Agesa 1006 Update

[DuronBurgerMan]

Some of you may remember a ranty thread of mine some time back about Agesa 1004, and all the problems it created. Well, Agesa 1006 fixed everything. I'm running the beta BIOS (803) for my Asus X370 Prime, and now my memory is back up to 2800. 2933 *sort of* works.

On the original BIOS 5xx, I was running 2933 at CL 15 (though it always reported 16 in CPUID, for whatever reason). Agesa 1004 came and reduced me to 2400 CL 15, or 2666 with really relaxed timings (I forget what, exactly, but it was really crappy. Wound up reverting to 2400).

Anyway with Agesa 1006, 2800 worked right out of the box with the XMP settings. And 2933 booted, but then behaved... odd. I think I might be able to tweak it a bit and get it stable. But if not, 2800 CL 15 is not bad. I'm running 32GB in two DIMMs, dual rank. So I'm, quite literally, one of the WORST configurations for RAM overclocking on this platform.

Agesa 1006 also benchmarks a bit better. If you remember, my Cinebench and AotS testing had 1004 @ 2400 RAM > Original @ 2933 RAM. Agesa 1006 @ 2800 RAM is now well over 1004 @ 2400 RAM. I'm a good 6% higher than the original, shipped BIOS at the same CPU speed.

Stability is good. Overclocking on the CPU is, so far, basically the same. 4 GHz practical ceiling. It can boot at 4.1 and 4.2, but stability and temps not so good there. I'm running air cooling, so it's possible water cooled folks may see improvement. I don't know.

Given that this is a beta BIOS, it's very good. So far, Agesa 1006 update looks promising. AMD delivered this time.

 

[ManofGod]

Excellent. I ended up bricking one of my boards and send it off to Asus. Since I did not want to wait, I bought a Asrock X370 Taichi for home, swapped my working X370 Pro into my 2nd computer and left it at 0801 bios. Unfortunately, during my farting around with ram swapping, I damaged some cpu pins on my 1700 when it came out of the socket with the heatsink. Bought a 1600 since it was cheaper and the second computer is running fine. (2933 speed on the memory, which is 3000 speed memory.) Unfortunately, my 2800 ram is not running above 2133 speed with the 1.0.0.6 update which is why I was trying the ram swap.

Anyways, my friend is going to try to straighten the pins and I should eventually receive my board back from RMA. I am hoping it will work but, we will see. (I am not overclocking my 1600 and I am leaving the insides of these machines alone for a long while.) I have not damaged a cpu nor bricked a board in over a decade, talk about timing. Not trying to steal your thunder or thread, glad your machine is running better now.

 

[jbc029]

Same experience here. Running 2x16 GB @ 2400 12-12-12-30 for last couple months with 1.0.0.4. 1.0.0.6 was the first time my XMP outright worked on my ASRock B350 Pro4. No cold boot issues, though I lost 25 MHz off the top of my stable OC (down to 3.95 from 3.975). Now running at 16-15-15-35 @ 2933.

 

[pgaster]

I agree, but there is one potential catch with these bios updates, at least on Asus: Command Rate
After enabling XMP, or Asus calls their version DOCP, the command rate will typically be set at 1T, but most RAM should have 2T if you look at the specs. Many people might not notice this, and maybe it won't even cause problems for some people, but it's probably best to run RAM at proper spec just to be safe.
The most logical thing most people will do it just change the 1T or 2T setting to 2T, save, and forget about it. Unfortunately this doesn't work on Asus at least. You still only get 1T.
For 2T you have to go to a setting they call Gear Down and disable that. Leave the 1T or 2T setting on Auto.

 

[sirmonkey1985]

if you guys could put what brand/model ram you're using on what model board i think it would definitely help out others that are just browsing or searching the forums trying to figure out what ram to buy for their setup.. just a suggestion since i notice it happening in a lot of threads(not just on this forum).

 

[jbc029]

if you guys could put what brand/model ram you're using on what model board i think it would definitely help out others that are just browsing or searching the forums trying to figure out what ram to buy for their setup.. just a suggestion since i notice it happening in a lot of threads(not just on this forum).

MB - ASRock B350 Pro4, BIOS 2.6.0
Memory kit - Gskill TridentZ 2x16 F4-3000C15D-32GTZ

 

[Formula.350]

I agree, but there is one potential catch with these bios updates, at least on Asus: Command Rate
After enabling XMP, or Asus calls their version DOCP, the command rate will typically be set at 1T, but most RAM should have 2T if you look at the specs. Many people might not notice this, and maybe it won't even cause problems for some people, but it's probably best to run RAM at proper spec just to be safe.
The most logical thing most people will do it just change the 1T or 2T setting to 2T, save, and forget about it. Unfortunately this doesn't work on Asus at least. You still only get 1T.
For 2T you have to go to a setting they call Gear Down and disable that. Leave the 1T or 2T setting on Auto.

