cancel
Showing results for 
Search instead for 
Did you mean: 

Community Update #4: Let's Talk DRAM!

Staff
Staff
18 72 150K

Welcome to the fourth installment of the Ryzen Community Updates! If you’re checking into this series for the first time, this is where we let our community know about all the exciting updates that are on their way for the AMD Ryzen™ processor. We’ve covered a lot of ground in a short amount of time: game updates, new Windows® power plans, stability improvements, feature additions and much more. Today’s update is the one many of you have been most excited about: the AGESA that focuses on overclocked memory. There’s some great news for IOMMU/ACS users, too!

What is an “AGESA”?

AGESA is an acronym that stands for “AMD Generic Encapsulated System Architecture.” As a brief primer, the AGESA is responsible for initializing AMD x86-64 processors during boot time, acting as something of a “nucleus” for the BIOS for your motherboard. Motherboard vendors take the core capabilities of our AGESA updates and build on them with their own “secret sauce” to create the BIOS files you download and flash. Today, the BIOS files for AMD AM4 motherboards are largely based on AGESA version 1.0.0.4.

Beginning this month, as we promised to you, we began beta testing a new AGESA (v1.0.0.6) that is largely focused on aiding the stability of overclocked DRAM (>DDR4-2667). We are now at the point where that testing can begin transitioning into release candidate and/or production BIOSes for you to download. Depending on the QA/testing practices of your motherboard vendor, full BIOSes based on this code could be available for your motherboard starting in mid to late June. Some customers may already be in luck, however, as there are motherboards—like my Gigabyte GA-AX370-Gaming5 and ASUS Crosshair VI—that already have public betas.

Starting With Virtualization

If you’re the kind of user that just needs (or loves!) virtualization every day, then AGESA 1.0.0.6-based firmware will be a blessing for you thanks to fresh support for PCI Express® Access Control Services (ACS). ACS primarily enables support for manual assignment of PCIe® graphics cards within logical containers called “IOMMU groups.”  The hardware resources of an IOMMU group can then be dedicated to a virtual machine.

This capability is especially useful for users that want 3D-accelerated graphics inside a virtual machine. With ACS support, it is possible to split a 2-GPU system such that a host Linux® OS and a Windows VM both have a dedicated graphics cards. The virtual machine can access all the capabilities of the dedicated GPU, and run games inside the virtual machine at near-native performance.

This is certainly a complicated setup for most users, but I have no doubt that there will be a whole lot of you enthusiastically nodding at this news. We’re grateful for your feedback and your patience, and we hope the new support for ACS serves you well.

What's Next For Memory

AGESA 1.0.0.6 officially adds 26 new parameters that can improve the compatibility and reliability of DRAM, especially for memory that does not follow the industry-standard JEDEC specifications (e.g. faster than 2667, manual overclocking, or XMP2 profiles).

The following table spells out all the new parameters, and provides a few words on what they do. Keeping in mind that this is overclocking territory, manual or automated control of these parameters should nevertheless make it a little more straightforward to use DDR4-3200 modules—or faster if you have the talent!1


