AMD Ryzen 9 9900X CPU Review & Benchmarks vs. 14700K, 7900X, 9950X, & More
video description
Date: 2024-08-17
Related videos
Comments and reviews: 20
wolpumba4099
Summary
0:00 - AMD Ryzen 9 9900X Review: This video reviews the AMD Ryzen 9 9900X, a 12-core, 24-thread CPU priced at $500.
2:02 - Performance: The 9900X shows a performance improvement of 0-10% compared to the previous generation 7900X. In some production applications, this improvement can be as high as 9-12%.
8:27 - Efficiency: While efficiency can be better in some cases (like Blender), it can also be slightly worse than the previous generation in other workloads, depending on factors like whether all cores are fully utilized.
14:30 - Gaming: The 9900X is capable of gaming, but is not designed to be a top performer. AMD's 7800X3D and Intel's 14700K (pending Intel resolving ongoing issues) offer better gaming performance.
27:35 - Value Proposition: Currently, Gamers Nexus recommends the previous generation Zen 4 CPUs, like the 7950X and 7800X3D, over the 9900X for better value. This is based on the price-to-performance ratio and efficiency.
28:20 - Intel Alternatives: The Intel i7-14700K is competitive with the 9900X, but Gamers Nexus has temporarily paused recommending any 13th or 14th Gen Intel CPUs due to concerns about microcode updates and support processes.
Key Takeaways:
The 9900X offers modest performance gains over the previous generation.
Efficiency gains are workload-dependent.
It is not the best choice for a pure gaming build.
Zen 4 and potentially the 14700K offer better value currently.
28:38 - Recommendation: Consider Zen 4 CPUs (like the 7950X for heavily threaded tasks or the 7800X3D for gaming) as better alternatives to the 9900X right now. Keep an eye on the Intel 14700K as a potential competitor if Intel resolves its current issues.
Summarized by AI model: gemini-1.5-pro-exp-0801
Cost (if I didn't use the free tier): $0.1228
Input tokens: 31985
Output tokens: 1034
reply
Summary
0:00 - AMD Ryzen 9 9900X Review: This video reviews the AMD Ryzen 9 9900X, a 12-core, 24-thread CPU priced at $500.
2:02 - Performance: The 9900X shows a performance improvement of 0-10% compared to the previous generation 7900X. In some production applications, this improvement can be as high as 9-12%.
8:27 - Efficiency: While efficiency can be better in some cases (like Blender), it can also be slightly worse than the previous generation in other workloads, depending on factors like whether all cores are fully utilized.
14:30 - Gaming: The 9900X is capable of gaming, but is not designed to be a top performer. AMD's 7800X3D and Intel's 14700K (pending Intel resolving ongoing issues) offer better gaming performance.
27:35 - Value Proposition: Currently, Gamers Nexus recommends the previous generation Zen 4 CPUs, like the 7950X and 7800X3D, over the 9900X for better value. This is based on the price-to-performance ratio and efficiency.
28:20 - Intel Alternatives: The Intel i7-14700K is competitive with the 9900X, but Gamers Nexus has temporarily paused recommending any 13th or 14th Gen Intel CPUs due to concerns about microcode updates and support processes.
Key Takeaways:
The 9900X offers modest performance gains over the previous generation.
Efficiency gains are workload-dependent.
It is not the best choice for a pure gaming build.
Zen 4 and potentially the 14700K offer better value currently.
28:38 - Recommendation: Consider Zen 4 CPUs (like the 7950X for heavily threaded tasks or the 7800X3D for gaming) as better alternatives to the 9900X right now. Keep an eye on the Intel 14700K as a potential competitor if Intel resolves its current issues.
Summarized by AI model: gemini-1.5-pro-exp-0801
Cost (if I didn't use the free tier): $0.1228
Input tokens: 31985
Output tokens: 1034
reply
D3lor34n
Just my two cents since you scraped the topic in the intro. I'm absolutely fine with older generations dropping out of charts and all that, but the biggest struggle for me is always finding equivalents. First ten seconds of a chart are me figuring out which components are on the screen, how they vaguely relate to my setup and then extrapolating how it related to the reviewed subject. I'm not asking for a superchart with 500 entries, but i wonder if some kind of translation table, maybe once a genration or year is a feasable goal. Maybe a website link in the footnote
I don't have a solution, just saying even from RTX 2000 it's sometimes a challenge with one 2080 data point if you're lucky. You can leap through generations and land in 4060 land but with DLSS generations even these numbers are sometimes varied. Same goes for older but not ancient CPUs. I know, at some point it's more or less all the same, doesn't matter if new generation is 110 or 115% faster than my old setup, but what i'm struggling with (a gripe i have for decades of pc building now) is a definite data point for reference after a 1-2 year cliff. Yes i'd like to know how a 4090 performs on a 9900k vs 14900k for example, just so i can gauge the impact of upgrades. I know there are older reference points in every review but it's still mental gymnastics more times than not when you're a little bit obscessed with exact comparison to your current setup.
