• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

First Tentative Alder Lake DDR5 Performance Figures Leak

TheLostSwede

News Editor
Joined
Nov 11, 2004
Messages
18,855 (2.50/day)
Location
Sweden
System Name Overlord Mk MLI
Processor AMD Ryzen 7 7800X3D
Motherboard Gigabyte X670E Aorus Master
Cooling Noctua NH-D15 SE with offsets
Memory 32GB Team T-Create Expert DDR5 6000 MHz @ CL30-34-34-68
Video Card(s) Gainward GeForce RTX 4080 Phantom GS
Storage 1TB Solidigm P44 Pro, 2 TB Corsair MP600 Pro, 2TB Kingston KC3000
Display(s) Acer XV272K LVbmiipruzx 4K@160Hz
Case Fractal Design Torrent Compact
Audio Device(s) Corsair Virtuoso SE
Power Supply be quiet! Pure Power 12 M 850 W
Mouse Logitech G502 Lightspeed
Keyboard Corsair K70 Max
Software Windows 10 Pro
Benchmark Scores https://valid.x86.fr/yfsd9w
The first indication of what to expect in terms of early DDR5 memory performance on Intel's Alder Lake platform has leaked, courtesy of a Chinese forum and Twitter. It's not the first time we've seen DDR5 performance figures, but this time the CPU is clocked at much higher speeds compared to the figures leaked by Longsys back in March.

Although the memory is still running at 6400 MHz, performance is up significantly with read speeds of over 90 GB/s. As we're most likely still looking at an ES sample and an early UEFI build, there should be room for improvement here. The leaked performance numbers appear to be from a Dell system of unknown model.




The not so great news is that the memory latency is still very high at 92.5 ns, although the 40-40-40-85 timings of the DDR5 modules used most likely has something to do with this. The CPU used for the test is said to be a Core i5 12600K and as an extra bonus, single core CPU-Z numbers of it were also posted, where it scores 785.6 points.

At this point in time, DDR5 still doesn't look very exciting, but as always when we transition to new memory technologies, there is usually an overlap where the fastest memory of the older generation always tends to outperform the first couple of JEDEC rated speeds of the new generation of memory. Once the memory module manufacturers get to play around a bit with DDR5, we expect to see much better performing modules in the market.

View at TechPowerUp Main Site
 
Hmm.. not at all what I was expecting.. thanks for sharing.
 
For an early result, this looks quite healthy, actually.
Breaking the 100GB/s barrier on dual channel machines is goin to be a breeze
 
For an early result, this looks quite healthy, actually.
Breaking the 100GB/s barrier on dual channel machines is goin to be a breeze
The unanswered question is how much doubling/tripling latency will affect things.
 
Still hoping CPU-Z will be updated to account separately for P and E cores.

Right now it acts like they are the same, adds them up and just calls it 10 cores, but that is not the whole story.
 
I would expect that to tighten up quite a bit as well
Yeah, so do I, but as with every recent generation of DRAM, we're not going to get close to the previous generation when it comes to latency.
 
Yeah, so do I, but as with every recent generation of DRAM, we're not going to get close to the previous generation when it comes to latency.
I mean we are. 40ns DDR5 6400 has roughly the same real world latency as CL20 DDR4 3200 or CL10 DDR3 1600. Yeah, that's nto great, but since this is first gen DDR5 I expect we'll see timings come down, wouldnt surprise me to see DDR5 8400 CL40 or 9600 CL40 in a years time.
 
Well, there's a lot more to this calculation. The increase in Cache capacity and complexity is a major contributor to a lot of the responsiveness effect of this ecosystem
 
Last edited:
It seems like everyone was waiting for the shortages to release new consoles, better gpu maker, new gpu's, new ram. Let's pretend we can all buy what we want at msrp
 
The unanswered question is how much doubling/tripling latency will affect things.
I just saw some random results from the "share your AIDA 64 cache and memory benchmark" thread, it seems the read, write and copy speed increase is already bigger than the latency toll. So not bad, although these are not standard speed and timings (4800 MHz with primary latency 40). And indeed the CPU result is impressive. Thanks for sharing.