ParameterFunctionValues
Memory clocksAdded dividers for memory clocks up to DDR4-4000 without refclk adjustment. Please note that values greater than DDR4-2667 is overclocking. Your mileage may vary (as noted by our big overclocking warning at the end of this blog).133.33MT/s intervals (2667, 2933, 3067, 3200, 3333, 3466, 3600, 3733, 3866, 4000)
Command rate (CR)The amount of time, in cycles, between when a DRAM chip is selected and a command is executed. 2T CR can be very beneficial for stability with high memory clocks, or for 4-DIMM configurations.2T, 1T
ProcODT (CPU on-die termination)A resistance value, in ohms, that determines how a completed memory signal is terminated. Higher values can help stabilize higher data rates. Values in the range of 60-96 can prove helpful.Integer values (ohms)
tWCL/tWL/tCWLCAS Write Latency, or the amount of time it takes to write to the open memory bank. WCL is generally configured equal to CAS or CAS-1. This can be a significant timing for stability, and lower values often prove better.Integer values (cycles)
tRCRow cycle time, or the number of clock cycles required for a memory row to complete a full operational cycle. Lower values can notably improve performance, but should not be set lower than tRP+tRAS for stability reasons.Integer values (cycles)
tFAWFour activation window, or the time that must elapse before new memory banks can be activated after four ACTIVATE commands have been issued. Configured to a minumum 4x tRRD_S, but values >8x tRRD_S are often used for stability.Integer values (ns)
tWRWrite recovery time, or the time that must elapse between a valid write operation and the precharging of another bank. Higher values are often beneficial for stability, and values < 8 can quickly corrupt data stored in RAM.Integer values (ns)
CLDO_VDDP

Voltage for the DDR4 PHY on the SoC. Somewhat counterintuitively, lowering VDDP can often be more beneficial for stability than raising CLDO_VDDP. Advanced overclockers should also know that altering CLDO_VDDP can move or resolve memory holes. Small changes to VDDP can have a big effect, and VDDP cannot not be set to a value greater than VDIMM-0.1V (not to exceed 1.05V). A cold reboot is required if you alter this voltage.

Sidenote: pre-1.0.0.6 BIOSes may also have an entry labeled “VDDP” that alters the external voltage level sent to the CPU VDDP pins. This is not the same parameter as CLDO_VDDP in AGESA 1.0.0.6.

Integer values (V)
tRDWR / tWRRDRead-to-write and write-to-read latency, or the time that must elapse between issuing sequential read/write or write/read commands.Integer values (cycles)
tRDRD / tWRWRRead-to-read and write-to-write latency, or the time between sequential read or write requests (e.g. DIMM-to-DIMM, or across ranks). Lower values can significantly improve DRAM throughput, but high memory clocks often demand relaxed timings.Integer values (cycles)
Geardown ModeAllows the DRAM device to run off its internally-generated ½ rate clock for latching on the command or address buses. ON is the default for speeds greater than DDR4-2667, however the benefit of ON vs. OFF will vary from memory kit to memory kit. Enabling Geardown Mode will override your current command rate.On/Off
RttControls the performance of DRAM internal termination resistors during nominal, write, and park states.Nom(inal), WR(ite), and Park integers (ohms)
tMAWMaximum activation window, or the maximum number of times a DRAM row can be activated before adjacent memory rows must be refreshed to preserve data.Integer values (cycles)
tMACMaximum activate count, or the number of times a row is activated by the system before adjacent row refresh. Must be equal to or less than tMAW.Integer values (cycles)
tRFCRefresh cycle time, or the time it takes for the memory to read and re-write information to the same DRAM cell for the purposes of preserving information. This is typically a timing automatically derived from other values.Integer values (cycles)
tRFC2Refresh cycle time for double frequency (2x) mode.  This is typically a timing automatically derived from other values.Integer values (cycles)
tRFC4Refresh cycle time for quad frequency (4x) mode. This is typically a timing automatically derived from other values.Integer values (cycles)
tRRD_SActivate to activate delay (short), or the number of clock cycles between activate commands in a different bank group.Integer values (cycles)
tRRD_LActivate to activate delay (long), or the number of clock cycles between activate commands in the same bank group.Integer values (cycles)
tWRWrite recovery time, or the time that must elapse between a valid write operation and the precharging of another bank. Higher values are often better for stability.Integer values (ns)
tWTR_SWrite to read delay (short), or the time between a write transaction and read command on a different bank group.Integer values (cycles)
tWTR_LWrite to read delay (long), or the time between a write transaction and read command on the same bank group.Integer values (cycles)
tRTPRead to precharge time, or the number of clock cycles between a READ command to a row and a precharge command to the same rank.Integer values (cycles)
DRAM Power DownCan modestly save system power, at the expense of higher DRAM latency, by putting DRAM into a quiescent state after a period of inactivity.On/Off

Until next time

What are you interested in hearing more about in our next AMD Ryzen Community Update? Let us know on Twitter @AMDRyzen!