I think you get what i mean, just a very occasional link to the past to better comprehend the actual value of an upgrade. Keep up your great work, GN is not even a national treasure since you shine around the world but a beacon of light in this industry.
reply
Just my two cents since you scraped the topic in the intro. I'm absolutely fine with older generations dropping out of charts and all that, but the biggest struggle for me is always finding equivalents. First ten seconds of a chart are me figuring out which components are on the screen, how they vaguely relate to my setup and then extrapolating how it related to the reviewed subject. I'm not asking for a superchart with 500 entries, but i wonder if some kind of translation table, maybe once a genration or year is a feasable goal. Maybe a website link in the footnote
I don't have a solution, just saying even from RTX 2000 it's sometimes a challenge with one 2080 data point if you're lucky. You can leap through generations and land in 4060 land but with DLSS generations even these numbers are sometimes varied. Same goes for older but not ancient CPUs. I know, at some point it's more or less all the same, doesn't matter if new generation is 110 or 115% faster than my old setup, but what i'm struggling with (a gripe i have for decades of pc building now) is a definite data point for reference after a 1-2 year cliff. Yes i'd like to know how a 4090 performs on a 9900k vs 14900k for example, just so i can gauge the impact of upgrades. I know there are older reference points in every review but it's still mental gymnastics more times than not when you're a little bit obscessed with exact comparison to your current setup.
I think you get what i mean, just a very occasional link to the past to better comprehend the actual value of an upgrade. Keep up your great work, GN is not even a national treasure since you shine around the world but a beacon of light in this industry.
reply
intoamutecrypt5782
I think the whole Bigger uplifts on Linux/Bigger uplifts on admin mode reveals a fundamental difficulty of revieving: You can't do every test. GN's test suite is reflective of a fairly common user: Someone unwilling or unable to leave Windows, who just wants to sit down, build their PC and then use it without sitting down and carefully optimising their specific use case. While it'd be conceivable to add tests outside that (What happens on Linux), those tests would take time and labour - and GN is already devoting all their time and labour to capturing that generic Windows user's use case really, really well. Adding Linux testing would mean making some undesirable tradeoffs for GN, such as not capturing that generic Windows user as well, or pushing so hard for test quantity that test quality starts to slip (or coming out substantially after the first wave of reviews, which is really just unfeasible from a commercial perspective).
This is why it's important for there to be multiple different reviewers in the space. Other reviewers can offer different perspectives, such as catering more towards Linux users. As consumers, a healthy review space with diverse perspectives allows us to pick and choose based on what our use cases actually are, and build a better picture than any single reviewer can offer.
reply
I think the whole Bigger uplifts on Linux/Bigger uplifts on admin mode reveals a fundamental difficulty of revieving: You can't do every test. GN's test suite is reflective of a fairly common user: Someone unwilling or unable to leave Windows, who just wants to sit down, build their PC and then use it without sitting down and carefully optimising their specific use case. While it'd be conceivable to add tests outside that (What happens on Linux), those tests would take time and labour - and GN is already devoting all their time and labour to capturing that generic Windows user's use case really, really well. Adding Linux testing would mean making some undesirable tradeoffs for GN, such as not capturing that generic Windows user as well, or pushing so hard for test quantity that test quality starts to slip (or coming out substantially after the first wave of reviews, which is really just unfeasible from a commercial perspective).
This is why it's important for there to be multiple different reviewers in the space. Other reviewers can offer different perspectives, such as catering more towards Linux users. As consumers, a healthy review space with diverse perspectives allows us to pick and choose based on what our use cases actually are, and build a better picture than any single reviewer can offer.
