Does this even affect real world performance for a normal person (as in not for work related activities)? I've never been able to detect much difference in a gaming scenario.
"Does this even affect real world performance for a normal person (as in not for work related activities)? I've never been able to detect much difference in a gaming scenario."
If you're using integrated graphics? Absolutely yes for games. Otherwise probably not.
That's the wrong question to ask in the context of Haswell-E and X99. Anyone buying these is by definition not a normal person. Ask again once DDR4 hits meinstream platforms. For iGPU it will provide obvious benefits, whereas for regular operation you'll probably see a modest performance increase, equivalent to 1 or 2 CPU frequency bumps (it's been like that for DDR2 and 3 as well).
Using Anandtech's "Performance Index" (speed divided by CAS latency), this rates fairly low. Will all DDR4 start out this way or are we looking at a budget part here?
It will probably all start out that way. DRR3 had high latencies when it first came out, DDR2 did to, and DDR1.... Actually, I don't remember how DDR1 compared to SD RAM anymore, but it certainly got better as it matured.
That should be inevitable. The ram manufacturers have years of experience tweaking parts for the old standard; but have only just figured out how to make the new one in mass quantities.
It will absolutely get better with age. Think of this more like DDR3 800; it was a thing for a little while at launch, but was quickly bumped up to 1333 with lower latency, and then 1600+ as things went forward. We will have the better part of a year to get things a right before mainstream platforms start using this, so I would not worry about it.
As others have said, it should quickly get better. But there's more: DDR4 is supposed to be a bit faster at similar specs due to improved efficiency. Let's see how this works out!
Furthermore: this is for businesses who wouldn't run current Intels at anything higher than DDR3-1600, because they're not specified for anything faster. and if they can keep up DDR4-2133 15-15-15 for 16 GB modules it's actually not that bad.
Pricing's the most important part, the cost will be absorbed by system integrators regardless but whether there's price parity or an increase will determine how quickly enthusiasts adopt a platform that uses DDR4.
I imagine everyone will soon slap spreaders on modules regardless of whether it's needed or not, tho if you're making bland green DIMMs you might as well.
performance memory at 1600+ could probably use at least a basic spreader, especially when you don't know what kind of environment it is going to be installed in. I am a firm believer in AC in by house, but some crazy people somehow enjoy 80*f temps, and computer parts may not deal so well for them. So better safe than sorry I suppose. Still, with modern processing, lower voltage, and a cycle time, I can't imagine these getting all that warm.
These are honestly just garbage sticks. I see no reason anyone would buy DDR4 sticks that: 1. have only 8 GB for a DDR4 set 2. provide only 2133 MHz bandwidth for DDR4 3. add to this worst you can get CL15 timings 4. also these are not even LPDDR4/notebook oriented modules so you can't excuse the specs
What we have here is a product that is targeted at desktops, but every DDR3 2133 set outperforms it. The only advantage of this set -- low voltage -- doesn't matter for a desktop PC.
Maybe because hammering those DDR3 2133 set sticks into a platform that uses DDR4 won't work? Considering a Haswell-E here but might skimp on the RAM to buy a decent set later.
If it was possible to make DRAM itself dramatically faster, we'd never have gone down the whole DDR1/2/3/4 cycle with each generation roughly corresponding to doubling the databus speed and doubling the on-dimm parallelism to feed it. PC-133, DDR1-266, DDR2-533, DDR3-1066, and DDR4-2133 all have the dram running at the same speed; and just use faster/more parallel controllers to read/write to more locations within the ram chips at once.
This is also why we should expect to see DDR4 speeds ramp up fairly quickly; as the engineers get more practice in building faster controllers; the actual dram has plenty of headroom before it catches up with the speeds at which it's been made to operate in DDR3.
With equal clockspeeds and lower timings, wouldn't the DDR3 example above outperform these DDR4 modules? What's the point in upgrading for a desktop user?
