Wednesday, March 17th 2021

DDR5-6400 RAM Benchmarked on Intel Alder Lake Platform, Shows Major Improvement Over DDR4

As the industry is preparing for a shift to the new DDR standard, companies are trying to adopt the new technology and many companies are manufacturing the latest DDR5 memory modules. One of them is Shenzhen Longsys Electronics Co. Ltd, a Chinese manufacturer of memory chips, which has today demonstrated the power of DDR5 technology. Starting with this year, client platforms are expected to make a transition to the new standard, with the data center/server platform following. Using Intel's yet unreleased Alder Lake-S client platform, Longsys has been able to test its DDR5 DIMMs running at an amazing 6400 MHz speed and the company got some very interesting results.

Longsys has demoed a DDR5 module with 32 GB capacity, CAS Latency (CL) of 40 CL, operating voltage of 1.1 V, and memory modules clocked at 6400 MHz. With this being an impressive memory module, this is not the peak of DDR5. According to JEDEC specification, DDR5 will come with up to 8400 MHz speeds and capacities that are up to 128 GB per DIMM. Longsys has run some benchmarks, using an 8-core Alder Lake CPU, in AIDA64 and Ludashi. The company then proceeded to compare these results with DDR4-3200 MHz CL22 memory, which Longsys also manufactures. And the results? In AIDA64 tests, the new DDR5 module is faster anywhere from 12-36%, with the only regression seen in latency, where DDR5 is doubling it. In synthetic Ludashi Master Lu benchmark, the new DDR5 was spotted running 112% faster. Of course, these benchmarks, which you can check out here, are provided by the manufacturer, so you must take them with a grain of salt.
Source: Longsys
Add your own comment

49 Comments on DDR5-6400 RAM Benchmarked on Intel Alder Lake Platform, Shows Major Improvement Over DDR4

#1
lynx29
is that latency going to be ok for gaming though??? i feel like this is great for synthetics but not great for gaming
Posted on Reply
#2
robb
"proceeded to compare these results with DDR4-3200 MHz CL22"

what kind of garbage DDR4 comes with such a crappy CL for that speed? CL16 is pretty much the norm for that speed on good memory kits.
Posted on Reply
#3
Crackong
If we count the latency with MHz and CL only
DDR4 2400 CL15 will be 1/2400*15 = 0.00625
This DDR5 6400 CL40 will be 1/6400*40 = 0.00625

So it is very roughly DDR4 2400 JEDEC performance
Posted on Reply
#4
londiste
CrackongIf we count the latency with MHz and CL only
DDR4 2400 CL15 will be 1/2400*15 = 0.00625
This DDR5 6400 CL40 will be 1/6400*40 = 0.00625

So it is very roughly DDR4 2400 JEDEC performance
Also DDR4 3200 CL20 is the same, this would be DDR4 3200 JEDEC spec.
Posted on Reply
#5
ViperXTR
isn't latency continuously increasing ever since for each generation and we dont have issues with them?
Posted on Reply
#6
Midland Dog
utter trash, i get high 50 gb/s and 36ns of latency on some trash tier 3600 cl17 bdie bin huhuhuhuhuhu
Posted on Reply
#7
bxcounter
Midland Dogutter trash, i get high 50 gb/s and 36ns of latency on some trash tier 3600 cl17 bdie bin huhuhuhuhuhu
Latency is trash , sure, but that 35 gb/s bandwith is single channel. I'm positive that in few years time DDR5 will break 100+ gb/s in dual channel mode.
Posted on Reply
#8
watzupken
I think best to wait and see if we will see any benefits over time. The JEDEC standards are usually very low, and we should see better timings from overclocked versions.
Posted on Reply
#9
Et3rNa1
Is there a mistake in the article?
Elsewhere the same news is reported with 4800Mhz


4800, with those latency and bandwidth is very different than 6400. ;)
Posted on Reply
#10
ppn
is it Quad channel, 2 per controller.
Posted on Reply
#11
Athlonite
Et3rNa1Is there a mistake in the article?
Elsewhere the same news is reported with 4800Mhz