Robert Hallock is a technical marketing guy for AMD's CPU division. His postings are his own opinions and may not represent AMD’s positions, strategies or opinions. Links to third party sites are provided for convenience and unless explicitly stated, AMD is not responsible for the contents of such linked sites and no endorsement is implied.


1. WARNING: Overclocking memory will void any applicable AMD product warranty, even if such overclocking is enabled via AMD hardware and/or software.  This may also void warranties offered by the system manufacturer or retailer or motherboard vendor.  Users assume all risks and liabilities that may arise out of overclocking memory, including, without limitation, failure of or damage to RAM/hardware, reduced system performance and/or data loss, corruption or vulnerability.  GD-112

Tags (4)
72 Comments
Challenger
Challenger

Sweet !

Forerunner
Forerunner

I am extremely excited about all the improvements in such a short time - can't wait to see further implementations by motherboard manufacturers!

So far I've been on all the beta BIOS' from Asus for the Crosshair VI Hero and things are really looking promising!

Adept I
Adept I

Very exciting news. Now I need MSI to get a beta to me.

Journeyman III
Journeyman III

"DDR4-4000 without refclk adjustment" - can't wait to get my kit up and running hell yeah!!! Awesome job AMD!!!

Forerunner
Forerunner

superb news, specially IOMMU and ACS!!!!! happy panda. The forum with beta bios for Aorus GA-X370-Gaming K7, 5 and 3 is here​.

Nice weekend ahead. Testing new Ryzen AGESA, Monaco F1 GP and Fernando Alonso running Indy 500.

Adept I
Adept I

Ryzen has been out for only a few months but the number of updates/improvements are amazing (especially considering AMD had to "catch up")! My trust in AMD wasn't displaced when opting for a Ryzen 7 series CPU on release. Great job AMD!

Adept II
Adept II

Awsome news. Thanks Robert, and thanks to every one at AMD who is working so hard to get these updates to us.

Adept II
Adept II

@ rhallock

You mentioned you have a Gigabyte GA-AX370-Gaming5 (like I do) with a AGESA 1006 beta UEFI, but well, neither on the official productpage nor in the BETA thread on gigabyte forum I can find it. So where do I get it? Want to test if KVM OVMF GPU passthrough is working now w/o disabling npt.

Regards

Journeyman III
Journeyman III

Already went from Crucial Ballistix Tactical 2666mhz 1.2v to 3200mhz 1.36 with good timings after updating to MSI Tomahawk B350 Arctic beta bios with this. Wasn't able to post at all with over 2666 before using any settings.

For 3333mhz now I would probably need to adjust some of these bios variables listed as it doesn't always post(5 memory tries per post) but when it does, passes all memory tests(even prime95 25min with memory test settings).

Journeyman III
Journeyman III

Yeah I used the link above to get f6d bios last week and with it I was able to get my 32gb (2x16) GSkill 3200mhz ram to 2900 which previously wouldn't run stable beyond 2400. It would run at 2667 but would eventually crash. Still haven't found timings that will boot beyond that but have had very limited time to experiment. I'm on vacation this weekend and will resume efforts Monday evening.

Journeyman III
Journeyman III

What timings did you use??

Journeyman III
Journeyman III

16-17-17-36, this is sold at at 16-17-17-38 for 2666mhz. <- that's quite maximum, if I lower any timings there's no bios post.

EDIT: Said good timings as I didn't need to lower anything from 3000->3333mhz but if I touch anything no post. And even at 3333mhz whatever timings and volts there is rarely posting so I went back to 3200 for now.

