AMD Ryzen 9 3900X Review & Benchmarks: Premiere, Blender, Gaming, & More
video description
Date: 2020-05-06
Related videos
Comments and reviews: 10
William
Would be nice for all these tech reviewers to test latency. For people who work in areas that demand low latency, we'd all like to know. Anyone who does music production for example. Although I'm sure stock trading is out of the equation, that's another area where latency matters. There's a reason that games are better on Intel. And, yep, you guessed it. Intel has lower latency than AMD (clock for clock even. So, as someone who uses multiple Kontackt, Diva, Serum patches, I look at these game guides and see that the i9 9900k is still beating the Ryzen, despite the Ryzen having more Cores, then I can only speculate that latency is still higher on the Ryzen chips. Frequency does not matter so much for DAWs and music production, certainly not as much as latency, for live recordings. However, for post production and exporting, then having more cores is better, as at that point you can simply whack up the audio buffer to a level that would be unusable if you tried to play live. It was this that caused me to purchase the i9, despite me prefering the Ryzen 7 at the time. The Ryzen 9 (3rd Gen) would be fantastic, especially the promised 16 core version. But, I do like to play live, with all my patches and tracks 'as is' (ie, not actually rendered. And, unfortunately, as is the case with gaming, I would be at a disadvantage if I went with AMD. However, this latency issue could be addressed with BIOS updates, and the upcoming better scheduler from Microsoft. It doesn't have to beat Intel. It just has to come close enough not to be an issue. The old Ryzen 7 (2nd Gen, would have caused me to lower my audio buffer two whole stops. So I'm very curious to where the new Ryzen lies. If I can run at 128 samples on the audio buffer, I'd be happy. Happy enough to sell on my i9 9900k, that's for sure. If I have to run at 256 audio samples, then, that would be a tough call, but still certainly viable for the additional cores (as I could have more instances, a lot more. But If I'm having to run at 512 samples, then I'm out. So yeah, it would be nice if tech reviewers used the DAW bench analysis in their productivity suites. You'd get a LOT more views, and it would show the differentials between the two platforms as well.
reply
Would be nice for all these tech reviewers to test latency. For people who work in areas that demand low latency, we'd all like to know. Anyone who does music production for example. Although I'm sure stock trading is out of the equation, that's another area where latency matters. There's a reason that games are better on Intel. And, yep, you guessed it. Intel has lower latency than AMD (clock for clock even. So, as someone who uses multiple Kontackt, Diva, Serum patches, I look at these game guides and see that the i9 9900k is still beating the Ryzen, despite the Ryzen having more Cores, then I can only speculate that latency is still higher on the Ryzen chips. Frequency does not matter so much for DAWs and music production, certainly not as much as latency, for live recordings. However, for post production and exporting, then having more cores is better, as at that point you can simply whack up the audio buffer to a level that would be unusable if you tried to play live. It was this that caused me to purchase the i9, despite me prefering the Ryzen 7 at the time. The Ryzen 9 (3rd Gen) would be fantastic, especially the promised 16 core version. But, I do like to play live, with all my patches and tracks 'as is' (ie, not actually rendered. And, unfortunately, as is the case with gaming, I would be at a disadvantage if I went with AMD. However, this latency issue could be addressed with BIOS updates, and the upcoming better scheduler from Microsoft. It doesn't have to beat Intel. It just has to come close enough not to be an issue. The old Ryzen 7 (2nd Gen, would have caused me to lower my audio buffer two whole stops. So I'm very curious to where the new Ryzen lies. If I can run at 128 samples on the audio buffer, I'd be happy. Happy enough to sell on my i9 9900k, that's for sure. If I have to run at 256 audio samples, then, that would be a tough call, but still certainly viable for the additional cores (as I could have more instances, a lot more. But If I'm having to run at 512 samples, then I'm out. So yeah, it would be nice if tech reviewers used the DAW bench analysis in their productivity suites. You'd get a LOT more views, and it would show the differentials between the two platforms as well.