Problem is, that's usually what it's rated at in.... XMP.... which is set to 2T because it was required to operate on INTEL chips. You can definitely try the settings that are programmed in XMP, but in all practical applications you should only use them as a guideline. Like I don't even know if we can still apply all the memory tweaking knowledge accumulated from past AMD chips. I seem to recall that the FAW on my Llano APU was temperamental, and the XMP profile's value was pretty close to what I'd have to run it at for the speeds I was hitting. However, my current kit lists the FAW as like, 39T in the profile at 3200. My board, on Auto, sets it to 34T. Yet it can further be tweaked down into the mid-low 20s!

But yes, some kits likely do need 2T due to the chips used, but like with Corsair they don't generally stick with one chip for the same part number kit, and will use chips that are capable of the original kit advertised timings. Therefore, between revisions, one may work flawless with 1T and another will require 2T.

I can only speak to my current kit's XMP profile, but G.Skill didn't program in a tCR for it, it's a blank line. Works perfect at 1T at 3200, and I assume at 3333, but I've not played around enough with 3333 to be Prime stable, so I dunno if 2T will make a difference or not.



As far as AGESA goes, I do indeed like 1.0.0.6 a lot more than 1.0.0.4! Not just for the subtimings finally being unlocked, but because it let me use my RAM @ 3200 again

 

[ManofGod]

I have really, really good news...... So, my friend straightened out the pins on my damaged 1700 and well I was waiting for my new X370 Pro, I decided to try the processor in my existing X370 Pro. Shoot, only 16GB of the 32GB of ram was shown as usable in the bios, oh well. So, I boxed that up and put my 1600 back in and everything worked with that, as per usual.

I received my new X370 Pro back from RMA so I decided to try the processor again but in this board now. Therefore, I pulled 16 GB of ram from my other Pro, installed it on the bench and all 16GB showed. Then I moved it to the other 2 slots and they worked in there as well. Now, I figured, why not try all 32GB in this new board. BAMMMM! It all worked! So, I have a fully working 1700 in a brand new fully working X370 Pro with the latest bios installed.

The other board and processor will be used in a build I am making for a friend in August. (He is paying for it so now, the only cost I incurred was the $50 or so dollars higher cost of the Asrock X370 Taichi that I have installed at home. Too say I am happy would be an understatement, to say I have been blessed would be just about right.

 

[pgaster]

Unfortunately, during my farting around with ram swapping, I damaged some cpu pins on my 1700 when it came out of the socket with the heatsink.

I had something similar happen with my 1500X. The heatsink would just not come off with some twisting, as has pretty much every other AMD heatsink I've pulled off since the Socket A days. I was mostly twisting back and forth with some gentle pulling/tiping and all of a sudden it just popped off. I was relieved it came loose for a split second until I saw the empty socket and then I just about crapped my pants. I never had that happen before. Thankfully I didn't have any bent pins, but that was probably close to a miracle.
I also had a 1700, but that heatsink came off much easier. After I thought about it a bit, I think I know why. I took the 1700 heatsink off when it was cold. The computer hadn't been on that day yet at all. The 1500X was different. I took that heatsink off right after it had been running for a few hours. I bet that stock TIM is more like glue when it's warm. If you need to take off the stock heatsink for the first time, do it cold. At least I think that's the key.

 

[DooKey]

I have really, really good news...... So, my friend straightened out the pins on my damaged 1700 and well I was waiting for my new X370 Pro, I decided to try the processor in my existing X370 Pro. Shoot, only 16GB of the 32GB of ram was shown as usable in the bios, oh well. So, I boxed that up and put my 1600 back in and everything worked with that, as per usual.

I received my new X370 Pro back from RMA so I decided to try the processor again but in this board now. Therefore, I pulled 16 GB of ram from my other Pro, installed it on the bench and all 16GB showed. Then I moved it to the other 2 slots and they worked in there as well. Now, I figured, why not try all 32GB in this new board. BAMMMM! It all worked! So, I have a fully working 1700 in a brand new fully working X370 Pro with the latest bios installed.

The other board and processor will be used in a build I am making for a friend in August. (He is paying for it so now, the only cost I incurred was the $50 or so dollars higher cost of the Asrock X370 Taichi that I have installed at home. Too say I am happy would be an understatement, to say I have been blessed would be just about right.

Good deal. So far I've been very happy with my X370 Pro. Solid board. My wife is enjoying the PC and it is BOINCing very well. Ryzen is a solid purchase for all but the most demanding gamers. I'm one of those but I have my Intel rig for that. I'm getting the best of both worlds.