EDIT2: At 2666mhz I used 14-13-13-32 but I rather wanted more memory speed on ryzen and now it's possible. I had a quicksave at Prey. It went from 3000mhz->3200mhz +10fps. for 3200 -> 3333mhz it went up +2fps. I don't think it's so important for me to hit that 3333mhz.

Miniboss
Miniboss

Thanks for the community update, keep em coming

Adept I
Adept I

Thanks for the update!

For the most part, the 1006 AGESA update has been working quite well on my motherboard despite the BIOS still labeled under Beta status. I'm currently using a Gigabyte Aorus GA-AX370 K7 with 2x16 GB T-Force Night Hawk 3200 memory and it has been running at its rated speed (3200 at 16-18-18-38) thus far using the F4D BIOS. I just enable the XMP preset and it simply works! I could try fiddling with some of the timings, but I'm pretty content with the performance right now-- maybe I'll tighten up the timings after the BIOS leaves beta status.

Aside from that, thanks for the update and keep up the great work!

Journeyman III
Journeyman III

First off, thanks for the hard work!

I really appreciate you guys taking the time tuning the memory compatibility, however, the new 1006 AGESA on my GA-AX370 Gaming 5 did not help my G.Skill F4-3200C16D-32GTZ (2X16GB) to run at 3200mhz 16-18-18-36 advertised XMP2 speed.

One thing I noticed on the chart you provided though, is that the CR configuration on these RAMs did not work. Such as right now I am running the RAMs at 2933 18-17-17-34 CR=1, even though I configured in the BIOS to have it run at CR=2. So if the BIOS can actually set the CR to what I set it to be, I think it may help on the stability.

Please investigate, and let me know that whether this is Gigabyte's problem or the AEGSA's.

Journeyman III
Journeyman III

hi thanks for improving the ryzen life, i currently have the msi x370 carbon gaming pro and i also bought the G Skill Ryzen ram 3200 called flare x. it will boot at 2933 at the timings that are advertised of 14 14 14 34, but will not boot at 3200 with 14 14 14 34. even if i loosen numbers i havent got a boot in 3200. other than that it runs great and i get great fps and benchmarks on 2933 and 14 14 14 34. i am wondering and hoping that my ryzen and approved flare x g skill ram will be able to get its full potential and run at 3200  with this new agesa bios update, if you can give me some or any answers/hope i would greatly appreciate it thank you

Forerunner
Forerunner

Played around with the latest beta BIOS from Asus (9945) on my C6H and was able to keep my memoryat 3200MHz without needing to increase BCLK - just a straight default selection. Added my DRAM timings (14-14-14-34-48) and tested for a couple of hours on AIDA64X and Prime95. Results are shown here: AMD Ryzen 7 1800X @ 3699.14 MHz - CPU-Z VALIDATOR

The CPU temp shown is CPU-z's crappy output - the actual temp is almost half that (31C/47C load). I have 3600MHz RAM coming in soon so I'll be able to test that as well. Next step is to OC the CPU a bit, but I"m not worried about it - I think I lucked out on the Silicone-lottery for this one as I've been able to push it without any issues to 4.2!

Exciting times indeed!

Journeyman III
Journeyman III

I wonder why is DDR4-2800 excluded? Since 2667+133=2800...

Exemplar
Exemplar

Tested the beta 2.36 UEFI for the ASRock X370 gaming professional.  Initially I could only get 2 DIMMs to boot at 3200 MHz CL14, while 4 DIMMs were limited to 2667 MHz CL14.  With the beta UEFI all DIMMs are now at 3200 MHz CL14.  System posted and booted on the first try and so far no instability detected.  I reran my GEARs 4 benchmark to see how things had improved.

RAM 3200 32GB.png

Here is how things compare to my previous benchmarks.  Keep in mind that all other benchmarks were done with an AGESA 1.0.0.3 based UEFI (1.55)

RAM2.PNG

RAM.PNG

The platform just keeps getting better.  I'll attempt to push things a bit further when the final UEFI for this AGESA are released.

