• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Ryzen Owners Zen Garden

I doubt they got new XT processors about to launch this close to Zen 3, so its either new APU's or zen 3 IMO
 
Sold my B450I Aorus board, bought a B550I Aorus Pro AX, the ridiculously heavy and surprisingly affordable one with a full metal backplate. It's a solid (quite literally) board, BIOS is a fair bit better, and the VRM and temps are light-years ahead of its B450 predecessor (and most other ITX boards for that matter). Things look good, except for this:

ryzen master vdram.png

what 1.33v looks like.png


Yeah, that's 1.33V. What the hell is this?? This board overvolts RAM by 0.06V above what you set in BIOS. I know Gigabyte loves to do this as my B450 board allowed up to 0.03V above any given setting, but this is getting ridiculous. I finished installing Windows and was doing some HCI when I noticed the default 1.35V XMP VDRAM setting was taking the RAM up to 1.419V.

Putting the geriatric X99-era 4Gb Samsung E-die through its paces for the first time, still lots of work to do:

aida 3200 cl14 e-die.png 3200 cl14 timings e-die.png
 
Yeah, that's 1.33V. What the hell is this?? This board overvolts RAM by 0.06V above what you set in BIOS. I know Gigabyte loves to do this as my B450 board allowed up to 0.03V above any given setting, but this is getting ridiculous.
I thought my B450 Tomahawk was bad for over voltage with 1.368 above 1.35.
 
I thought my B450 Tomahawk was bad for over voltage with 1.368 above 1.35.

I guess it just takes some getting used to. The difference is (kind of?) consistent, so just mental math every time I wanna adjust VDRAM.

I have absolutely no idea what's going on, but ol' dinosaur 4Gb Samsung E-die that was rated for 3200 16-16-16-36 just straight up booted 3600 16-18-18-38, and trashed the much newer 8Gb Hynix DJR in both AIDA and membench, at comparable voltage, and to add insult to injury it did it with Geardown enabled. o_O

Both are Trident Z sticks, the E-die is a 2x8GB dual rank kit, the DJR is a RGB 2x16GB dual rank kit.

e-die shames djr.png aida 3600 16-18-18-38 e-die!.png e-die 3600.png

What is going on lmao I thought it was ancient and not supposed to even boot 3600 ever, maybe it's really like some users said it was, a sleeper mini B-die

Or maybe the memory traces on this 8-layer board are really that good. My head's spinning, I didn't see this one coming.
 
Yeah, that's 1.33V. What the hell is this?? This board overvolts RAM by 0.06V above what you set in BIOS. I know Gigabyte loves to do this as my B450 board allowed up to 0.03V above any given setting, but this is getting ridiculous. I finished installing Windows and was doing some HCI when I noticed the default 1.35V XMP VDRAM setting was taking the RAM up to 1.419V.
I'm wondering if there is a technical reason for this. There's got to be something specific and not just a bug/defect/glitch.
 
Doesn't it also change depending on the load?
 
I'm wondering if there is a technical reason for this. There's got to be something specific and not just a bug/defect/glitch.

Since GB does this all the time I'm guessing it's for compatibility purposes. You know, pump a lil extra voltage to make sure the auto 1.35V XMP voltage can run all sorts of ICs without problems.

But I seriously question the need for this on a beefy 8-layer ITX board, where the extra voltage can conceivably result in some serious hotboxing in the cases that one might expect to find these boards. And we know that B-die are temperature sensitive, so lose stability when over 50C.

Doesn't it also change depending on the load?

Funny you should ask that. On Auto (1.35V), the voltage occasionally drops a bit, but when manually entering voltage it only ever goes up. Setting is 1.33V in BIOS, at idle it's 1.38V, during heavy HCI Memtest it's continuously 1.392-1.404V, occasionally goes up to 1.416V for a second.

It looks like the memory VRM has no droop. And no LLC that's at least accessible to the user.

So the E-die does 3600 16-17-17-36 completely stable without a bump in voltage from 1.38V. Sadly, while 16-17-16 benched fine, it's gonna need more than 1.38V to bring tRP and tRCD below 17. And I can't give it more voltage, NCASE M1 gets real hot in games.

e-die timings 3600.png memtest e-die normal length 226.png memtest e-die 16-17-17-36.png

Overall this was a very pleasant surprise from a kit that Thaiphoon believed was "4Gb D-die". tRFC 351 @ 3600 is 195ns, which is either on the high end for B-die or just outside B-die range depending on who you ask. Might be able to go lower. Still, nice to have better performance than CJR/DJR and Rev.E, all at a very reasonable voltage.
 
