RIP CPU + Motherboard: R9 3900X & 5700 XT Overclock Stream Recap
video description
Date: 2020-05-06
Related videos
Comments and reviews: 10
xyz360400
It seems like NVIDIA restricts power/OC to a certain extent to stay within TDP/overall power draw numbers to make their efficiency look good (especially since they're almost always last to a new node so AMD has the advantage of smaller silicon) but the upside is that you end up with more OC headroom, whereas AMD, usually on a new node, pushes their silicon just about as far as it can go without frying itself just to try and compete reasonably with NVIDIA which leaves virtually no headroom in their silicon for overclocking. It seems similar with Intel vs AMD for CPUs where Intel (at least used to; the latest Coffee Lake refresh seems to be an exception given that they're using an 8 core chip to try and compete with 12 core and eventually 16 core CPUs) leaves some wiggle room with regards to overclocking clock speeds and voltages, where AMD tends to have their chips binned and rated right at or maybe JUST below the max for their chips under normal conditions (i. e. not insane voltages under LN2/extreme cooling. I hope that you guys are able to successfully power-mod Navi, but unfortunately I suspect you won't see much gain out of it (at least with any amount of stability) and I foresee a lot of artifacts and crashes in your future, but I hope I'm wrong. If AMD binned their GPUs the way that NVIDIA does, where their performance level is right around the top of their efficiency curve rather than being pushed to the max for optimum performance, power and thermals be damned, then they would look a lot better for laptops and other lower power/thermally limited scenarios, though it would hurt their bottom line since they'd have to sell them as lower-tier graphics cards which is likely why they don't do it.
reply
It seems like NVIDIA restricts power/OC to a certain extent to stay within TDP/overall power draw numbers to make their efficiency look good (especially since they're almost always last to a new node so AMD has the advantage of smaller silicon) but the upside is that you end up with more OC headroom, whereas AMD, usually on a new node, pushes their silicon just about as far as it can go without frying itself just to try and compete reasonably with NVIDIA which leaves virtually no headroom in their silicon for overclocking. It seems similar with Intel vs AMD for CPUs where Intel (at least used to; the latest Coffee Lake refresh seems to be an exception given that they're using an 8 core chip to try and compete with 12 core and eventually 16 core CPUs) leaves some wiggle room with regards to overclocking clock speeds and voltages, where AMD tends to have their chips binned and rated right at or maybe JUST below the max for their chips under normal conditions (i. e. not insane voltages under LN2/extreme cooling. I hope that you guys are able to successfully power-mod Navi, but unfortunately I suspect you won't see much gain out of it (at least with any amount of stability) and I foresee a lot of artifacts and crashes in your future, but I hope I'm wrong. If AMD binned their GPUs the way that NVIDIA does, where their performance level is right around the top of their efficiency curve rather than being pushed to the max for optimum performance, power and thermals be damned, then they would look a lot better for laptops and other lower power/thermally limited scenarios, though it would hurt their bottom line since they'd have to sell them as lower-tier graphics cards which is likely why they don't do it.
reply
John
Heat/Cold Stress: Yes, Slam Bake ovens with LN2 assisted cooling are used to STRESS parts. In fact due to fact that shoving a frozen part in a hot oven drops the ovens temperature, it makes sense to overheat the oven above the desired soaking temperature while the Unit Under Test is in the Freezer. Then at the instant the Frozen part is shoved into the oven the oven's setpoint is adjusted to the desired soaking temperature. That way you get sharper edges on the temps in the STRESS test. The opposite happens to the freezer when the part is in the oven. Over cool the freezer and shove a hot part in it while resetting the freezer temperature. That way you get MAXIMUM STRESS and can literally bust a part open. Love it. Engineers didn't like that much stress because too many parts would fail. They preferred the method of heating the freezer to oven temps and back. sort of like hauling an Alaskan Iceberg to South America. it melts before you get there. Least stress, cheating.
reply
Heat/Cold Stress: Yes, Slam Bake ovens with LN2 assisted cooling are used to STRESS parts. In fact due to fact that shoving a frozen part in a hot oven drops the ovens temperature, it makes sense to overheat the oven above the desired soaking temperature while the Unit Under Test is in the Freezer. Then at the instant the Frozen part is shoved into the oven the oven's setpoint is adjusted to the desired soaking temperature. That way you get sharper edges on the temps in the STRESS test. The opposite happens to the freezer when the part is in the oven. Over cool the freezer and shove a hot part in it while resetting the freezer temperature. That way you get MAXIMUM STRESS and can literally bust a part open. Love it. Engineers didn't like that much stress because too many parts would fail. They preferred the method of heating the freezer to oven temps and back. sort of like hauling an Alaskan Iceberg to South America. it melts before you get there. Least stress, cheating.
reply
Joshua
Totally agree driver updates can break overclocking options. I remember back when the GTX 500 series came out, the driver which the GTX 400 series shared, broke shader overclocking separate from the memory clocks, which generally gave the GTX 400 series a competitive edge on the GTX 500 series and deterred people from initially upgrading to the 500 series. After that, more people adopted the 500 series upgrade, because alot of the games and benchmarks upgraded to not work on older drivers. Was a real shady move by Nvidia to force people to upgrade.