Challenger
Challenger

Good info, I'm hoping Asus will release the new bios containing AGESA 1.0.0.6 soon. I have an Asus Prime X370-Pro waiting to be updated. Although there is discussion on another thread that the memory command rate is an issue that I believe is causing many of the stability and stopcode issues that can be avoided if motherboard manufacturers would make the memory command rate available as an option.

Journeyman III
Journeyman III

Looks like it's the Dual Rank vs Single Rank issue on RAMs that is limiting the speed. My 32GB (2x16) kit is not getting past 2933mhz on my Gigabyte AX370 Gaming 5. However, with the new AGESA 1.0.0.6, it's now allowing much lower timings than before. Even though it's on 2933mhz only, i'm getting 14-14-14-32 1T; I was only able to boot at 18-18-18-34 1T on 1.0.0.5.

I really hope something can be done on the dual rank RAM kits.

Exemplar
Exemplar

Dual rank with 2 DIMMs should perform the same or better than single rank with four.  Here is the original RAM support data from my manual for Ryzen processors.  So buying a 32 GB kit with the same specs, one being dual rank 2X16 and the other single rank 4X8 should favor the dual rank slightly.  However, I'm not sure if this is actually what happens in the real world.

pastedImage_0.png

Adept III
Adept III

I have downloaded the latest two   agesa 1.0.06 beta bios updates for my MSI Tiotanium board and my Ryzen 1800X processor. I had a four dimm configuration all with 8GB DDR4 3200mhz B-die dimms. The updates have had minimal impacxt on memory speed . Playing with all appropraite timings and voltages I was able to only inch up from 2667mhz to 2800 mhz. I expected at least 2933mhz. Too many Ryzen chiips have underperforming IMC. Only a few golden cpus have  the higher performing IMC and they mostly went to reviewers. Veryy disappointed. I am selling off my two 16 GB kits of FlareX DDR4 320 and buying two 16GB double rank G. Skill Trident X ddr4 3200 Samsung B-die 14-14-14-34 in a kit. G. Skill told me even though these 2 chips are double rank they put less stress on the memory controller than a 4 dimm setup. Of course AMD NEVER let this cat out of the bag. I like transparency. There is NOT enough of it in this industry.

Exemplar
Exemplar

Sorry you're having issues,  but it sounds like part of the problem may be that you bought two individual 16 GB kits and combined them for your 4 DIMM setup.  That has historically been problematic, even when the kits are the same speed and manufacturer.  The best bet is to buy a complete 4 DIMM kit, as those DIMMS will have been thoroughly tested together.  With the beta 1.0.0.6 BIOS my 4 DIMM kit is running just fine at CL14 3200MHz, and I'm certainly not a reviewer.

" G. Skill told me even though these 2 chips are double rank they put less stress on the memory controller than a 4 dimm setup. Of course AMD NEVER let this cat out of the bag. I like transparency."

See the table in my previous post, it was put out by AMD when Ryzen lauched.  This particular version is from the manual of my motherboard.  In fact, I'll post it again.

pastedImage_0.png

These were AMDs quoted speeds for the Ryzen memory controller.  You can clearly see that 2 dual rank DIMMS (2400-2667) are rated at higher speed than the 4 single rank (2133-2400).  That information has been there all along.  Not sure how much more transparent they could be.

Challenger
Challenger

It's interesting that my 16GB FlareX DDR4-3200 running on an Asus Prime B350-Plus with the 0613 Bios which is still the AGESA 1.0.0.4 code has no issues running at 3200MHz, the inconsistencies between motherboard manufacturer bios's are puzzling. System stability seems better on the B350 mainstream chipset as opposed to the X370 chipset. I know that the memory command rate and it's availability within some bios's has caused memory issues for many that have it set at 1T vs 2T.

As for the X370 chipset not being able to get DDR4-3200 to run at 3200MHz is very strange.