4800, with those latency and bandwidth is very different than 6400. ;)
That also looks like it's running in single channel mode :kookoo: or am I missing something shouldn't the second stick be in Controller 1 Channel 0 I think someone at Intel needs a :slap:
Posted on Reply
#12
ssdpro
I must have reached knowledge obsolescence. I have no idea what I am looking at. For what it is, it looks like junk. I don't know how anyone can tell what channel mode this is in. If single, why? What would be the intent? To show you can run in single channel mode with this new DDR5 at crippled latency and 10 year old speeds? I have been sitting on an old skylake 6700k and DDR4 for about 5 years now waiting for that next upgrade and it outperforms whatever is shown here by a mile.
Posted on Reply
#13
londiste
ppnis it Quad channel, 2 per controller.
DDR5 has 2 channels per DIMM.

DDR4 has 72-bit channel: 64-bit data + 8-bit ECC.
DDR5 has two 40-bit channels: 32-bit data + 8-bit ECC each.
Posted on Reply
#15
metalfiber
"Major improvement"....Am i missing something? :confused:

Posted on Reply
#16
Toothless
Tech, Games, and TPU!
Everyone being all like "hurdedur I'm faster" while benching dual and quad channels against a single stick.
Posted on Reply
#17
lexluthermiester
lynx29is that latency going to be ok for gaming though???
It has the potential for being a problem for gaming.
robb"proceeded to compare these results with DDR4-3200 MHz CL22"

what kind of garbage DDR4 comes with such a crappy CL for that speed? CL16 is pretty much the norm for that speed on good memory kits.
Agreed, those are crap specs. At CL22 DDR4 should be running at 4000mhz nominal.
Posted on Reply
#18
Tomorrow
CrackongIf we count the latency with MHz and CL only
DDR4 2400 CL15 will be 1/2400*15 = 0.00625
This DDR5 6400 CL40 will be 1/6400*40 = 0.00625

So it is very roughly DDR4 2400 JEDEC performance
Unless DDR5 does some magic behind the scenes i dont see it being any faster than DDR4. Atleast in terms of latency:

4800 Mhz @ CL40 is 16,66ns

3200 Mhz CL16 DDR4 is 10,00ns.
3800 CL14 is 7,36ns.
4800 CL18 is 7,91ns.

So intial DDR5 latency will be double that of DDR4. Bit less when compared to standard 3200 CL16 kits.
I also calculated numbers based on speeds we should be getting in the coming years. Assuming the latency does not rise even more:

6400 Mhz CL40 is 12,50ns.
8400 Mhz CL40 is 9,52ns.

So only when DDR5 crosses 8000 Mhz at CL40 does it achieve latency parity with 3200 CL16 but it will still lose to faster DDR4 kits.
So here's hoping that they are no just hoping to brute force performance with raw bandwidth and there are some other things DDR5 does to achieve better latency than a simple Mhz/CL calculation would suggest.
ViperXTRisn't latency continuously increasing ever since for each generation and we dont have issues with them?
That is bevcause thus far the speed increase has been able to offset latency increase. DDR5 will need 8000Mhz to be able to do that. And those speeds are still years away.
bxcounterLatency is trash , sure, but that 35 gb/s bandwith is single channel. I'm positive that in few years time DDR5 will break 100+ gb/s in dual channel mode.
This was dual channel. As inheritly it would be impossible to run DDR5 on single channel even with just one stick seeing as all DDR5 will have dual channel from a single stick.
ppnis it Quad channel, 2 per controller.
That's not how memory channels work. You do not get Quad channel simply by using more sticks. Its still dual channel as determined by the motherboard and CPU.
AthloniteThat also looks like it's running in single channel mode :kookoo: or am I missing something shouldn't the second stick be in Controller 1 Channel 0 I think someone at Intel needs a :slap:
DDR5 is unable to run in single channel (unless there is a BIOS option for force disable half the stick) as a single stick will appear to the system as dual channel with all the performance benefits.

But yeah those numbers are atrocious. Dual channel ~30GB/s with 112ns latency. I already get nearly 60GB/s at half that latency on DDR4. As anyone who has experienced DDR generation change - skip the 1st gen. It's not worth it. Only 4800Mhz. DDR4 can already achieve this (with much lower latency). Double or even triple the DDR4 latency and miniscule voltage savings (1.1v vs 1.35v etc).

