Benchmarks showcase impressive DDR5-4800 RAM performance

nanoguy

Posts: 1,355   +27
Staff member
Something to look forward to: We won't have to wait much longer for DDR5 to make its way into consumer products, and it looks like the benefits over DDR4 will be worth it. The new memory modules will feature higher capacity, higher bandwidth, and lower power consumption, not to mention two data channels per DIMM and integrated error correction.

Earlier today, Intel announced its 11th-gen Rocket Lake CPUs, solid performers that will come at more affordable price points than AMD's Ryzen 5000 series CPUs. However, they are a dead end for LGA 1200 motherboards, which also means they don't support DDR5 memory, and are limited to the now mature DDR4 technology.

Last year, JEDEC released the DDR5 specification, generating a lot of excitement among gamers and enthusiasts. At least on paper, the new memory tech will offer four times the capacity per die and double the effective bandwidth, while also leveraging a more power-efficient architecture when compared to DDR4 SDRAM.

Must read: Anatomy of RAM

Industry leaders like Micron and SK Hynix started sampling DDR5 to partners in the enterprise space at the beginning of last year, and others like TeamGroup are scrambling to validate their modules with motherboard manufacturers. That said, we didn't see any of these companies showcasing what performance would look like in practice.

Today, Chinese manufacturer Shenzhen Longsys Electronics revealed the first benchmarks of DDR5-4800 memory using one of Intel's Alder Lake-S engineering samples. The company is currently working on two DDR5 modules, a 16GB single-rank design and a 32GB dual-rank design, both operating at 1.1V with a CAS Latency of 40.

Longsys compared the DDR5-4800 memory with what appears to be DDR4-3200, using a combination of AIDA64 and Ludashi, the latter being a popular Chinese benchmarking software. In AIDA64's read, write, and copy tests, the DDR5 memory module scored 39 percent, 36 percent, and 12 percent higher, respectively. However, it also showed almost double the latency versus the DDR4 modules.

In the Ludashi Master Lu benchmark, the DDR5 module obtained a score of just over 190,000 points, while the DDR4 module scored less than 92,000 points. This is an impressive result, and eventually DDR5-6400 memory will come to surpass it with even better numbers.

According to roadmaps, the earliest we could taste that kind of DDR5 performance in a consumer product will be when Intel releases Alder Lake-S processors. The company says that could be as soon as the second half of this year, but that timeline could very well slip into 2022 as the company is gearing to retake its position of "unquestioned leader in process technology."

Permalink to story.

 
I won’t be building a new PC till I can have DDR5.

I wonder what generation Intel Core will be available then???

I wonder how my 3090 FTW3 will perform on that new motherboard with that new memory?

It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.
 
It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.


DDR4 was released in 2014.

I was able to buy 32GB for just $300 during the height of the crypo ponzi scheme madness in 2017.

My only concern is whether or not the speed will justify the upgrade prices, but my only point is, I am not gonna build a new PC till DDR5 is available so I can have a new motherboard that handles it and newer CPU along with newer tech SSD.

I only want solid state with liquid cooling. No moving parts allowed.
 
It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.
Meh... I adopted DDR4 on 5820k and it is still a strong platform. Was worth it for the longevity. The bandwidth still stacks up.
 
DDR4 was released in 2014.

I was able to buy 32GB for just $300 during the height of the crypo ponzi scheme madness in 2017.

My only concern is whether or not the speed will justify the upgrade prices, but my only point is, I am not gonna build a new PC till DDR5 is available so I can have a new motherboard that handles it and newer CPU along with newer tech SSD.

I only want solid state with liquid cooling. No moving parts allowed.
Yep I bought 32GB DDR4 quad channel kit for $510 AUD in Nov 2014. About $440 USD at the time. ROI I think it paid itself off.

Liquid cooling has a pump right? That's a moving part isn't it?
 
It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.

Not necessarily true at all. If DDR5 hits the market before widespread acceptance and adoption it could be very reasonably priced. This may be a case of early adopters getting the very best prices in fact. If history repeats itself. Of course, in this crazy world anything is possible. Remember, the laws of supply and demand....
 
This is one of the reasons why I will simply wait for AMD new cpu and socket.

Also, maybe rdna3 will be out or perhaps, rx6900 would be available at a more reasonable price.

Fun times ahead!
 
Never the early adopter, never preorder, never the paying beta tester. I have something called patience.

I like to see the tech mature a little, get over bugs and growing pains and then I'll see DDR5 in 2023 (or even later) in my PC.

Not to mention that those that think DDR5 will have great price/performance at launch are incredibly naive, to say it mildly.
 
40 CAS at 4800MHz seems very high. That's worse than highest latency for official JEDEC DDR4(16ns vs 15ns - and with just XMP, you can get below 10ns). If that's indicative of what the performance will be like at release, I would expect to see little to no benefit over DDR4(and possibly even performance regression, everything else being equal) except for iGPUs and some niche workloads.
 
40 CAS at 4800MHz seems very high. That's worse than highest latency for official JEDEC DDR4(16ns vs 15ns - and with just XMP, you can get below 10ns). If that's indicative of what the performance will be like at release, I would expect to see little to no benefit over DDR4(and possibly even performance regression, everything else being equal) except for iGPUs and some niche workloads.
It won't be the case of regression and not even equal performance, the leaks so far showed that despite the high latency DDR5 is still faster than the best DDR4.

That being said, the DDR5 that will come after 2-3 years, that one would be much much better and will also have the best price/performance ratio vs the best DDR4. So yeah, the first iterations of DDR5 will have drawbacks, but not as much as a complete move backwards. No one would buy it if that would be the case...
 