Exemplar
Exemplar

Hello!

I think the issue there has more to do with the number of DIMMs.  When I only installed two of the DIMMs from my kit (2X8GB) I was able to hit 3200 Mhz with every iteration of the UEFI on my X370 based system.  Even the original 1.5 (AGESA 1.0.0.3) UEFI I had on the board was able to that.

The problem arose when I installed the other two and ran in a 4X8GB configuration.  Then I was limited to 2667MHz CL14.  The AGESA 1.0.0.6 beta UEFI is the first that allows me to run at full rated speed with 4 DIMMs installed.

Challenger
Challenger

I agree, I'm wondering if that's why AMD is using the quad-channel configurations for the next generation of processors.

Adept III
Adept III

I do not think so. The threadripper and series has the same IMC as

Ryzen as far as I know, just a different chipset. Memory overclocking

should be just as challenging on thread rippper as on Ryzen. I am sure

with 4 memory channels it will definitely have greater bandwidth and

thus performance , but do not expect to see much 3600mhz dimms running

at 3600mhz. Better success will occur for thioose only putting one dimm

in each channel.

Adept III
Adept III

I have 4 G.Skill FlareX DDR 3200 dimms . They are Samsung B-die. I

can NOT run them over 2800mhz with the latest agesa 1.00.06 beta bios

1.73 for MSI X370 Titanium. You must be using the refernece clock

overclocking to get above 2800mhz or have a golden chip. I guarantee you

9 out of 10 with the same configuration will never getnear 3200mhz with

4 dimms. RYzen IMC is variable but very few are good enough for 3200mhz

on 4 dimms. I am selling mione and buying two 16GB dimms that are B-die

Trident Z. You have a better chance of getting 3200mhz even on double

rank dimms with 2 dimms versus 4 dimms even if single rank. I confirmed

that with G.Skill technical support.

Adept III
Adept III

There are no 16GB FlareX DDR4 3200 dimms. There are two 8GB dimms

in the 3200mhz kit.Of course two *GB dims , a 16GB kit, will run at

3200mhz. I am talking about a 4 dimm 32GB configuration. Know yor apples

from oranges.

Adept III
Adept III

Do you even know what your talking about? G.Skill FlareX does NOT

make 4 dimm kits at 3200 MHZ.

Challenger
Challenger

That's what I mean I have a 2x8GB DDR4-3200, misunderstood.

Adept III
Adept III

My problem is want and need 32 GB of memory. That is why I am selling

off these two 16GB kits for two 16GB dimms. They will have the same CL

14-14-14-34 settings as my Flare-X. They are Samsung B-die and at very

least I will be able to get to 2933mhz and probably 3200mhz.

Exemplar
Exemplar

I did not claim that they did.  I was merely pointing out the fact that combining kits has historically led to compatibility issues on many different platforms not just Ryzen, and it is small wonder your having issues.   Buying a complete 4X kit from a different G. Skill series will likely lead to better results than taking two FlareX kits and mashing them together.  The 2 DIMM Trident series you are planning on purchasing has a better chance for success as they are a single kit, and dual rank performs the same or better than 4 single rank.

What AMD has told us is that its memory controller should work identically with 1 single rank, 1 dual rank, or 2 single rank DIMMs.  2 dual rank DIMMs may hit the same speeds but may not, while 4 single rank DIMMs WILL see a drop off.   On your Flare X kits,  for example, they were tested in a two single rank config and rated for 3200 speed.  You wouldn't expect merging two kits to yield the same speed, you'd probably only see 2933 or even 2667.

In the case of my kit (a Corsair Dominator 4X8GB) it is rated at 3200 Mhz CL14 and tested with all four DIMMS.  While it was tested on Intel platforms, that does give it a better chance of hitting those speeds.  What that implies however, is that by removing two of the DIMMs and running 2 single rank I should be able to hit even higher speeds (3466 for example).  I haven't attempted that yet, and I'm actually perfectly happy running 3200 CL14 on four DIMMs. 