A rule of thumb in processor power consumption is that the max load power is proportional to the clock frequency times the square of the voltage. I promise you that these sticks use much less power than their ddr3 equivalent. That being said they're pretty low performance and in a HEDT environment the tiny amount of power that ddr3 isn't an issue. It matters more in server blades with hundreds of sticks or in laptops were every mW counts.
Is it just me or does there seem to be a lot of wasted PCB space? Would it really cost that much more for them to just make the PCB shorter? Would be nice for those big tower heatsinks
I think the vertical heights for the PCB are set in the DDR4 spec itself; even if they don't need the space internally for routing connections in the PCB the height may have been selected for mechanical reasons. If they kept the length the same and but made them shorter they'd become more prone to bending or breaking.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
28 Comments
Back to Article
vLsL2VnDmWjoTByaVLxb - Thursday, July 31, 2014 - link
"One of the hot topics in computer upgrades for the next couple of years is going to be the move from DDR4. "Think you mean "to"?
Ryan Smith - Thursday, July 31, 2014 - link
Indeed. Thanks for pointing that out; it has been corrected.invinciblegod - Thursday, July 31, 2014 - link
Does this even affect real world performance for a normal person (as in not for work related activities)? I've never been able to detect much difference in a gaming scenario.lagokc - Friday, August 1, 2014 - link
"Does this even affect real world performance for a normal person (as in not for work related activities)? I've never been able to detect much difference in a gaming scenario."If you're using integrated graphics? Absolutely yes for games. Otherwise probably not.
MrSpadge - Friday, August 1, 2014 - link
That's the wrong question to ask in the context of Haswell-E and X99. Anyone buying these is by definition not a normal person. Ask again once DDR4 hits meinstream platforms. For iGPU it will provide obvious benefits, whereas for regular operation you'll probably see a modest performance increase, equivalent to 1 or 2 CPU frequency bumps (it's been like that for DDR2 and 3 as well).Scabies - Thursday, July 31, 2014 - link
Using Anandtech's "Performance Index" (speed divided by CAS latency), this rates fairly low. Will all DDR4 start out this way or are we looking at a budget part here?Mr Perfect - Thursday, July 31, 2014 - link
It will probably all start out that way. DRR3 had high latencies when it first came out, DDR2 did to, and DDR1.... Actually, I don't remember how DDR1 compared to SD RAM anymore, but it certainly got better as it matured.B3an - Thursday, July 31, 2014 - link
It's stupid how a new DDR standard is always slower than the previous one when it starts out.DanNeely - Friday, August 1, 2014 - link
That should be inevitable. The ram manufacturers have years of experience tweaking parts for the old standard; but have only just figured out how to make the new one in mass quantities.CaedenV - Thursday, July 31, 2014 - link
It will absolutely get better with age. Think of this more like DDR3 800; it was a thing for a little while at launch, but was quickly bumped up to 1333 with lower latency, and then 1600+ as things went forward. We will have the better part of a year to get things a right before mainstream platforms start using this, so I would not worry about it.MrSpadge - Friday, August 1, 2014 - link
As others have said, it should quickly get better. But there's more: DDR4 is supposed to be a bit faster at similar specs due to improved efficiency. Let's see how this works out!Furthermore: this is for businesses who wouldn't run current Intels at anything higher than DDR3-1600, because they're not specified for anything faster. and if they can keep up DDR4-2133 15-15-15 for 16 GB modules it's actually not that bad.
Impulses - Thursday, July 31, 2014 - link
Pricing's the most important part, the cost will be absorbed by system integrators regardless but whether there's price parity or an increase will determine how quickly enthusiasts adopt a platform that uses DDR4.I imagine everyone will soon slap spreaders on modules regardless of whether it's needed or not, tho if you're making bland green DIMMs you might as well.
Assimilator87 - Friday, August 1, 2014 - link
I do love Geil's Black Dragon series for memory without heatspreaders.Mr Perfect - Thursday, July 31, 2014 - link
They don't really need the heatsinks though, do they? I was under the impression they're usually there for the aesthetics.