reply
sandwich2473
Now, the thing with measuring performance in a Digital Audio Workstation is that it's very dependent on a whole bunch of different factors from what VSTs you're using, if you're working with live audio on top of that, what input devices you're using, etc
The most important thing I think is the input delay between your midi keyboard and the sound coming out of the speakers/headphones
If you were to load up a significantly complex project file with several instruments, audio snippets, instruments that are linked to other ones, with effects that were linked between them as well, then you'd probably see a good amount of delay between keystroke and sound generation however I don't know if this would really be all that significant
Going from an FX 4100 to 6700K to a 3600, I didn't notice the delay becoming all that worse when using effects - maybe it's a natural thing over time that I didn't notice because I was trying to avoid it, and I slowly grew into the new headroom I afforded myself with each CPU upgrade, but realistically the only thing I notice being any different is the render time which really isn't important to me at all
It'd give you a fixed number and it's a repeatable test but when people want a good CPU for a DAW they're probably not talking about render times
reply
Now, the thing with measuring performance in a Digital Audio Workstation is that it's very dependent on a whole bunch of different factors from what VSTs you're using, if you're working with live audio on top of that, what input devices you're using, etc
The most important thing I think is the input delay between your midi keyboard and the sound coming out of the speakers/headphones
If you were to load up a significantly complex project file with several instruments, audio snippets, instruments that are linked to other ones, with effects that were linked between them as well, then you'd probably see a good amount of delay between keystroke and sound generation however I don't know if this would really be all that significant
Going from an FX 4100 to 6700K to a 3600, I didn't notice the delay becoming all that worse when using effects - maybe it's a natural thing over time that I didn't notice because I was trying to avoid it, and I slowly grew into the new headroom I afforded myself with each CPU upgrade, but realistically the only thing I notice being any different is the render time which really isn't important to me at all
It'd give you a fixed number and it's a repeatable test but when people want a good CPU for a DAW they're probably not talking about render times
reply
pestilencekills
In regards to Music Production testing - look up DAW Bench, it's a long-running project that benchmarks CPU performance in DAWs. They use Reaper for a few reasons - it's very well-behaved with multi-threading (some industry standards, like Cubase... not so much), you can restrict how many cores it uses, how much buffering it does (beyond audio interface latency). You can also write your own LUA scripts within Reaper, which might be useful for creating test scripts and ensuring repeatability. Super customisable, and it also has a decent CPU usage stats (eg. how much CPU each specific instance of a plugin is using). One of the big UK PC retailers use their methodology to test new CPUs.
DAW Bench uses two versions of basically the same test - they load up a specific plugin until playback fails. One pass is using a VST plugin (which is almost entirely CPU bound), and one uses a virtual instrument (which also loads up the RAM).
Just straight rendering/encoding/exporting audio is unlikely to give you more insight than your video rendering tests already do.
reply
In regards to Music Production testing - look up DAW Bench, it's a long-running project that benchmarks CPU performance in DAWs. They use Reaper for a few reasons - it's very well-behaved with multi-threading (some industry standards, like Cubase... not so much), you can restrict how many cores it uses, how much buffering it does (beyond audio interface latency). You can also write your own LUA scripts within Reaper, which might be useful for creating test scripts and ensuring repeatability. Super customisable, and it also has a decent CPU usage stats (eg. how much CPU each specific instance of a plugin is using). One of the big UK PC retailers use their methodology to test new CPUs.
DAW Bench uses two versions of basically the same test - they load up a specific plugin until playback fails. One pass is using a VST plugin (which is almost entirely CPU bound), and one uses a virtual instrument (which also loads up the RAM).
Just straight rendering/encoding/exporting audio is unlikely to give you more insight than your video rendering tests already do.
reply
KingdomScorned
As an audio engineer with 15 years of experience, I can offer some insights into software that's useful for testing CPU performance. Reaper stands out as one of the best-optimized DAWs for multi-threaded workloads, making it highly efficient for larger mixing sessions. It tends to show almost linear performance gains with stronger CPUs.
A common method for testing both CPU and DAW performance involves loading multiple instances of a demanding plugin, such as a popular guitar amplifier simulator like Neural DSP, alongside a direct guitar performance audio file. The goal is to see how many tracks with the same plugin the CPU can handle before audible artifacts appear in playback.
Another common test is rendering a multi-track session to a single stereo file and comparing the render times across different CPUs. These tests evaluate the CPU's performance differently, as real-time audio processing and non-real-time rendering place distinct demands on the processor.
reply
As an audio engineer with 15 years of experience, I can offer some insights into software that's useful for testing CPU performance. Reaper stands out as one of the best-optimized DAWs for multi-threaded workloads, making it highly efficient for larger mixing sessions. It tends to show almost linear performance gains with stronger CPUs.
A common method for testing both CPU and DAW performance involves loading multiple instances of a demanding plugin, such as a popular guitar amplifier simulator like Neural DSP, alongside a direct guitar performance audio file. The goal is to see how many tracks with the same plugin the CPU can handle before audible artifacts appear in playback.
Another common test is rendering a multi-track session to a single stereo file and comparing the render times across different CPUs. These tests evaluate the CPU's performance differently, as real-time audio processing and non-real-time rendering place distinct demands on the processor.
reply
P3X967
There is a diference between benchmarking for music production and for mixing. First one focuses more on latency number of instrument instances, while the second one focuses on track count number of effect instances. Every DAW will have a different result. Every instrument or effect will have a different result. I'd recommend using reaper since its free/ cheap DAW and pair it with free plugins - free instrument and free chanel strip (eq compressor saturation). I wouldn't use stock plugins since they are integrated and exclusive to the selected DAW. A system needs to be tested on all sample rates starting with 44.1kHz and on all buffer sizes. Here the results will differ based on audio driver selected. The best drivers come with RME devices, but you can use a free one ASIO4ALL for realtek sound card or whatever is integrated in the mobo.