Exemplar
Exemplar

I 100% agree with os2wiz here.  Threadripper likely will use a doubled IMC configuration, where each controls a set of 2 channels/2DIMMs per channel.  In that configuration you would expect it to perform identically to the current Ryzen chips.  If there is a single memory controller it will likely perform worse, as having to interpolate 4 channels of 2 DIMMs each is way more overhead than the 2 and 2 version

Adept III
Adept III

You know less than you think. I have been doing my own builds while

you were in diapers. I have done this before with no issues. The problem

I have has little or nothing to do with 2 different kits. It is simply

the issue of using 4 dimms on a dual channel system. Most people even

with 4 dimm kits are NOT getting their rated frequency on Ryzen. The IMC

needs work. Hopefully AMD will get it right with the next series of Ryzen.

Adept III
Adept III

I have gone to 2800mhz on all 4 of these dimms. 2933 was obtained but

could not be stabilized. You are telling me things that I alreadsy know.

The Ryzen IMC is not as good as Intels as more often than nolt you would

wind up with a higher frequency than we are obtaining even with the

latest bios for a 4 dimm configuration.

Exemplar
Exemplar

"Of course AMD NEVER let this cat out of the bag. I like transparency. There is NOT enough of it in this industry."

That was your original complaint.  My point was AMD told everyone right from the get go that "its memory controller should work identically with 1 single rank, 1 dual rank, or 2 single rank DIMMs.   Expecting that you could just add another one and have it work at the same speed is a shot in the dark.  If you buy a single rank kit, 2 DIMMs, 4 DIMMs or otherwise, and the highest speed you can get with a single DIMM installed is 3200, it will NOT run at 3200 with all four.  As AMD stated in the chart, there will be a drop off from a single DIMM.   In the case of your FlareX kit, it was sold as a 2x8 at 3200 which is all G. Skill tested.  They make no claim that it will work in a 4X config, they never tested it, and if 3200 is the max you can do with only 2 DIMMs installed it will certainly not reach the same speed with four.

In the case of my Dominator kit, it is rated at 3200 CL14 as a four DIMM kit, meaning it was tested at that speed with all four DIMMs installed.  Granted, it was tested on an Intel platform, and there was no guarantee that Ryzen would produce the same results.  Indeed with AGESA 1.0.0.3, and 1.0.0.4a it did not, and I was limited to 2667 CL14.  But with AGESA 1.0.0.6 I can load the XMP profile and it boots right up, no bclk adjustment required. 

"You have a better chance of getting 3200mhz even on double rank dimms with 2 dimms versus 4 dimms even if single rank. I confirmed that with G.Skill technical support."

You really didn't need too.  AMD already told us that when the product launched.  See the Ryzen memory support table?  Notice than line 4 with the 2 (DRs) is above the line with 4 SRs with a higher speed listed.

Exemplar
Exemplar

I agree.  If you have need of the RAM and want the fastest speed possible, 2 dual rank gives you a better shot than 4 single rank of identically rated memory from the same manufacturer.

Exemplar
Exemplar

"I have done this before with no issues". The problem I have has little or nothing to do with 2 different kits. It is simply the issue of using 4 dimms on a dual channel system. Most people even with 4 dimm kits are NOT getting their rated frequency on Ryzen. The IMC needs work."

I understand that a large percentage of the time, you can buy two identical kits and use them together with no issue.  But, the fact of the matter is a vendor only guarantees a kit to function at the speed labeled, on the platform tested, in the configuration shipped.  If you use the RAM off label you incur the risk that it will not work, end of story.  This is not a problem isolated to Ryzen.  The issue appears consistently in 4 DIMM kits rated for Intel dual channel processors, purchased in duplicate and used in quad channel X series configurations.  Does that mean Broadwell-X has a bad memory controller? 