reply
Tech
I think is important to mention that the tsmc 7nm silicon is kind of limited by the imposed amd locks so that i stays very efficient and because of that when one of the 3 locks is hit even if the other 2 are not, youre cpu will stop increasing the frequency and you hit that frequency wall. Here are the silicon frequency locks described on the anandtech article: Package Power Tracking (PPT): The power threshold that is allowed to be delivered to the socket. This is 88W for 65W TDP processors, and 142W for 105W TDP processors. Thermal Design Current (TDC): The maximum amount of current delivered by the motherboard s voltage regulators when under thermally constrained scenarios (high temperatures) This is 60A for 65W TDP processors, and 95A for 105W TDP processors. Electrical Design Current (EDC): This is the maximum amount of current at any instantaneous short period of time that can be delivered by the motherboard s voltage regulators. This is 90A for 65W TDP processors, and 140A for 105W TDP processors. I think Intel has a way wider margin on those, and i think is one of the reasons why those intel cpus clock much higher but also draw much more juice, amd wanted those to be efficient, the vrms on x570 are way overdone for those power draw limits, also in terms of cooling wouldnt be a problem for some high end noctua cooler or good aio, but it doesnt matter since your locked by amd and that pbo it think doesnt really cross the power boundaries, or if it does is very limited %
reply
I think is important to mention that the tsmc 7nm silicon is kind of limited by the imposed amd locks so that i stays very efficient and because of that when one of the 3 locks is hit even if the other 2 are not, youre cpu will stop increasing the frequency and you hit that frequency wall. Here are the silicon frequency locks described on the anandtech article: Package Power Tracking (PPT): The power threshold that is allowed to be delivered to the socket. This is 88W for 65W TDP processors, and 142W for 105W TDP processors. Thermal Design Current (TDC): The maximum amount of current delivered by the motherboard s voltage regulators when under thermally constrained scenarios (high temperatures) This is 60A for 65W TDP processors, and 95A for 105W TDP processors. Electrical Design Current (EDC): This is the maximum amount of current at any instantaneous short period of time that can be delivered by the motherboard s voltage regulators. This is 90A for 65W TDP processors, and 140A for 105W TDP processors. I think Intel has a way wider margin on those, and i think is one of the reasons why those intel cpus clock much higher but also draw much more juice, amd wanted those to be efficient, the vrms on x570 are way overdone for those power draw limits, also in terms of cooling wouldnt be a problem for some high end noctua cooler or good aio, but it doesnt matter since your locked by amd and that pbo it think doesnt really cross the power boundaries, or if it does is very limited %
reply
stevae
It's really strange to listen to your attempt at making gaming sound more important than productivity. I maintain that anyone who builds a computer system ONLY to game is freaking ignorant! There is more to life than just gaming. And it's a WASTE of money and a good device. While gaming is fun, and a cool way to pass some time, it's NOT important in LIFE! There are thousands of things you can use your computer for, and gaming is only one small part of that. It's funny how you say the 3900X isn't great at everything, but continuously have praised the 9900k for being great for gaming. Well, considering how the 3900X beats the crap out of the 9900k in just about EVERYTHING else besides a few game titles, I think that the 9900k not doing hardly anything great, other than gaming, is a stronger statement about that chip. In a total comparison, from what I have seen over hundreds of videos, the 3900X beats the 9900k in Almost ALL other applications, outside of gaming. And even in gaming, it's close enough on most titles that you can't tell the difference on the actual screen. You ONLY can tell the 9900k is a little higher in fps by numbers in a benchmark. I'll take the 3900X every day of the week over the 9900k, and twice on Sunday. Lastly, the 9900k is well dated, at the end of its life, and the 3900X is just starting. The choice is clear to me.
reply
It's really strange to listen to your attempt at making gaming sound more important than productivity. I maintain that anyone who builds a computer system ONLY to game is freaking ignorant! There is more to life than just gaming. And it's a WASTE of money and a good device. While gaming is fun, and a cool way to pass some time, it's NOT important in LIFE! There are thousands of things you can use your computer for, and gaming is only one small part of that. It's funny how you say the 3900X isn't great at everything, but continuously have praised the 9900k for being great for gaming. Well, considering how the 3900X beats the crap out of the 9900k in just about EVERYTHING else besides a few game titles, I think that the 9900k not doing hardly anything great, other than gaming, is a stronger statement about that chip. In a total comparison, from what I have seen over hundreds of videos, the 3900X beats the 9900k in Almost ALL other applications, outside of gaming. And even in gaming, it's close enough on most titles that you can't tell the difference on the actual screen. You ONLY can tell the 9900k is a little higher in fps by numbers in a benchmark. I'll take the 3900X every day of the week over the 9900k, and twice on Sunday. Lastly, the 9900k is well dated, at the end of its life, and the 3900X is just starting. The choice is clear to me.
