Friday, May 10th 2019

AMD Ryzen 3000 "Zen 2" a Memory OC Beast, DDR4-5000 Possible

AMD's 3rd generation Ryzen (3000-series) processors will overcome a vast number of memory limitations faced by older Ryzen chips. With Zen 2, the company decided to separate the memory controller from the CPU cores into a separate chip, called "IO die". Our resident Ryzen memory guru Yuri "1usmus" Bubliy, author of DRAM Calculator for Ryzen, found technical info that confirms just how much progress AMD has been making.

The third generation Ryzen processors will be able to match their Intel counterparts when it comes to memory overclocking. In the Zen 2 BIOS, the memory frequency options go all the way up to "DDR4-5000", which is a huge increase over the first Ryzens. The DRAM clock is still linked to the Infinity Fabric (IF) clock domain, which means at DDR4-5000, Infinity Fabric would tick at 5000 MHz DDR, too. Since that rate is out of reach for IF, AMD has decided to add a new 1/2 divider mode for their on-chip bus. When enabled, it will run Infinity Fabric at half the DRAM actual clock (eg: 1250 MHz for DDR4-5000).

This could turn into an additional selling point for AMD X570 chipset motherboards, as they'll have a memory frequency headroom advantage over boards based on older chipsets as their BIOS will include not just the increased memory clock limit, but also the divider mode. Of course this doesn't mean that you can just magically overclock any memory kit to these 5 GHz speeds - it's probable that only the best of the best modules will be able to get close to these speeds.

1usmus also discovered that the platform adds a SoC OC mode and VDDG voltage control. We've heard from several sources that AMD invested heavily in improving memory compatibility, especially in the wake of Samsung discontinuing its B-die DRAM chips.
Add your own comment

102 Comments on AMD Ryzen 3000 "Zen 2" a Memory OC Beast, DDR4-5000 Possible

#51
Xuper
for DDR4 4000 or above, you need a Mobo with beefy SoC VRM which almost all of those Tier 2/3 Mobo don't have it.
Posted on Reply
#52
bug
NdMk2o1o said:

Don't be ridiculous, you can make anything true on the Internet, I want to know where amd said this, oh you won't spoon feed me because its not true :rolleyes:
No, I won't waste my time digging up statements from two years ago because of your hostile attitude.
Maybe someone else will jump in and lend a hand, but I actually have no time to waste right now.
Posted on Reply
#54
NdMk2o1o
bug said:

No, I won't waste my time digging up statements from two years ago because of your hostile attitude.
Maybe someone else will jump in and lend a hand, but I actually have no time to waste right now.
That old chestnut ie: full of it :laugh:

Andrev said:

The specs sheet of Asrock B350m shows 3200+(OC) mhz speed. I'm curios where didn you saw that 4000mhz for that board.
In the bios and since multiple agesa updates since it was first released there have been more multipliers and speeds gradually added.
Posted on Reply
#55
TheLostSwede
<div class="youtube-embed" data-id="QrVZ1Yi6gq4:265"><img src="https://i.ytimg.com/vi/QrVZ1Yi6gq4:265/hqdefault.jpg" /><div class="youtube-play"></div><a href="https://www.youtube.com/watch?v=QrVZ1Yi6gq4:265" target="_blank" class="youtube-title"></a></div>
Posted on Reply
#56
londiste
Midland Dog said:
IMC is on the I/O die
I/O die is on the CPU package. The context of my comment is that IMC is not on the motherboard which would make it possible for old motherboards to support the divider with BIOS update.
Posted on Reply
#57
cucker tarlson
Tired of tech sites and channels doing that.
Can we accept Ryzen 3000 for what it once it launches or do we have to read clickbait "news" like this every day just to see for the 1000th time the real product never meets expecations if they're crazy high.
Posted on Reply
#58
TheLostSwede
cucker tarlson said:

Tired of tech sites and channels doing that.
Can we accept Ryzen 3000 for what it once it launches or do we have to read clickbait "news" like this every day just to see for the 1000th time the real product never meets expecations if they're crazy high.
Not really click bait, simply information collected from new UEFI files.
Posted on Reply
#60
R0H1T
TheLostSwede said:

<div class="youtube-embed" data-id="QrVZ1Yi6gq4:265"><img src="https://i.ytimg.com/vi/QrVZ1Yi6gq4:265/hqdefault.jpg" /><div class="youtube-play"></div><a href="https://www.youtube.com/watch?v=QrVZ1Yi6gq4:265" target="_blank" class="youtube-title"></a></div>
So zen3 has 4 way SMT & stacked HBM(?) for the new Xbox, might as well make a weekly AMD "all you can throw" rumors thread & I'm not being sarcastic :peace:
Posted on Reply
#61
InVasMani
I'm curious how BCLK frequency might play into the 1:2 divider. If you can offset it a bit via that and it's not tied to PCIE/SATA then I'm quite alright with that seems like a easy enough remedy potentially in such a case.
Posted on Reply
#62
Imsochobo
MikeMurphy said:

Maybe memory bandwidth is important because they supposedly need to feed 16 cores and 32 threads on a dual 64-bit memory channel architecture.



Integrated HBM, my dude.
Yes and no, a few times memory bandwidth is important.
But take a i9 14 core, remove two sticks and do dual channel, surprise!
Doesn't affect it that much at all!

Zens major limitation is latency more than bandwidth.
But surprise! Frequency decreases latency.

So yeah, to some extent bandwidth but I feel the latency is really what they are after.
Posted on Reply
#63
Melvis
Sounds great! but I would be happy if I could just get my 3600MHz Ripjaws to run at there rated speed instead of 3200MHz lol
Posted on Reply
#64
Chloe Price
And I'm running DDR4-2400 with my 2600 :D
Posted on Reply
#66
Jism
Even i can do 3600Mhz with a Hynix 2x 16GB kit. Have you guys configured the modules in the right slot(s) for example? This 470-F board only does these speeds when you hit slot 2 and 4 and not 1 and 3, for some weird reason.

Ontopic: I think with a half divider of IF and 2500Mhz memory speeds, you can still pack more bandwidth then ever and still get better scores. I dont think it's a useless feature but one beneficial.
Posted on Reply
#67
GoldenX
Jism said:

Even i can do 3600Mhz with a Hynix 2x 16GB kit. Have you guys configured the modules in the right slot(s) for example? This 470-F board only does these speeds when you hit slot 2 and 4 and not 1 and 3, for some weird reason.

Ontopic: I think with a half divider of IF and 2500Mhz memory speeds, you can still pack more bandwidth then ever and still get better scores. I dont think it's a useless feature but one beneficial.
Same here, slots 0 and 2 are useless for high frequencies (can't even go higher than 2933), 1 and 3 are the good ones.
Posted on Reply
#68
kapqa
I am also running Ryzen 2600 (with Adata 3000 but apparently it is only 2666 Mhz Memory) , and on Passmark the Memory Factor scores quite low (3 and half star out of 5).
Posted on Reply
#69
Camm
It should be noted it isn't exactly all roses on the other side of the fence. Since jumping over to a 9900K, my G.Skill DDR4 3866 4x16GB Dual Rank kit I can only hit 3200mhz with. On my AsRock X470 Taichi w/ 2700X, I could hit about 3333mhz.

I know I'm a unique usecase with full DIMM population with dual ranked modules, but I was a little surprised how weak Intel's memory controller is in this situation.
Posted on Reply
#70
Eselmio
I actually have my memory at 3400 CL 15-17-17-30 CR1 @ 1.35v it is a Quad Channel kit Trident Z RGB 4x8 GB 3200 Mhz CL16-18-18-38 1.35v , I've tested with memtest86+ for stability (chips are Hynix AFR)

My motherboard is an Extreme Zenith paired with a Threadripper 1950X overclock at 4.0 GHz with 1.37-1.39 using the compensating voltage mode.

Before I was using 8X4 GB of Corsair Dominator, you can not make them work at their advertised speed, work at the advertising speed of 3000 MHz CL15-17-17-35, could only make then work at 2800 Mhz

4 sticks were Hynix chips and the other Micron.