CaedenV - Thursday, July 31, 2014 - link
performance memory at 1600+ could probably use at least a basic spreader, especially when you don't know what kind of environment it is going to be installed in. I am a firm believer in AC in by house, but some crazy people somehow enjoy 80*f temps, and computer parts may not deal so well for them. So better safe than sorry I suppose.Still, with modern processing, lower voltage, and a cycle time, I can't imagine these getting all that warm.
Peeping Tom - Thursday, July 31, 2014 - link
Not unless they're overclocked beyond spec IMOsemyon95 - Thursday, July 31, 2014 - link
These are honestly just garbage sticks. I see no reason anyone would buy DDR4 sticks that:1. have only 8 GB for a DDR4 set
2. provide only 2133 MHz bandwidth for DDR4
3. add to this worst you can get CL15 timings
4. also these are not even LPDDR4/notebook oriented modules so you can't excuse the specs
What we have here is a product that is targeted at desktops, but every DDR3 2133 set outperforms it. The only advantage of this set -- low voltage -- doesn't matter for a desktop PC.
Kjella - Friday, August 1, 2014 - link
Maybe because hammering those DDR3 2133 set sticks into a platform that uses DDR4 won't work? Considering a Haswell-E here but might skimp on the RAM to buy a decent set later.Peeping Tom - Thursday, July 31, 2014 - link
Please don't drag out the latency improvements until we're on the cusp of DDR5/Nextgen.DanNeely - Friday, August 1, 2014 - link
If it was possible to make DRAM itself dramatically faster, we'd never have gone down the whole DDR1/2/3/4 cycle with each generation roughly corresponding to doubling the databus speed and doubling the on-dimm parallelism to feed it. PC-133, DDR1-266, DDR2-533, DDR3-1066, and DDR4-2133 all have the dram running at the same speed; and just use faster/more parallel controllers to read/write to more locations within the ram chips at once.This is also why we should expect to see DDR4 speeds ramp up fairly quickly; as the engineers get more practice in building faster controllers; the actual dram has plenty of headroom before it catches up with the speeds at which it's been made to operate in DDR3.
abrowne1993 - Thursday, July 31, 2014 - link
With equal clockspeeds and lower timings, wouldn't the DDR3 example above outperform these DDR4 modules? What's the point in upgrading for a desktop user?drothgery - Thursday, July 31, 2014 - link
DDR4 won't be available for a standard desktop platform for a while. Haswell-E is a server/workstation part.bmayer0122 - Thursday, July 31, 2014 - link
As there is a drop in voltage (1.65 to 1.2v) I would expect the power usage to go down on a per unit of memory basis. Is there any word on that?Thank you
willis936 - Thursday, July 31, 2014 - link
A rule of thumb in processor power consumption is that the max load power is proportional to the clock frequency times the square of the voltage. I promise you that these sticks use much less power than their ddr3 equivalent. That being said they're pretty low performance and in a HEDT environment the tiny amount of power that ddr3 isn't an issue. It matters more in server blades with hundreds of sticks or in laptops were every mW counts.celestialgrave - Friday, August 1, 2014 - link
Is it just me or does there seem to be a lot of wasted PCB space? Would it really cost that much more for them to just make the PCB shorter? Would be nice for those big tower heatsinksDanNeely - Friday, August 1, 2014 - link
I think the vertical heights for the PCB are set in the DDR4 spec itself; even if they don't need the space internally for routing connections in the PCB the height may have been selected for mechanical reasons. If they kept the length the same and but made them shorter they'd become more prone to bending or breaking.Notmyusualid - Friday, August 1, 2014 - link
CL15....Ugly.
StrangerGuy - Sunday, August 3, 2014 - link
The reason why DDR4 still hasn't went mainstream by now is because demand for NAND and LPDDRx is going up while demand for PC DRAM is slowing down.BTW I predict DDR4 will be the very last consumer DRAM to come in DIMMs.