reply
Will4May
I am an AMD fan, they are consumer freindly with opensource, freesync, pricing and able to use the same motherboard over three generations. However the things that are hurting AMD is Raw clock speed, but you can forgive them that, since Intels cpus are on old optimized to death architecture and Ryzen is new, AMD has had 5. 0ghz cpus just not very well designed. On the 3900x at least having to turn off SMT to get more speed for some games isnt ideal. And AMD themselves, they are not really known for being nasty to others like Intel and Nvidia, but tend to hurt themselves instead, TSMC struggling right now hasnt helped tbh, right when AMD really need them at full speed. Either way I am still going to continue to support AMD, and without them PC part pricing would be insane, it is still expensive but at least achievable for a lower class worker (eventually.
reply
I am an AMD fan, they are consumer freindly with opensource, freesync, pricing and able to use the same motherboard over three generations. However the things that are hurting AMD is Raw clock speed, but you can forgive them that, since Intels cpus are on old optimized to death architecture and Ryzen is new, AMD has had 5. 0ghz cpus just not very well designed. On the 3900x at least having to turn off SMT to get more speed for some games isnt ideal. And AMD themselves, they are not really known for being nasty to others like Intel and Nvidia, but tend to hurt themselves instead, TSMC struggling right now hasnt helped tbh, right when AMD really need them at full speed. Either way I am still going to continue to support AMD, and without them PC part pricing would be insane, it is still expensive but at least achievable for a lower class worker (eventually.
reply
bnccristi
Question about the Premiere Pro Export time. As we know the Premiere Pro is able to use the Integrated Graphics Card UHD 530, UHD 630 inside the Intel Chip starting with the 6 series on chips. Myself I own a I7 8086k Anniversary edition which is a basically a I7 8700k. When I export an 4K project on Premiere Pro H. 265, Hardware Encoding (not software. the system is using a the integrated chip about 90-100%, CPU about 60-80%, Nvidia 1080 ti about 35%, that is reducing the export time comparing to software encoding or with only the dedicated card. The question is: Can you compare the export time on a I7 or I9 with the integrated graphics card enabled Vs AMD Ryzen 9 3900x on Hardware Encoding, not software encoding? I will really appreciate, i couldn't find any information or comparison
reply
Question about the Premiere Pro Export time. As we know the Premiere Pro is able to use the Integrated Graphics Card UHD 530, UHD 630 inside the Intel Chip starting with the 6 series on chips. Myself I own a I7 8086k Anniversary edition which is a basically a I7 8700k. When I export an 4K project on Premiere Pro H. 265, Hardware Encoding (not software. the system is using a the integrated chip about 90-100%, CPU about 60-80%, Nvidia 1080 ti about 35%, that is reducing the export time comparing to software encoding or with only the dedicated card. The question is: Can you compare the export time on a I7 or I9 with the integrated graphics card enabled Vs AMD Ryzen 9 3900x on Hardware Encoding, not software encoding? I will really appreciate, i couldn't find any information or comparison
reply
Alper
what is next 320p lowest test? low resource testing is an unproven myth. you are effectively handicapping L3 cache advantage of new ryzen, in reality it does have real performance effects. and stop using single core test of multicore applications, because i have yet to see anyone using blender-cinebech etc. with single core. If you have single core applications natively/naturally use that instead. you will have 9900K or 3900X and 2080ti and use it 1080p? no but we are doing that to draw cpus future performance? well that doesnt work well for 7700K owners now does it. I will give you a hint for GCC, use a single file compile 100000 times and use a single core, this setting will hide intels weakness if not give huge advantage. But in reality no one ever does that.
reply
what is next 320p lowest test? low resource testing is an unproven myth. you are effectively handicapping L3 cache advantage of new ryzen, in reality it does have real performance effects. and stop using single core test of multicore applications, because i have yet to see anyone using blender-cinebech etc. with single core. If you have single core applications natively/naturally use that instead. you will have 9900K or 3900X and 2080ti and use it 1080p? no but we are doing that to draw cpus future performance? well that doesnt work well for 7700K owners now does it. I will give you a hint for GCC, use a single file compile 100000 times and use a single core, this setting will hide intels weakness if not give huge advantage. But in reality no one ever does that.