If this will become true, it will actually be very good.
Posted on Reply
#71
GoldenX
Camm said:

It should be noted it isn't exactly all roses on the other side of the fence. Since jumping over to a 9900K, my G.Skill DDR4 3866 4x16GB Dual Rank kit I can only hit 3200mhz with. On my AsRock X470 Taichi w/ 2700X, I could hit about 3333mhz.

I know I'm a unique usecase with full DIMM population with dual ranked modules, but I was a little surprised how weak Intel's memory controller is in this situation.
Yeah, there is no processor that likes all 4 dimms full. Weaklings.
Posted on Reply
#72
lsevald
How well does 4x8GB single ranked samsung b-die work on current Ryzen 2000/Intel? I got my trigger finger ready on buying 2 kits of 2x8GB G.Skill 3200 Flare X cl14 for a build after Ryzen 3000 is out. I guess I could also end up with an Intel setup in the unlikely event that Ryzen 3000 disappoints (performance or cost wise). Bad idea?
Posted on Reply
#73
londiste
lsevald said:
How well does 4x8GB single ranked samsung b-die work on current Ryzen 2000/Intel? I got my trigger finger ready on buying 2 kits of 2x8GB G.Skill 3200 Flare X cl14 for a build after Ryzen 3000 is out. I guess I could also end up with an Intel setup in the unlikely event that Ryzen 3000 disappoints (performance or cost wise). Bad idea?
Less memory modules is better than more memory modules. Go for 2x16GB instead of 4x8GB (assuming same speeds etc).
Posted on Reply
#74
mat9v
1/2 divider is kinda useless?
I mean, you get very fast memory and have to push it through a thin straw of halved IF to CPU cores. Sure if IF at 1250Mhz is significantly faster than memory at 5000 then it is not a big issue but if not - you loose that gained speed from increasing memory frequency. The more important thing when running IF at half speed would be timings. Whatever you gain from better timings in memory modules you loose much more in IF lagging behind - memory chip optimisation would gain us 10-20ns while IF running at half speed would loose us a 50ns easily? Consider that without any optimisations we get over 100ns for 2133 memory and corresponding 1066 IF speed and around 80 for 3200 memory and 1600 IF link - that is 20ns lag for 500Mhz of IF link speed. Sure increasing IF link speed would give us decreasingly smaller gains but still... going from 4000 memory with IF at 2000 to IF at 1000 would "kill the performance dead" ;)
It's good only for breaking speed records.

Now, if AMD would be to introduce 2x IF multiplier, that would enable use to lower delays introduced by slow IF link...
Consider somewhat slow memory at 3000 that results in IF running at 1500 and apply 2x to that to get IF running at 3000? IF link with delays on the order of 25ns... and even with low quality memory running at 2133 or 2400 we would get IF running at 2133 and 2400. It would eliminate the problem with IF bus concurrency when communicating with memory controller and PCIEx (GPUs) that can happen in current scenarios. All that, if IF would be able to run at that speeds.... but considering that some memory kits are able to run 3733 on ZEN+ and that means 1866 for IF it is not really far from 2133...

One can dream of course.
Posted on Reply
#75
londiste
mat9v said:
I mean, you get very fast memory and have to push it through a thin straw of halved IF to CPU cores. Sure if IF at 1250Mhz is significantly faster than memory at 5000 then it is not a big issue but if not - you loose that gained speed from increasing memory frequency. The more important thing when running IF at half speed would be timings. Whatever you gain from better timings in memory modules you loose much more in IF lagging behind - memory chip optimisation would gain us 10-20ns while IF running at half speed would loose us a 50ns easily?
What Ryzen so far benefits from with fast memory is the increased IF link speed between CCXs which should be slower no matter what memory timings or speed gets to be.
Edit: reading up on IF speeds in current Zen, I assume AMD doubled the width of IF links from CCX to RAM, otherwise the divider would also limit RAM bandwidth.
mat9v said:
Now, if AMD would be to introduce 2x IF multiplier, that would enable use to lower delays introduced by slow IF link...
Consider somewhat slow memory at 3000 that results in IF running at 1500 and apply 2x to that to get IF running at 3000?
This is the entire problem and reason for introducing the divider - IF cannot run at that high a clock. There is hope that AMD has improved IF in Zen2 but we will ahve to wait and see what approach they have taken with it.
Posted on Reply
Add your own comment