ASUS' Crosshair VI Hero now supports 3200MHz+ memory multipliers

Wow... Was the bios that was released on launch pre alpha? Has anyone tried new memory profiles. I am very interested in scaling performance if the board can actually pull off those memory speeds.
 
Wow... Was the bios that was released on launch pre alpha? Has anyone tried new memory profiles. I am very interested in scaling performance if the board can actually pull off those memory speeds.

Lots of people on Reddit claim to have been getting excellent results:

Reddit Users:

-aylan1196:
"Finally gskillz 3600 32gb 4 8xgb running at 3466 with 1.37 v and my CPU at 4.0 1.4v .... thanks ASUS at last I can push this amd beast build."


-ShikatsuWatercooled (XFX R9 Fury | R7 1800X):
"Finally got my G.Skill memory (F4-3600C16D-16GVK) running at it's XMP specs: 3600MT/s CL16-16-16-36 with the x18 strap.

Great job from AMD and ASUS!"


-digitalfrost (1800X@4Ghz | ASUS CH6 | G.Skill 3200Mhz CL14 | R9 290X):
"Insane BIOS, my RAM is currently at 3466Mhz.

So glad I got this board."


-MaldiavoloSapphire (Tri-X R9 Fury):
"AGESA 1.0.0.6 is what is allowing the higher speeds. It's from AMD."
 
Last edited:
I'm on 9945, running at 3466mhz @ 15-15-15-35-51trc-1T. Still can't boot 3600, so it's not quite there *yet*, but every update is one step closer!
 
You noticing improved performance with every update?
Honestly, I haven't done *extensive* testing, but using AIDA64s memory and cache benchmark, 3466 nets me the fastest copy speed and lowest latency so far. I've run 2966 @ 12-12-12-24, 3200@14-14-14-34, 3333 @ 15-15-15-35, and 3466@15-15-15-35 scores the highest so far. Cinebench R15 also sees a tiny bump (1699, my highest score so far). My 1700 seems to be an OCing turd though, as it needs crazy volts, so it's staying at 3.8Ghz for now. But in the small amount of testing I've done, I'm seeing improvements. And more importantly, it's stable. I think! :)
 
Last edited:
I guess memory timings don't matter so much and just the overall speed throughput is what matters for Ryzen. Pretty interesting. Hopefully you can clock better with a future BIOS. Hitting 4ghz and running memory at 3466 would be quite a feat
 
4Ghz is beyond hope for my CPU I'm afraid. My ram is awesome, since i paid extra $$$ for 3600 @ CL15. If I can get that running at 3600, I'll be happy. Almost there, thanks to Asus releasing these test bioses for us to tinker around with. The new ones open up a slew of individual memory timings, the ability to choose 1T or 2T, etc in addition to the AGESA updates. CH6 owners should be all over those, especially the hard core tinkerers!
 
I have F4-3600C16D-16GTZSW modules with the C6H. Using the 9945 bios, I am running 3600MHz with 16-16-16-16-36-1T. It boots consistently, and in general use has been flawless, but I need to tune further to pass torture testing.

You will need to learn some new approaches with using things like CLDO_VDDP settings, trDRDsc_sm, and tuning ProcODT is more important.
 
I guess memory timings don't matter so much and just the overall speed throughput is what matters for Ryzen. Pretty interesting. Hopefully you can clock better with a future BIOS. Hitting 4ghz and running memory at 3466 would be quite a feat

With Ryzen memory latencies are a bit larger than what is seen on Intel, so timings could have an effect on things. It is something that will need to be tested in the future.

There is a reason why the last AGESA update focused on lowering memory latencies.

I will include a table from Crucial below to help explain that memory latency is more than just CAS latency. IE, clocks and CAS timings are a factor.

c3-speed-vs-latency-table.png
http://www.crucial.com/usa/en/memory-performance-speed-latency

With Ryzen's natural memory latency being a little higher than Intel's, AMD could potentially benefit from memory with tighter CAS timings, though as always this will benefit select scenarios only.

This is something that will require dedicated testing in the future.
 
Without the new microcode update from AMD and running bios 1002 I've been running my Corsair vengeance Red Led 3200mhz at 3200mhz and it's Hynix memory, I have the R7 1700 @3.9ghz.
 
Last edited:
With Ryzen memory latencies are a bit larger than what is seen on Intel, so timings could have an effect on things. It is something that will need to be tested in the future.

There is a reason why the last AGESA update focused on lowering memory latencies.

I will include a table from Crucial below to help explain that memory latency is more than just CAS latency. IE, clocks and CAS timings are a factor.

c3-speed-vs-latency-table.png
http://www.crucial.com/usa/en/memory-performance-speed-latency

With Ryzen's natural memory latency being a little higher than Intel's, AMD could potentially benefit from memory with tighter CAS timings, though as always this will benefit select scenarios only.

This is something that will require dedicated testing in the future.

I know it's higher than Intel. But it still doesn't matter. As we have various benches already done it is clear latency is not as important to Ryzen as is the overall throughput. Timings can go up but increasing the speed gives higher performance. Seeing as 3600 is the limit and no one has gone over it yet, we don't know where the ideal speed/timing curve starts to drop off. Because going from 34400 to 3600 still gives a boost to performance, although less so compared to 2933 to 3200.
 
I dont think AMD should try to push the 3200 MHz to higher frequencies.
They should focus on lifiting the 2133/2400 Mhz people up to 3000/3200 MHz.
 
I dont think AMD should try to push the 3200 MHz to higher frequencies.
They should focus on lifiting the 2133/2400 Mhz people up to 3000/3200 MHz.

That is what AGESA 1006 is aimed for. I have heard a lot of reports of people achieving higher frequencies than before with the new AMD code.

The higher multipliers just give AMD users more options, especially on the high-end. The main benefit here will be better support for more DDR4 memory types.
 
I know it's higher than Intel. But it still doesn't matter. As we have various benches already done it is clear latency is not as important to Ryzen as is the overall throughput. Timings can go up but increasing the speed gives higher performance. Seeing as 3600 is the limit and no one has gone over it yet, we don't know where the ideal speed/timing curve starts to drop off. Because going from 34400 to 3600 still gives a boost to performance, although less so compared to 2933 to 3200.

Yeah, from the tests I've seen, memory timings does not play as big of a role as memory speed.
 
Back
Top