig into it though . thanks!

 

[.vodka]

Easier said than done...? AMD AM4 motherboards are supposedly finicky when it comes to which RAM they like?
I'll dig into it though . thanks!

The problem with Ryzen appears when you try to overclock memory past what's officially supported. This is required to extract maximum performance out of the CPU, so it's necessary. You want Samsung B-die based memory to reliably hit high memory speeds. That's DDR4 3200 14-14-14 stuff and higher, those specs pretty much guarantee you that kind of memory. Of course you have to do some research before buying so you are sure.

AMD is going to release another big AGESA code update in May that should increase compatibility with other memory makers based memory and high speeds (> DDR4 2933), so watch out for that and the corresponding BIOS updates from each manufacturer as they implement the update, and people report on their other memory kits that are currently problematic.

Either way, you can't go wrong with said B-die stuff for Ryzen.

 

[harobikes333]

The problem with Ryzen appears when you try to overclock memory past what's officially supported. This is required to extract maximum performance out of the CPU, so it's necessary. You want Samsung B-die based memory to reliably hit high memory speeds. That's DDR4 3200 14-14-14 stuff and higher, those specs pretty much guarantee you that kind of memory. Of course you have to do some research before buying so you are sure.

AMD is going to release another big AGESA code update in May that should increase compatibility with other memory makers based memory and high speeds (> DDR4 2933), so watch out for that and the corresponding BIOS updates from each manufacturer as they implement the update, and people report on their other memory kits that are currently problematic.

Either way, you can't go wrong with said B-die stuff for Ryzen.

Thanks for the informative post. ... how exactly can a person tell if it's "Samsung B-die based memory" ?

 

[.vodka]

A quick google search returned MSI's memory compatibility list for one of their motherboards. It would seem they've been doing quite extensive testing.

https://www.msi.com/Motherboard/support/B350M-MORTAR.html#support_support

Look for the memory kits marked Samsung. Those are the ones you want.


For the rest of the kits in the list based on other memory manufacturers.. let's hope the May update gets them working better at high speeds, just like the Samsung based stuff does now.

 

[billbobaggins87]

What's a guy supposed to do if they want to build a new AMD rig when VEGA hits...?

Looking like by the time Vega shows up dram process will have cooled down. And hasn't shown anything good on drop date.

Sent from my Nexus 6P using Tapatalk

 

[Malogeek]

Thanks for the informative post. ... how exactly can a person tell if it's "Samsung B-die based memory" ?

If you have to sell your first born to get it, likely it's Samsung b-die.

 

[JimmiG]

Thanks for the informative post. ... how exactly can a person tell if it's "Samsung B-die based memory" ?

Generally it's 3200 C14 or faster (and more expensive), where as 3200 C16 and slower is usually Hynix.

 

[IEC]

If you have to sell your first born to get it, likely it's Samsung b-die.

The trick is to buy the Ripjaws V DDR4-3200 CL14 when it's on sale. Or the Galax HOF. Or other lesser known SKUs that use Samsung B-die ICs.

I barely paid a premium for my B-die.

 

[BonzaiDuck]

I should start my own thread of inquiry, but I'll see if the interested parties here can offer some advice.

My sig Skylake has 16GB of TridentZ 3200 14-14-14. It was great: just pop them in, set the XMP profile in BIOS, and they tested out flawless with HCI Memtest 64.

This is -- for just about everything -- plenty of RAM. In fact -- after trial-testing Romex PrimoCache back in 2014, I've been using it ever since. And I've used some of the RAM for that -- as much as 4 to 6 GB leaving 10GB for everything else. The caching program caches to SSD as well as RAM. But I've begun to wonder what sort of performance I'll get by expanding the RAM allocated to the caching.

So for months, I've been conflicted about (a) getting another "identical" TridentZ kit, or (b) replacing this kit with a set of 2x16 = 32GB of 3200 14-14-14.

I contacted G.SKILL tech-support, and they're always good about getting back by e-mail -- great folks there.

"If I had an identical kit, will I be able to run them without flaw at XMP DDR4-3200 14-14-14?"
"We can't guarantee that. We match out 4x8=32GB kits to assure it. But we can't guarantee that running the two [identical] kits will make that possible."
"What about dropping the speed to DDR4-3000 and the same 14-14-14 timings?"
"We could work with that!" [Meaning -- they could support it for RMA or whatever purposes. G.SKILL has always been stellar for quick replacement of defective sticks.]

That's option "(a)". Option (a) will cost me $180+ at today's prices throwing in tax and shipping. Option (b) will cost me north of $324.

I consider that I have to test the whole enchilada, either with option (b) or option (a), but I'm sure I can get it done running HCI's program in Windows. [I used to prefer using the self-booting CD version of the test program -- stupid, I know. . . ]

WHAT SHOULD I DO? [and why should I do it?]

With a system like this, there is no hurry at this point except for the topic of this thread. But I'm "curious" about the extra RAM possibilities, no longer worried about writes to SSD from hibernation files when Win 10 creates a "Full" hiberfil.sys of only 40%, and no longer worried about writing pagefile when auto-management by Windows doesn't choose more than ~3 GB for pagefile size.

 

[nopainnogain]

Considering that option (a) is somewhat risky and option (b) is expensive... The best option (IMO) would be to wait. I mean, if you can cope with waiting...

Having said that, option (b) is (again IMO) a better solution.

 

[BonzaiDuck]

Here's some more I can think of.

I could buy a 16 GB set of TridentZ DDR4-3400's with a matching product code. If the parts were similar enough to give a chance of a match, I could then set the speed to 3200, and expect the timings to settle in to the loosest possible denominator. As a fallback to that, I could probably clock them at 3000 14-14-14. But that poses more risk than an identical kit. G.SKILL felt pretty confident that a double pair of 3200's would work at those latencies for 200Mhz less. The latencies would be identical, and at the lower clock speed, it should all work. In fact, I could change the granularity of the bus-to-memory ratio, and try downclocking them by something like 33 or 66 Mhz.

Then I would only have spent the ~ $170. Eventually, after testing them that way and even using them, I could see what happens at 3200.

On the other hand . . . .

Well, who else? What would you do?

What does waiting give me? How long would I have to wait for prices to go up, and then down again? For the double-kit or the double-gigabytes, like South Park's Butters Stotch, I could be bi-curious or simply try-curious. For the 2x16's, it would be a more confident Blackjack-double-down.

This is like having an itch in your brain . . . And I can't decide how to scratch it . . .

 

[EXCellR8]

I'm still not too happy about shelling out for more DDR4 on the X370 platform, especially since the whole compatibility thing, but at the end of the day I've made wayyyyyyy worse purchases.

and yes, pretty much all PC-related

 

[Fir]

Yes it is! I was paying $169 for 32GB Crucial kits (16x2) DDR4-2400 back in December and now they're $239.

 

[Ajay]

Annoying indeed - darn cell phone gobbling up DDR4 wafer production
I know new NAND capacity is coming online 2H2017, but don't know about DRAM production.

 

[Hans Gruber]

Until the justice department shakes down the memory manufactures again with fines probably north of $1 billion this time around. All you have to do is google memory price fixing. Several years ago all the big memory makers paid huge fines for price fixing.

 

[harobikes333]

Micron: Goldman Cuts to Hold as DRAM Market Set to Cool
On Barron's website it appears Goldman Sach's thinks the DRAM prices will be cooling down.. real question is WHEN..

 

[Ajay]

Well, based on what Mark Delaney and other's have said, prices for DRAM & NAND will start sliding back down in 2018 [stabilization of demand and added capacity for newer memory technologies). To do better than that, you need a crystal ball.