Since GB does this all the time I'm guessing it's for compatibility purposes. You know, pump a lil extra voltage to make sure the auto 1.35V XMP voltage can run all sorts of ICs without problems.
While that is a fair point, 1.419v seems a bit much. I'm wondering of there is a power drain effect happening that the extra voltage compensates for?
 
Out of curiosity to see what level of OC & RAM speed 1:1 is needed for Zen+ match Zen2@stock?
Managed to do a 4.4GHz manual OC on my trusty old 2600X & discovered (at least according to CPU-Z) it now matches Zen2 in single thread performance albeit with 3533 CL16 ram 1:1

CPUZ-Bench@4.4_3533.JPG
 
Isn't Zen+ and Zen1 always 1:1? You don't get control over the Infinity Fabric dividers.
Besides that, nice IPC difference on Zen2.
 
Managed to do a 4.4GHz manual OC on my trusty old 2600X
I don’t want to ruin it for you with the OC. Just that I’d be a little concerned with that high voltage of 1.431.

Edit: Seems the voltage should be between 1.375-1.4 for Zen+ when OC?
 
Last edited:
I don’t want to ruin it for you with the OC. Just that I’d be a little concerned with that high voltage of 1.431.

Edit: Seems the voltage should be between 1.375-1.4 for Zen+ when OC?
Ain't ruining anything, this is not a daily OC. I've had this chip up to 4.5GHz all core/thread before purely for benching.
The Vdroop on this board is typical for it's price range. Even fiddling with LLC options.
Enabling 'extreme' in LLC is the only way out. But didn't want to do for the purposes of this test.

Isn't Zen+ and Zen1 always 1:1? You don't get control over the Infinity Fabric dividers.
Besides that, nice IPC difference on Zen2.
You can change the synchronicity with BLCK . Some boards will do it, some won't.
 
I don't know if you guys remember me seeking help with memory issues back around page 89. My memory couldn't run at XMP sometimes, didn't overclock at all, and sometimes couldn't run anything above 2133mhz sometimes, and I had unusually high latency when I could get it to run at XMP speeds. We concluded that it was corsair memory not playing well with AMD systems. I finally decided to buy new ram and see if it makes a difference. I will post results later this week.
 
Hey guys, I just found this. Looks interesting. I am setting up to play with it.
 
Hey guys, I just found this. Looks interesting. I am setting up to play with it.
I just got done messing with it. I'm leaving posts in this thread.
 
I don't know if you guys remember me seeking help with memory issues back around page 89. My memory couldn't run at XMP sometimes, didn't overclock at all, and sometimes couldn't run anything above 2133mhz sometimes, and I had unusually high latency when I could get it to run at XMP speeds. We concluded that it was corsair memory not playing well with AMD systems. I finally decided to buy new ram and see if it makes a difference. I will post results later this week.
Sooooo.............. it must be the cpu because even this memory doesn't like to boot up at the XMP speeds.
 
Swapped cases, stuck more fans in, cleaned heatsink, got a good 10c drop on the core. Decided to play around with CPU-z benchies and here we go.
d443f27215.png


Still a 0.5v vdroop and I can't shut that off since the BIOS doesn't seem to care what level LLC it's on, so maybe it can stay there and whatever. I'm convinced that if I put in the time I can keep 4.6 for day to day use since under load that was around 1.35v and 1.4v idle (similar to AMD stock spec from what I've seen)

Take that 3800x. I can match and beat. :rockout:
 
well shit, thats a high clocking 3700x
 
I'm tempted to see if I can go faster with some sketchy voltages, but between this, the SR-2 rig, and my 4790k stuff I'm pooped out for the day.
 
Sooooo.............. it must be the cpu because even this memory doesn't like to boot up at the XMP speeds.
What's the new ram?
 
What's the new ram?
I went with the F4-3600C16D-32GTZNC rather than the patriot because it was on sale and had the "made for ryzen" thing going on. It is in the QVL list. I have tried two different motherboards, two different sets of memory, and I still have trouble running ram at XMP speeds. That only leaves the CPU.
 
I went with the F4-3600C16D-32GTZNC rather than the patriot because it was on sale and had the "made for ryzen" thing going on. It is in the QVL list. I have tried two different motherboards, two different sets of memory, and I still have trouble running ram at XMP speeds. That only leaves the CPU.
Good thing I never bothered with XMP after a few tries of it not working. Manual settings are flawless though.
 
Back
Top