First Tentative Alder Lake DDR5 Performance Figures Leak

[Sybaris_Caesar]

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?

I have at least a decade worth of salt as tech reader I'll have you know.

 

[Zubasa]

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.

Technically that DDR5-6400 still counts as an OC, because the JEDEC spec for Alder Lake is only DDR5-4800.

 

[GorbazTheDragon]

This seems roughly the DDR5 equivalent to a 3200c16 DDR4 bin when comparing to the JEDEC DDR5 specs... I think anyone who was expecting DDR5 to bring performance uplifts on the CPU side is foolish, it's pretty obvious from the timings that it will not improve latency.

It will be pretty interesting to see how big the gap is comparing good dual rank B-die setups to these early DDR5 ICs.

 

[ThrashZone]

Hmm.. not at all what I was expecting.. thanks for sharing.

Hi,
Yeah but you'll need a twitter feed lol

Source should be on top not at the bottom of stories.

 

[dicktracy]

my 5800x only gets 670 in singlethread. beast. 3D stacking won't be enough against this

 

[Makaveli]

my 5800x only gets 670 in singlethread. beast. 3D stacking won't be enough against this

I score this in the Cpu z benchmark. And we will need to see a full review with more than 1 app. ADL will most certainly have a ST lead as it should. However its abit soon to be writing off the 3d stacking chips based on a single leak benchmark.

 

[ZoneDymo]

my 5800x only gets 670 in singlethread. beast. 3D stacking won't be enough against this

You dont have a 5800x, stop being silly and desperate.

 

[Metroid]

single 785? my ryzen 5900 does only 633, 633 + 63 + 63 + 24 = 783 = 5900 + 10% +10% + 4% = 24%, impressive.

 

[NSX05]

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.

Agreed. That is just encouraging false advertising.

 

[GorbazTheDragon]

Agreed. That is just encouraging false advertising.

I don't really have a problem, people are way too fixated on core counts with modern processors IMHO... There's not really a good way to say with a single number how many cores a hybrid design like this, and anyway the mobile guys set the precedent with this long before anyone realised x86 big little would be a thing...

 

[Fouquin]

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.

The round trip latency doesn't show you the individual operation latencies. We're looking at a really bog standard 12.5ns primary latency target here, and tRAS at 26.6ns is substantially faster than DDR4's debut at 32.8ns. Poor overall latency is almost assuredly down to IMC or board deficiencies and not the DIMMs themselves.

 

[Tardian]

I'm not holding my breath for DDR5 to be cheap and fast.

 

[GorbazTheDragon]

The round trip latency doesn't show you the individual operation latencies. We're looking at a really bog standard 12.5ns primary latency target here, and tRAS at 26.6ns is substantially faster than DDR4's debut at 32.8ns. Poor overall latency is almost assuredly down to IMC or board deficiencies and not the DIMMs themselves.

tRAS is not a relevant timing for random operation latency, it simply denotes when the RAS signal is dropped which is normally after the whole access (RCD+CAS) cycle completes anyway.

DDR5 has one key advantage over DDR4 which is the move to 4 independent channels. This enables more granularity of operations and should reduce latency under load as concurrently issued memory accesses are less likely to get in each other's way. It is not yet clear how much this will compensate for the worse timings.

 

[Fouquin]

tRAS is not a relevant timing for random operation latency

I don't recall saying it was.

 

[Ed_1]

According to Toms HW, this was tested in Gear 4 mode.

Alder Lake Tested With DDR5-6400 Memory, Has Unexpectedly High Latency

Gear 4 is good for extremely high memory speeds, but may be terrible for gaming.

www.tomshardware.com

 

[Makaveli]

Gear 4 was probably needed to get it to clock to DDR5-6400 speeds. I wonder what the same module would do clocked at 4800 with lower timings and in Gear 1 or 2 mode.

 

[MxPhenom 216]

Hi,
At 6400 mhz latency is going to be high c40 lol

This is why I always tend to skip the first 2-3 platforms that support a new DDR. Latency is usually high initially, than it drops down over a few years. DDR3 and 4 were the same way. DDR4 was exceptionally bad initially from latency standpoint.

 

[Crackong]

Since it is using DDR5
There is no point judging Alder Lake performance with these figures alone, it could be "boosted" by DDR5, or the opposite.


Still waiting for a Alder Lake "leak" using DDR4.

 

[Makaveli]

Since it is using DDR5
There is no point judging Alder Lake performance with these figures alone, it could be "boosted" by DDR5, or the opposite.


Still waiting for a Alder Lake "leak" using DDR4.

Comparing ADL on DDR4 vs DDR5 should be an interesting test.

 

[Crackong]

Comparing ADL on DDR4 vs DDR5 should be an interesting test.

Yes it will be a very useful topic since many budget oriented motherboard will have DDR4 support instead of 5

 

[Space Lynx]

I just hope we get some benchmarks with next gen gpu/cpu builds with ddr4 vs in ddr4 3600 cas 14 systems... and that is the only variable that changes. If we can get that, it would be great.

Hopefully the higher latency DDR5 ram doesn't hurt gaming smoothness. Probably won't hurt frame rates, but ms/smoothness... hopefully we get specific testing for that. I'm sure we will.

@TheLostSwede

 

[ZoneDymo]

I don't really have a problem, people are way too fixated on core counts with modern processors IMHO... There's not really a good way to say with a single number how many cores a hybrid design like this, and anyway the mobile guys set the precedent with this long before anyone realised x86 big little would be a thing...

they are, core count matters more for those who dont know and because of that calling 6 performance cores and 4 eco cores simple "10 cores" compared to a 5800X 8 performance cores, is basically a loophole to false marketing.

and you dont have to use a single number, just say something like " 6+4 " on the box and we are all good from that perspective, and CPU-Z and other programes should be aware and show that difference because if we just call it 10 cores, it could be 10 Pcores or 8 Pcores and 2 Ecores or 6 Pcores and 4 Ecores or 4 Pcores and 6 Ecores or 2 Pcores and 8 Ecores or hell, all 10 Ecores, and those would simply preform differently and would attract different consumers.

 

[Chrispy_]

That's actually an impressive 11% faster than rocket lake for single-threaded CPU-Z bench scores, Not that CPU-Z is a particularly good indicator.

 

[TheLostSwede]

they are, core count matters more for those who dont know and because of that calling 6 performance cores and 4 eco cores simple "10 cores" compared to a 5800X 8 performance cores, is basically a loophole to false marketing.

and you dont have to use a single number, just say something like " 6+4 " on the box and we are all good from that perspective, and CPU-Z and other programes should be aware and show that difference because if we just call it 10 cores, it could be 10 Pcores or 8 Pcores and 2 Ecores or 6 Pcores and 4 Ecores or 4 Pcores and 6 Ecores or 2 Pcores and 8 Ecores or hell, all 10 Ecores, and those would simply preform differently and would attract different consumers.

Well, there are some claimed early benchmarks, which suggests the small cores are actually doing a lot more than expected. I would take this with a healthy dose of natrium chloride.

Alleged Intel Alder Lake Core i9-12900KF, Core i7-12700KF, Core i5-12600KF CPU Benchmarks & Prices Leak Out

The first benchmarks of Intel Alder Lake Core i9-12900KF, Core i7-12700KF, & Core i5-12600KF CPUs along with retail prices have leaked out

wccftech.com

 

[HenrySomeone]

my 5800x only gets 670 in singlethread. beast. 3D stacking won't be enough against this

Yup, Ryzens won't stay competitive much longer, especially considering it looks like even Meteor Lake will be out before Zen4...yikes!