 
The unanswered question is how much doubling/tripling latency will affect things.
Hi,
Doesn't seem like much of a hit
Just compare Intel and AMD now
Intel upper 30s
AMD upper 50s
 
Hi,
Doesn't seem like much of a hit
Just compare Intel and AMD now
Intel upper 30s
AMD upper 50s
Almost three times doesn't seem like much of a hit for you?
Can't really compare with AMD, as that's a potato to tomato comparison.
 
Almost three times doesn't seem like much of a hit for you?
Can't really compare with AMD, as that's a potato to tomato comparison.
Hi,
At 6400 mhz latency is going to be high c40 lol
 
Hi,
At 6400 mhz latency is going to be high c40 lol
Well, this is JEDEC spec, so we'll hopefully see something closer to 34 over time.
After all the JEDEC spec for 3200MHz DDR4 is CL 20, but you can get CL 14 modules.
I guess it all comes down to a combination of production node, engineering expertise and luck.

Not saying we're seeing anything unexpected here for that matter, it's just a bigger jump that what we're used to.
 
That's a nice ~40% single core performance uplift over my 9900K (even with a 500Mhz clock speed deficit over my 5Ghz clocks) which considering this is an Intel CPU, pretty huge.

For someone heavily into emulators, where single core IPC is king, its a very tempting proposition.
 
Last edited:
Well, this is JEDEC spec, so we'll hopefully see something closer to 34 over time.
After all the JEDEC spec for 3200MHz DDR4 is CL 20, but you can get CL 14 modules.
I guess it all comes down to a combination of production node, engineering expertise and luck.

Not saying we're seeing anything unexpected here for that matter, it's just a bigger jump that what we're used to.
Hi,
At 6400 even c32 would be mediocre at best.
 
Well, this is JEDEC spec, so we'll hopefully see something closer to 34 over time.
After all the JEDEC spec for 3200MHz DDR4 is CL 20, but you can get CL 14 modules.
I guess it all comes down to a combination of production node, engineering expertise and luck.

Not saying we're seeing anything unexpected here for that matter, it's just a bigger jump that what we're used to.
I'm going to bet that instead of lower latency, CL40 will be the standard but will be seen on higher speeds, EG 6400 CL40, 7200 CL40, 8400 CL40, ece to the point that CL40 is no longer an issue.

Imagine DDR5 12600 CL40.....
 
That's a nice ~40% single core performance uplift over my 9900K (even with a 500Mhz clock speed deficit over my 5Ghz clocks) which considering this is an Intel CPU, pretty huge.

For someone heavily into emulators, where single core IPC is king, its a very tempting proposition.

ok sure but is your 9900k not sufficient for that purpose?
 
Ehhh? Hey I wanna post news articles too damnit!

I got seven stars like @TheLostSwede as well. So why can't I post shit?
 
Last edited:
ok sure but is your 9900k not sufficient for that purpose?

Not for some emulators it isn't. (RPCS3 (PS3) Xenia (Xbox 360) Yuzu (Switch)

Clock speeds aren't going to increase by much any time soon so the only way to get more performance is for IPC increases.
 
Ehhh? Hey I wanna post news articles too damnit!

I got seven stars like @ThePumaSwede as well. So why can't I post shit?
Sorry, the what..? :(

Nothing to do with stars, just helping out a bit. Also, I guess you don't have a decade plus worth of work history as a tech journalist?
 
Not for some emulators it isn't. (RPCS3 (PS3) Xenia (Xbox 360) Yuzu (Switch)

Clock speeds aren't going to increase by much any time soon so the only way to get more performance is for IPC increases.
Or wait for the software to mature. All three of those emulators are in their early stages, and no matter how much power you have they wont perform smoothly. The switch wont need much CPu power to emulate eventually.

Incidentally all 3 of the platforms you listted are multicore designs, and single core performance is only going to help you so much there.
 
Back
Top