It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.

Not necessarily. I got the first platform with DDR4 and it wasn't really that much more than DDR3 had been at the time.
 
Memory is probably the most boring part I buy, and I've never had to rush to get it. When 6400MHz is priced right I'll buy it when the time comes.
 
But how much are modern applications actually handicapped by RAM speed? Zen 3 CPU's already have huge 32MB caches and I'd expect caches to get larger over time so you wonder how much the RAM is even accessed. If the issue is loading graphics resources then the bottleneck is more likely to be in getting the data from SSD. If the RAM isn't the bottleneck within a modern system then speeding it up will obviously give no real improvement. It would be interesting to see where the bottleneck is in a modern PC running the latest applications (games or work) - is it CPU, GPU, SSD, RAM, PCIe etc? Does it differ between AMD and Intel? What about between Windows and Linux?
 
A boon for integrated graphics typically hamstrung performance wise by memory bandwidth. Particularly on notebook platforms.
 
Memory is probably the most boring part I buy, and I've never had to rush to get it. When 6400MHz is priced right I'll buy it when the time comes.

Yep, I time my mobo and cpu upgrades around memory releases but one generation behind. When DDR4 came out I upgraded to DDR3. Did the same when DDR5 was announced and just upgraded to DDR4. Not missing out on anything since CPU and memory have such negligible effect on gaming performance.
 
But how much are modern applications actually handicapped by RAM speed? Zen 3 CPU's already have huge 32MB caches and I'd expect caches to get larger over time so you wonder how much the RAM is even accessed. If the issue is loading graphics resources then the bottleneck is more likely to be in getting the data from SSD. If the RAM isn't the bottleneck within a modern system then speeding it up will obviously give no real improvement. It would be interesting to see where the bottleneck is in a modern PC running the latest applications (games or work) - is it CPU, GPU, SSD, RAM, PCIe etc? Does it differ between AMD and Intel? What about between Windows and Linux?
Some strategy games like sins of a solar empire and supreme commander show scaling with memory speeds, and even the highest DDR4 OCs still scale. DDR5 will be a boon for such games. And sure games dont need it *now* but as always it isnt a need, it'll just show upas newer platforms pulling ahead, like core 2 duos with DDR3 instead of DDR2 memory, or skylake with DDR4 VS DDR3
It will be VERY expensive when it does come out, so it may be several years after that before it makes any economic sense to build a PC with it.
It also makes no sense to jump when it first comes out. For the last 20 years any memory type released follows the same pattern. It comes out, its expensive, and 2 years later you can get significantly faster AND lower latency memory for less money.
 
DDR5 will bring ECC to the masses. Part of the DDR5 standard.

I honestly don't expect Alder Lake till early 2022, Same goes for Zen 4. So AMD and Intel should both be getting DDR5 around the same time. Demand and prices are going to be high.

My Plan is to wait for these DDR5 platforms. I have little faith in Alder Lake, but I hope Intel can use some of their old magic to get their core arch back on some type of track. As Rocket Lake is a full derailment.
 
40 CAS at 4800MHz seems very high. That's worse than highest latency for official JEDEC DDR4(16ns vs 15ns - and with just XMP, you can get below 10ns). If that's indicative of what the performance will be like at release, I would expect to see little to no benefit over DDR4(and possibly even performance regression, everything else being equal) except for iGPUs and some niche workloads.
JEDEC have specified 3 different variants (A, B, and C) for each DDR5 speed - A being the highest rated, and C the lowest.

DDR5-4800A is 34-34-34, B is 40-40-40, and C is 42-42-42; compared to something like this DDR4-4800, which is 46-24-24, the A-spec stuff is pretty good.

The advantage DDR5 will have over 4 is the far higher achievable transfer rates (currently max is DDR5-6400).
 
Are any of those numbers suppose to impress me?

My old AF X99 platform blows it out of the water...
Aida 64 Bench.png
 
DDR5 will bring ECC to the masses
Poor Intel, if it wasnt because they were murdered by AMD, they would have found a way to take this away from the consumers line and force you to a Xeon CPU just because they are that kind of d!cks.

Thats one of the reasons why I simply refuse to give them any money.
 
Are any of those numbers suppose to impress me?
Note the reported clock speed of the tested CPU: 805 MHz. It could be a reporting error, or it really could have been running at that rate in the test. If so, then yes - those figures are very impressive for that speed. Even if it is a reporting error, the processor is unlikely to be running at anywhere near full release speeds and so the memory controllers will be running quite slow too.
 
Note the reported clock speed of the tested CPU: 805 MHz. It could be a reporting error, or it really could have been running at that rate in the test. If so, then yes - those figures are very impressive for that speed. Even if it is a reporting error, the processor is unlikely to be running at anywhere near full release speeds and so the memory controllers will be running quite slow too.
800 Mhz was the base clock of the CPU as stated on a post I also read on Tom's, Aida doesn't always detect the maximum clock of the CPU, not to mention it is an unreleased CPU that Aida probably doesn't fully support yet, even in my capture it shows 3 GHz however my CPU is running at 4.5 GHz for example.
 
800 Mhz was the base clock of the CPU as stated on a post I also read on Tom's, Aida doesn't always detect the maximum clock of the CPU, not to mention it is an unreleased CPU that Aida probably doesn't fully support yet, even in my capture it shows 3 GHz however my CPU is running at 4.5 GHz for example.
If that's definitely the case, then it's still likely to be clocked very conservatively. Entries for other Alder Lake-S chips in the SiSoft Sandra database have higher base clocks, but none of them are especially high. That said, AL-S could just as well be an absolute dud :)
 
Back