"You are telling me things that I alreadsy know."

My apologies.   I assumed that when you had to confirm that two DR is better than four SR with G. Skill you were unaware that the information had already been supplied.  Not sure why you said AMD would never let the cat out of bag and weren't transparent since you knew all this already, but oh well.

If it's helpful, I can post the exact timings, voltage, SOC voltage etc I am currently running on my 4 DIMM kit.  But I'm not sure that will help anyone who isn't running the identical kit.

Exemplar
Exemplar

Here are some CPUz shots using the beta 2.36 UEFI.

pastedImage_0.png

No bclk increase needed.

pastedImage_1.png

pastedImage_2.png

Adept III
Adept III

You only have one cpuz shot and it it is useless. Where is the one of

the memory settings????

Exemplar
Exemplar

I have the CPU, Mainboard and Memory tabs displayed.  I can certainly add the SPD or Cache tab as well.

Adept III
Adept III

My email only contained the cpu tab.

Exemplar
Exemplar

Ah, were you able to see the other two in the forum then?  I only included the CPU tab to show I hadn't altered the base clock at all.  As you said, it isn't really necessary beyond that.

Exemplar
Exemplar

Made additional gains with the 2.40 full release UEFI with the 1.0.0.6.  I now have 3333MHz stable at CL14 with four single rank DIMMs installed.  Gears 4 benchmark continues to improve.  While this level of performance increase probably won't reproduce for all titles, it is a good demonstration of what faster RAM can do for Ryzen.

AMD4.png

Adept III
Adept III

This is a very atypical result for 4 dim co figurations

Sent from my Verizon, Samsung Galaxy smartphone

Exemplar
Exemplar

I agree.  Just looking across the forums, it does seem that there are still considerable issues with memory and the Intel speed ratings not reproducing with the Ryzen memory controller.  It doesn't seem to be related to the motherboard manufacturer either, more the specific RAM kit itself.  Hynix RAM continues to perform erratically as far as I can tell.  In the forums for my ASRock board, certain memory configs actually seem to perform worse on the new UEFI vs the older 1.0.0.4a based firmware.  Those experiences do seem to be in the minority however.  Most user see some improvement, how much seems to depend on the kit in question. 

Currently I am still stress testing the system at 3333 and everything seems fine.  Timespy stress test has been run numerous times back to back without issue and a 99.5%+ result.  Once all the testing is done I'll try to step up to 3466 MHz.  As we discussed already, 2 DIMMs at single rank will run better than 4 regardless.  So I am starting to wonder how much more could I get if I pulled 2 DIMMs back out?  3466?  3600?.  I may give that a shot once I hit the wall with 4 DIMMs.  Although, the AGESA 1.0.0.6 update allowing me to gain 667MHz in a 4 DIMM config has exceeded my expectations.

It is worth pointing out as well that my Firestrike scores don't really change with the faster RAM.  As I am always bottlenecked by the GPU, the absolute fps the system generates remains the same.  The standalone "physics" CPU test done by Firestrike doesn't really seem to gain anything from increased RAM speed.  The independent "CPU game" and "CPU Render" stats in the Gears 4 benchmark are largely effected, and mirror what the CPU would actually be doing in that DX12 title.  Looking at those alone, I have gained 40% fps just by increasing RAM speed.  However, as I was 100% GPU bound, even at 2133 MHz, I haven't actually gained any fps.  Who knows, with Vega, I might just be able to appreciate some of those gains.

AMD1.png   AMD2.png

AMD3.png

Exemplar
Exemplar

Bit of an update here.  I have managed to get 3466 MHz stable on 4x8Gb.  I had to play with the ProcODT a bit and change the command rate to 2T, but so far, I have been through HCI memtest 450%!

RAM test 3.PNG

Journeyman III
Journeyman III

CoreFreq_AMD0F_HTT_MC.png

Just need BKDG + Ryzen to program some IMC algorithms