DDR5 needs to achieve 8000Mhz (at CL40 or lower) to match DDR4-3200 CL16 latency. Preferrably we need something like DDR5-8000 at ~CL30 to establish a clear lead. And you can bet those kits are years away. Not before 2023 or even 2024. Not to mention the early adopter price. I feel like Alder Lake will be great for portable devices (in terms of battery life) with it's big.LITTLE design and lower voltage DDR5 but will offer no benefits on desktop.

Even if they manage to double the bandwidth shown in these pathetic results by Alder Lake's launch it will still match, not exceed DDR4. With triple the system latency to boot. But im sure they will advertise "great savings" in terms of power. A few watts at best considering how little power RAM uses. I feel like DDR5 is aimed at servers first are foremost (low voltage, big capacity etc) and regular desktop users and enthusiasts get shafted once again in the first years.
Posted on Reply
#19
boulard83
My old X99 quad channel is twice as fast....
Posted on Reply
#20
Punkenjoy
TomorrowUnless DDR5 does some magic behind the scenes i dont see it being any faster than DDR4. Atleast in terms of latency:

4800 Mhz @ CL40 is 16,66ns

3200 Mhz CL16 DDR4 is 10,00ns.
3800 CL14 is 7,36ns.
4800 CL18 is 7,91ns.

So intial DDR5 latency will be double that of DDR4. Bit less when compared to standard 3200 CL16 kits.
I also calculated numbers based on speeds we should be getting in the coming years. Assuming the latency does not rise even more:

6400 Mhz CL40 is 12,50ns.
8400 Mhz CL40 is 9,52ns.

So only when DDR5 crosses 8000 Mhz at CL40 does it achieve latency parity with 3200 CL16 but it will still lose to faster DDR4 kits.
So here's hoping that they are no just hoping to brute force performance with raw bandwidth and there are some other things DDR5 does to achieve better latency than a simple Mhz/CL calculation would suggest.


That is bevcause thus far the speed increase has been able to offset latency increase. DDR5 will need 8000Mhz to be able to do that. And those speeds are still years away.

This was dual channel. As inheritly it would be impossible to run DDR5 on single channel even with just one stick seeing as all DDR5 will have dual channel from a single stick.

That's not how memory channels work. You do not get Quad channel simply by using more sticks. Its still dual channel as determined by the motherboard and CPU.

DDR5 is unable to run in single channel (unless there is a BIOS option for force disable half the stick) as a single stick will appear to the system as dual channel with all the performance benefits.

But yeah those numbers are atrocious. Dual channel ~30GB/s with 112ns latency. I already get nearly 60GB/s at half that latency on DDR4. As anyone who has experienced DDR generation change - skip the 1st gen. It's not worth it. Only 4800Mhz. DDR4 can already achieve this (with much lower latency). Double or even triple the DDR4 latency and miniscule voltage savings (1.1v vs 1.35v etc).

DDR5 needs to achieve 8000Mhz (at CL40 or lower) to match DDR4-3200 CL16 latency. Preferrably we need something like DDR5-8000 at ~CL30 to establish a clear lead. And you can bet those kits are years away. Not before 2023 or even 2024. Not to mention the early adopter price. I feel like Alder Lake will be great for portable devices (in terms of battery life) with it's big.LITTLE design and lower voltage DDR5 but will offer no benefits on desktop.

Even if they manage to double the bandwidth shown in these pathetic results by Alder Lake's launch it will still match, not exceed DDR4. With triple the system latency to boot. But im sure they will advertise "great savings" in terms of power. A few watts at best considering how little power RAM uses. I feel like DDR5 is aimed at servers first are foremost (low voltage, big capacity etc) and regular desktop users and enthusiasts get shafted once again in the first years.
each DDR5 DIMM have it's own 2x32 memory channel instead of 1x64 like all previous version. So if you run it with what we call currently dual channel, you will have 4x32bit memory channel instead of 2x64 memory channel. This was made to reduce the channel number per core. When dual channel was implemented it was 2 memory channel per "core" as cpu were single core at the time. Now, it's 1 memory channel per 5 on Intel right now and 1 per 8 core on AMD best desktop CPU.
This will cut these number in half. The memory will burst a 64 bit line and will use the increased memory bandwidth to do it.

