- Joined
- Aug 16, 2009
From the link ......
I take this to mean just that ..... I just can't for the life of me understand why they would offset the temp like that.
Last edited by a moderator:
-
Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!
FEATURED AMD ZEN Discussion (Previous Rumor Thread)
- Thread starter Johan45
- Start date
From the link ......
I take this to mean just that ..... I just can't for the life of me understand why they would offset the temp like that.
- Joined
- Dec 18, 2000
So thanks to this new info, my 95C in Prime95 is only 75C which makes me feel a lot better. I couldn't understand why my 1700X was so much hotter than my two 1700s and now I know it isn't. Odd that AMD wouldn't simply subtract the 20C from the Ryzen Master temp display so users would get a correct reading.
- Joined
- Jan 1, 2008
- Location
- Colorado Springs, CO
Honestly, it would help tcase stay under 75C to ensure it boosts with XFR as much as possible. Atleast with the fan thing, it's the only logical explanation I can come up with.
- Joined
- Jun 28, 2012
This is exactly what they're doing, boosting the fan profile.
- Joined
- Jan 1, 2008
- Location
- Colorado Springs, CO
Yup, but bass was asking why, the only explanation is to bump the turbo/XFR time as temp is really lower than the Mobo/fan profile is really thinking.
OP
- Joined
- Dec 19, 2012
- Thread Starter
- #2,908
I don't think the temp reading issues all lie with the the way AMD implements things either. I got my replacement CHVI yesterday and flashed the new BIOS to it. The temps were way off when I started to OC. What I mean is that at stock I was getting mid 40s in idle. After bumping the multi up 200 and setting manual 1.2v I was getting idle in the 30s. Raja recommended a Sense MI skew of 272 which is their default value but apparently when I started to OC it must change that setting when left on auto. I set the skew and then temps "seemed" a bit more normal but I still felt they were low. I swapped to my loop (ambient) and things really didn't change a whole lot dropped maybe 5c. In the end I was running Small FFT at 4.0 , 1.4v and a temp of 70c
Reading all of the input about the temperature vs OC and voltage not doing much in the sense of mitigating, leads me to believe that the Ryzen is a very hungry current caterpillar. Some of you have mentioned that you see your VRM FETs continue to heat up as time goes on, and no additional cooling seems to help prevent it. Sounds like thermal runaway due to high current draw. This can happen when the FET doesn't "close" or "open" all the way allowing for all the current to seep through. Slowly the channel in the gate will relax allowing for current leakage. Current continues to leek into the CPU and into the board, which will ultimately heat up the entire board.
Anyone have a thermal camera? I bet you will see your entire board around the socket slowly heat up while running tests. To say the least that this is normal is true, but with the right cooling a board should never become overheated. However, I believe something else is happening and it looks like you will need a higher amount of cooling to keep these systems happy.
Anyone have a thermal camera? I bet you will see your entire board around the socket slowly heat up while running tests. To say the least that this is normal is true, but with the right cooling a board should never become overheated. However, I believe something else is happening and it looks like you will need a higher amount of cooling to keep these systems happy.
And that proves the point I was about to edit into my post.
I looked back at some of the posts here, and it seems the issues I mentioned are based off lower end boards of which will utilize lower quality parts. They will definitely be limited by how high you can push the voltage. This is true with just about all systems (Intel or AMD) but I think this will be seen a bit unusual because of the uniqueness of the AMD chips.
OP
- Joined
- Dec 19, 2012
- Thread Starter
- #2,912
So what you're saying is go big or stay home just like FX. If you want the highend OC you'll need a capable/quality board.
You got it Johan, just like the 9590, these chips seem to love current. Once I get my board I'll be able to tie the VCC and VDD rails into a current monitor at work. There I'll have nice scope shots of current spikes and the like for our Ryzen chips.
I just want my board back form RMA
I just want my board back form RMA
OP
- Joined
- Dec 19, 2012
- Thread Starter
- #2,914
You got it Johan, just like the 9590, these chips seem to love current. Once I get my board I'll be able to tie the VCC and VDD rails into a current monitor at work. There I'll have nice scope shots of current spikes and the like for our Ryzen chips.
I just want my board back form RMA
New BIOS seems OK still some quirks as I mentioned. Could have been the other BIOS but even BCLK and RAM adjustments were a lot easier this time. Didn't have one failure during boot. Started at 2400 strap and walked it up to get better subs
- Joined
- Jan 1, 2008
- Location
- Colorado Springs, CO
- Joined
- Dec 18, 2000
- Joined
- Jun 28, 2012
Here's a link instead of a screenshot so people can actually see the whole conversation...
http://www.overclockers.com/forums/showthread.php?p=7998105
http://www.overclockers.com/forums/showthread.php?p=7998105
Well there you have it. 75C after 5mins of stress. That's hot! I'm not surprised that some of the OC can only last for like 30mins under full load. Its not the CPU that is causing issues, its the mobo.
I'd say, keep those VRMs below 70C under 24/7 stress test, and you should see a bit more wiggle room. Really cool them, and you may get more but it will tapper off quickly.
I think the key to RyZen will be to keep the entire system as cold as possible rather than just the CPU like it has been in the past. Again no first hand experience just yet.
For those that are curious:
https://www.onsemi.com/pub/Collateral/AND8199-D.PDF
This paper by On Semi Conductors goes into a bit about what thermals looks like in terms of current output based on voltage input to the FET. Terms to be aware of:
ID = Current Drain from FET. This is the current that would be pushed into the CPU
Vgs = Voltage from Gate to Source. This is the voltage ratio difference between the FET gate to its Source. For MOSFETs you want the Voltage on the Drain Pin to be greater than the Vgs ratio to allow full saturation. When this number is too low, the FET will source lower amounts of current to the Drain pin. *NOTE* that the Vgs in the paper will be much much higher than what is on these mobos, this is just an example.
I'd say, keep those VRMs below 70C under 24/7 stress test, and you should see a bit more wiggle room. Really cool them, and you may get more but it will tapper off quickly.
I think the key to RyZen will be to keep the entire system as cold as possible rather than just the CPU like it has been in the past. Again no first hand experience just yet.
For those that are curious:
https://www.onsemi.com/pub/Collateral/AND8199-D.PDF
This paper by On Semi Conductors goes into a bit about what thermals looks like in terms of current output based on voltage input to the FET. Terms to be aware of:
ID = Current Drain from FET. This is the current that would be pushed into the CPU
Vgs = Voltage from Gate to Source. This is the voltage ratio difference between the FET gate to its Source. For MOSFETs you want the Voltage on the Drain Pin to be greater than the Vgs ratio to allow full saturation. When this number is too low, the FET will source lower amounts of current to the Drain pin. *NOTE* that the Vgs in the paper will be much much higher than what is on these mobos, this is just an example.
Last edited:
- Joined
- Jan 1, 2008
- Location
- Colorado Springs, CO
That board is the one without vrm sinks. I think it was the Asus B350-m (matx). Might not be as big a deal on some of the other B350 boards that have a passive sink on their VRMs.
Edit: not saying you are wrong by any means, just that I don't think that is a fair specimen. I'd be surprised if 8 cores didn't stress the VRM section of a Mobo.
Similar threads