 

[sirmonkey1985]

I had something similar happen with my 1500X. The heatsink would just not come off with some twisting, as has pretty much every other AMD heatsink I've pulled off since the Socket A days. I was mostly twisting back and forth with some gentle pulling/tiping and all of a sudden it just popped off. I was relieved it came loose for a split second until I saw the empty socket and then I just about crapped my pants. I never had that happen before. Thankfully I didn't have any bent pins, but that was probably close to a miracle.
I also had a 1700, but that heatsink came off much easier. After I thought about it a bit, I think I know why. I took the 1700 heatsink off when it was cold. The computer hadn't been on that day yet at all. The 1500X was different. I took that heatsink off right after it had been running for a few hours. I bet that stock TIM is more like glue when it's warm. If you need to take off the stock heatsink for the first time, do it cold. At least I think that's the key.

i think the reason we're seeing it so much now with ryzen is due to how friggin flat the heatspreader is compared to past cpu's. the last time i ever had issues was with the socket 754 clawhammers due to how big the heatspreaders were.

 

[tangoseal]

The latest bios from Asus 1201 for the CH 6 and this Agesa update is absolutely ROCK solid. I am getting 3.9g Pstate OC @ 3200 Ram and its just rocking and rolling. I am going to try and do an upgrade to 32GB and see if I can get that to run at 3200 stable also. Then I will be absolutely flabbergasted!

Lets hope that boards release with this kind of stability for Threadripper as I am planning to be a post 1 month adopter once reviews and the issues are mostly discovered an beginnings of hashing them out are evident.

 

[Pieter3dnow]

Problem is, that's usually what it's rated at in.... XMP.... which is set to 2T because it was required to operate on INTEL chips. You can definitely try the settings that are programmed in XMP, but in all practical applications you should only use them as a guideline. Like I don't even know if we can still apply all the memory tweaking knowledge accumulated from past AMD chips. I seem to recall that the FAW on my Llano APU was temperamental, and the XMP profile's value was pretty close to what I'd have to run it at for the speeds I was hitting. However, my current kit lists the FAW as like, 39T in the profile at 3200. My board, on Auto, sets it to 34T. Yet it can further be tweaked down into the mid-low 20s!
But yes, some kits likely do need 2T due to the chips used, but like with Corsair they don't generally stick with one chip for the same part number kit, and will use chips that are capable of the original kit advertised timings. Therefore, between revisions, one may work flawless with 1T and another will require 2T.

I can only speak to my current kit's XMP profile, but G.Skill didn't program in a tCR for it, it's a blank line. Works perfect at 1T at 3200, and I assume at 3333, but I've not played around enough with 3333 to be Prime stable, so I dunno if 2T will make a difference or not.
As far as AGESA goes, I do indeed like 1.0.0.6 a lot more than 1.0.0.4! Not just for the subtimings finally being unlocked, but because it let me use my RAM @ 3200 again

I have seen some posts on overclocker.net which shows what you can do with Samsung B die for rated kits CL14 and CL15+. Those timings work with Ryzen not Intel , settings thing manually is the way to go on this platform The Intel feature is pretty good for getting the "normal" OC to work but sadly does not translate very well to the AM4 platform if you look at the ratings The Stilt got from those tests.

He posted a bunch of pics which show very tight timings running above 3200mhz and they looked pretty good still on "default 1.35" voltage It seems that Ryzen can get some tweaked settings on the platform.

Too say I am happy would be an understatement, to say I have been blessed would be just about right.

Maybe the RMA department has a local priest blessing the new boards they ship out

 

[Formula.350]

I have seen some posts on overclocker.net which shows what you can do with Samsung B die for rated kits CL14 and CL15+. Those timings work with Ryzen not Intel , settings thing manually is the way to go on this platform The Intel feature is pretty good for getting the "normal" OC to work but sadly does not translate very well to the AM4 platform if you look at the ratings The Stilt got from those tests.

He posted a bunch of pics which show very tight timings running above 3200mhz and they looked pretty good still on "default 1.35" voltage It seems that Ryzen can get some tweaked settings on the platform.

I think even with the XMP profiles, they're tuned for a middle-ground, in case a memory controller on the Intel chip just doesn't want to pull off the subtimings the rest of the chips are, and that will avoid any problems. However, it may also simply be a case of the RAM brands not even applying any 'tweaks' to the subtimings and just basing it all off the math; granted, some of the specs are what Intel requires them to be in order to get XMP certified, but given they're so dang close to what Ryzen sets with Auto, I just can't help but think it's "by the numbers". JEDEC might not have official support for high speeds (2800 is the highest I've seen in their documents), but they've laid out enough that you can scale the numbers. And, realistically, the XMP are really only about getting the modules to run at the speed. Sure, the manufacturers will definitely be dropping in some tweaks to subtimings that don't have pre-defined requirements from Intel, but I don't think there's a whole lot of headroom left by that point heh Which is why when we see RAM comparisons they're generally really close, with the bigger variances coming down to different ICs between sticks.


Yea, I had seen those when I was posting stuff in the RAM Speeds thread. I saved em all to my laptop for reference I'ma go play with them right now, matter of fact!