reply
theexmann
Really? You recommend the Intel chip over the AMD processor for gaming? What was the average and median percent differences between the best AMD gaming FPS scores and the best Intel gaming FPS scores? I roughly figure it's around 5% or less? A human being would not be able to perceive that difference. Also, didn't you also say that the game play was overall more smooth on the AMD chip? I would say that would be more important than a 5% difference in FPS. Throw in PCIe4 as a platform bonus, and I would think the AMD would be the better chip on a better eco-system going forward. Also, AMD recommend 3600 memory for best results. I wonder if that 5% FPS gaming difference disappear. I think other reviewers using 3600 memory might point to that.
reply
Really? You recommend the Intel chip over the AMD processor for gaming? What was the average and median percent differences between the best AMD gaming FPS scores and the best Intel gaming FPS scores? I roughly figure it's around 5% or less? A human being would not be able to perceive that difference. Also, didn't you also say that the game play was overall more smooth on the AMD chip? I would say that would be more important than a 5% difference in FPS. Throw in PCIe4 as a platform bonus, and I would think the AMD would be the better chip on a better eco-system going forward. Also, AMD recommend 3600 memory for best results. I wonder if that 5% FPS gaming difference disappear. I think other reviewers using 3600 memory might point to that.
reply
Steve
Try compiling Chromium, it's painfully slow and it has my CPU hit 100% load for like an hour. [It might be I/O affected as well though, but who knows] Also, PLEASE add Tensorflow benchmarks. MKL-DNN uses AVX2, so you can push the CPU load there a lot if you choose to compile Tensorflow with that. From what I've seen in benchmarks, Tensorflow really emphasizes memory performance and core count. There is a MASSIVE performance difference between 2666mhz memory and 3600mhz memory with Ryzen 3000 CPUs under Tensorflow. Happy to help out here if needed.
reply
Try compiling Chromium, it's painfully slow and it has my CPU hit 100% load for like an hour. [It might be I/O affected as well though, but who knows] Also, PLEASE add Tensorflow benchmarks. MKL-DNN uses AVX2, so you can push the CPU load there a lot if you choose to compile Tensorflow with that. From what I've seen in benchmarks, Tensorflow really emphasizes memory performance and core count. There is a MASSIVE performance difference between 2666mhz memory and 3600mhz memory with Ryzen 3000 CPUs under Tensorflow. Happy to help out here if needed.
reply
Thanny
You cannot measure CPU power consumption by measuring the EPS 12V lines. First, you cannot assume that all that power goes to the CPU itself. Second, more importantly, that current does not go to the CPU directly. It goes through the VRM's, which are far from perfectly efficient, and lose a substantial amount of their input power as waste heat. It is easily possible that the figure you showed, when accounting for non-CPU usage and efficiency losses, shows less than 105W consumption by the CPU, putting it within its thermal design power.
reply
You cannot measure CPU power consumption by measuring the EPS 12V lines. First, you cannot assume that all that power goes to the CPU itself. Second, more importantly, that current does not go to the CPU directly. It goes through the VRM's, which are far from perfectly efficient, and lose a substantial amount of their input power as waste heat. It is easily possible that the figure you showed, when accounting for non-CPU usage and efficiency losses, shows less than 105W consumption by the CPU, putting it within its thermal design power.
reply
Nick
Pretty much cherrypicking for AMD real work scenarios, pretty much cherrypicking for intel the game side. AMD is way more ahead in productivity benches than intel in game benches. Why you throw in a ton of benchmarks of an comletly irrelevant 2 year old game (TW Warhammer2. Nobody cares for this old shit. There are some games, amd is beating intel. And on most games the gap is not that big. (around 5% margin) Why you dont test the 10 most popular games last 18 months?
reply
Pretty much cherrypicking for AMD real work scenarios, pretty much cherrypicking for intel the game side. AMD is way more ahead in productivity benches than intel in game benches. Why you throw in a ton of benchmarks of an comletly irrelevant 2 year old game (TW Warhammer2. Nobody cares for this old shit. There are some games, amd is beating intel. And on most games the gap is not that big. (around 5% margin) Why you dont test the 10 most popular games last 18 months?
reply
Add a review, comment
Other channel videos