reply
Totally agree driver updates can break overclocking options. I remember back when the GTX 500 series came out, the driver which the GTX 400 series shared, broke shader overclocking separate from the memory clocks, which generally gave the GTX 400 series a competitive edge on the GTX 500 series and deterred people from initially upgrading to the 500 series. After that, more people adopted the 500 series upgrade, because alot of the games and benchmarks upgraded to not work on older drivers. Was a real shady move by Nvidia to force people to upgrade.
reply
something1random23
One of the editors of PC Gamer was assuring me that memory overclocking is too hard, too unsafe, and offered virtually no benefit above 2666 MHz so it's a waste of time and money. At the same time he talked of how much more easy an overclock of the 8700K was and how it was up to 20% faster than Ryzen 3000 when overclocked in his testing (gaming. And sure it's faster in a lot of games, but I wasn't going to settle for that explanation. Thanks for proving the point on the Fclk and memory speeds repeatedly.
reply
One of the editors of PC Gamer was assuring me that memory overclocking is too hard, too unsafe, and offered virtually no benefit above 2666 MHz so it's a waste of time and money. At the same time he talked of how much more easy an overclock of the 8700K was and how it was up to 20% faster than Ryzen 3000 when overclocked in his testing (gaming. And sure it's faster in a lot of games, but I wasn't going to settle for that explanation. Thanks for proving the point on the Fclk and memory speeds repeatedly.
reply
Happy
Really wish AMD would have the pins on the socket, the motherboard is always cheaper than the CPU (unless you're doing something silly and running a expensive board with a i3/R3, also with the pins on the board you can have a socket that keeps the CPU locked in so when taking the heatsink off you don't have to worry about the CPU pulling out the socket. Really happy with my 3700x however that is all I'd like to see changed with the DDR5/AM5.
reply
Really wish AMD would have the pins on the socket, the motherboard is always cheaper than the CPU (unless you're doing something silly and running a expensive board with a i3/R3, also with the pins on the board you can have a socket that keeps the CPU locked in so when taking the heatsink off you don't have to worry about the CPU pulling out the socket. Really happy with my 3700x however that is all I'd like to see changed with the DDR5/AM5.
reply
TEAMHYBRID007
And my experience bent or broken pins are not a stopping point I have used pens removed via razor blade from useless CPUs and stuck them in the slot on the grid array where they belong and put the CPU on top I've also straightened CPU pins using a razor blade and or knife with great success and in some cases I've even used Staples cut to the correct length and inserted into the motherboard
reply
And my experience bent or broken pins are not a stopping point I have used pens removed via razor blade from useless CPUs and stuck them in the slot on the grid array where they belong and put the CPU on top I've also straightened CPU pins using a razor blade and or knife with great success and in some cases I've even used Staples cut to the correct length and inserted into the motherboard
reply
johnny5finger
I dont understand overclocking, you are risking your components to get an extra 30 fps. Who cares if you get 140 OC compared to 110 fps stock. If your looking for longevity of your components run them stock. 1. 6 volt is crazy overclocking. There is a reason AMD didn't ship the 3900x with a boost clock of 5. 2
reply
I dont understand overclocking, you are risking your components to get an extra 30 fps. Who cares if you get 140 OC compared to 110 fps stock. If your looking for longevity of your components run them stock. 1. 6 volt is crazy overclocking. There is a reason AMD didn't ship the 3900x with a boost clock of 5. 2
reply
WhyteDem0n
I wish my 3900X could hit 1900 on the IF. I can only get 1800 to really run stable. I'm not changing any voltages though because I'm not entirely sure what I'm doing. :-\ FWIW, I'm using an Aorus x570 Master, so I probably could get it to work if I knew what I was doing XD
reply
I wish my 3900X could hit 1900 on the IF. I can only get 1800 to really run stable. I'm not changing any voltages though because I'm not entirely sure what I'm doing. :-\ FWIW, I'm using an Aorus x570 Master, so I probably could get it to work if I knew what I was doing XD
reply
Joshua
Please make a metal band and call it Silicon Overcharge. Use overkill hardware to power overclocked guitars. Drink caffeine before every performance to overclock yourself. And whirlwind that metal AF hair until people mosh to your technologically dominant masculinity. \m/
reply
Please make a metal band and call it Silicon Overcharge. Use overkill hardware to power overclocked guitars. Drink caffeine before every performance to overclock yourself. And whirlwind that metal AF hair until people mosh to your technologically dominant masculinity. \m/
reply
SerpentXSF
Steve I am able to get 945 Mhz on memory, but depending on test, benchmark or Algo it may not be stable. The Non XT wont let us control memory voltage at all. even with More Power tool, unless there is a private version not released to public. Thanks for stream
reply
Steve I am able to get 945 Mhz on memory, but depending on test, benchmark or Algo it may not be stable. The Non XT wont let us control memory voltage at all. even with More Power tool, unless there is a private version not released to public. Thanks for stream
reply
Add a review, comment
Other channel videos