This shouldn't help a lot in synthetic benchmark, but it will greatly help for heavily threaded load.

Also, there are many other feature that will increase overall performance but still might not appear in synthetic benchmark. By example, Same bank refresh instead of whole dimm refresh will increase memory availability since only the chip being refresh will be unavailable.
Posted on Reply
#21
Chrispy_
CrackongIf we count the latency with MHz and CL only
DDR4 2400 CL15 will be 1/2400*15 = 0.00625
This DDR5 6400 CL40 will be 1/6400*40 = 0.00625

So it is very roughly DDR4 2400 JEDEC performance
Latency should always be compared in nanoseconds, yes.

Still, 6400 CL40 is the same as 3200 CL20 and nobody buys CL20 DDR4-3200 any more. It's CL16 for the cheapo-tier mass-market/OEM stuff and CL15/CL14 for the premium bins, meaning that this is a big step backwards over even the cheapest RAM on the market at the moment.

Still, it's only an early demo. Final retail DDR5 should be at least as good as the mainstream stuff when it arrives - in terms of absolute latency, with the added bonus of far more bandwidth (hello, IGP performance!)
Posted on Reply
#22
stimpy88
They found some DDR4-3200 MHz CL22??? You can bet your arse that they totally mangled the rest of the DDR4 memory sub-timings to be as high in latency as possible.

This is pure marketing BS, and will end up being the same as every other launch of a memory standard - The new stuff will be marginally worse than the best DDR4, but bandwidth will be better for DDR5. I hope Intel and AMD add some cache to the DDR5 memory controllers, to hide some of that ridiculous latency!
Posted on Reply
#23
lexluthermiester
Chrispy_Still, it's only an early demo. Final retail DDR5 should be at least as good as the mainstream stuff when it arrives - in terms of absolute latency, with the added bonus of far more bandwidth (hello, IGP performance!)
While that might be a bit optimistic, it would be very nice.
Posted on Reply
#24
Tomorrow
Punkenjoyeach DDR5 DIMM have it's own 2x32 memory channel instead of 1x64 like all previous version. So if you run it with what we call currently dual channel, you will have 4x32bit memory channel instead of 2x64 memory channel. This was made to reduce the channel number per core. When dual channel was implemented it was 2 memory channel per "core" as cpu were single core at the time. Now, it's 1 memory channel per 5 on Intel right now and 1 per 8 core on AMD best desktop CPU.
This will cut these number in half. The memory will burst a 64 bit line and will use the increased memory bandwidth to do it.

This shouldn't help a lot in synthetic benchmark, but it will greatly help for heavily threaded load.

Also, there are many other feature that will increase overall performance but still might not appear in synthetic benchmark. By example, Same bank refresh instead of whole dimm refresh will increase memory availability since only the chip being refresh will be unavailable.
Even when using two DDR5 sticks the system will still be dual channel (dual rank, 2 dimm's per channel) because that is determined by the motherboard. The fact that DDR5 enables dual channel from a single stick does not make the system magickly quad channel when using two or even four sticks. Quad channel requires motherboard and CPU support. Alder Lake-S is dual channel. That more than anything confirms that it will be dual channel.
Posted on Reply
#25
londiste
stimpy88They found some DDR4-3200 MHz CL22??? You can bet your arse that they totally mangled the rest of the DDR4 memory sub-timings to be as high in latency as possible.
JEDEC spec is CL20 to CL22.
DDR5 like other new generation memory standards are likely to get to servers first and there memory tends to be run at standard speeds.
TomorrowEven when using two DDR5 sticks the system will still be dual channel (dual rank, 2 dimm's per channel) because that is determined by the motherboard. The fact that DDR5 enables dual channel from a single stick does not make the system magickly quad channel when using two or even four sticks. Quad channel requires motherboard and CPU support. Alder Lake-S is dual channel. That more than anything confirms that it will be dual channel.
It does make it dual channel but if each channel is now 32-bit instead of 64-bit as with DDR4 the extra channel benefit is largely negated.
Posted on Reply
Add your own comment