I can go more into details but really dont intend to write a book here in yt comment section. Feel free to DM.
reply
There is a diference between benchmarking for music production and for mixing. First one focuses more on latency number of instrument instances, while the second one focuses on track count number of effect instances. Every DAW will have a different result. Every instrument or effect will have a different result. I'd recommend using reaper since its free/ cheap DAW and pair it with free plugins - free instrument and free chanel strip (eq compressor saturation). I wouldn't use stock plugins since they are integrated and exclusive to the selected DAW. A system needs to be tested on all sample rates starting with 44.1kHz and on all buffer sizes. Here the results will differ based on audio driver selected. The best drivers come with RME devices, but you can use a free one ASIO4ALL for realtek sound card or whatever is integrated in the mobo.
I can go more into details but really dont intend to write a book here in yt comment section. Feel free to DM.
reply
user-ci5gn2ko2n
Gaming Benchmarks are nice, but I capture everything that I play and have my own setup.
I lock most games to 60 FPS 4K and capture at 4K running 1 : 1 in OBS while using DLSS.
I am able to capture to the same machine (or send the data over to my capture/streaming PC).
Intel Nvidia is what I use and I have XTU profiles setup where I downclock and negative voltage offset for each game I play.
(12900K RTX 4090)
Let's say I just want something quick to play for 10 - 30 minutes and I want the greatest efficiency.
Well, for that I use a PC with an AMD G-Series APU. Connected to that system is a Arzopa Portable Monitor.
The entire system is connected to Solar Power and works just fine. I get around 4 - 6 hours of gaming at night
when the sun is down. My only two rules are 1080P Gaming must reach and lock 60 FPS.
reply
Gaming Benchmarks are nice, but I capture everything that I play and have my own setup.
I lock most games to 60 FPS 4K and capture at 4K running 1 : 1 in OBS while using DLSS.
I am able to capture to the same machine (or send the data over to my capture/streaming PC).
Intel Nvidia is what I use and I have XTU profiles setup where I downclock and negative voltage offset for each game I play.
(12900K RTX 4090)
Let's say I just want something quick to play for 10 - 30 minutes and I want the greatest efficiency.
Well, for that I use a PC with an AMD G-Series APU. Connected to that system is a Arzopa Portable Monitor.
The entire system is connected to Solar Power and works just fine. I get around 4 - 6 hours of gaming at night
when the sun is down. My only two rules are 1080P Gaming must reach and lock 60 FPS.
reply
fabrb26
I upgraded from a 5600 non X to a 5800X3D with the market blowing at my back i was able to sell it the price i got it so that was nice but i was still feeling like a missed a lot from not having a DDR5 / 7800X3D. Nice yo see i probably made the right choice considering selling price of last gen and those charts since i mainly play FFXIV anyway and still rocking a 2021 chaos era 6500XT who was... SADLY... The best and only new buy i could make at that moment for my remaining budget, being ripped off by some MF buying a 2nd hand Nvidia gpu at first... So i would say a GPU is the only thing i need for the next 10-15 years or so until TES6 come out
reply
I upgraded from a 5600 non X to a 5800X3D with the market blowing at my back i was able to sell it the price i got it so that was nice but i was still feeling like a missed a lot from not having a DDR5 / 7800X3D. Nice yo see i probably made the right choice considering selling price of last gen and those charts since i mainly play FFXIV anyway and still rocking a 2021 chaos era 6500XT who was... SADLY... The best and only new buy i could make at that moment for my remaining budget, being ripped off by some MF buying a 2nd hand Nvidia gpu at first... So i would say a GPU is the only thing i need for the next 10-15 years or so until TES6 come out
reply
MrHerrS
Thank you for adding more and more productivity tests.
As a Java, .Net / C# and Phyton developer it would be a bonus if we could see some productivity tests on Linux in the far future. Not sure about good testing methodology. What I usually struggle performance wise is.
- running complex sql statements (testing must be well prepared regarding buffers, execution plans and IO accesses)
- compiling big projects
- running a lot of applications in parallel. (Including containerized apps)
- JS performance on heavy websites (miro, atlassian products, postman on web )
- Training ML models (usually NNs)
reply
Thank you for adding more and more productivity tests.
As a Java, .Net / C# and Phyton developer it would be a bonus if we could see some productivity tests on Linux in the far future. Not sure about good testing methodology. What I usually struggle performance wise is.
- running complex sql statements (testing must be well prepared regarding buffers, execution plans and IO accesses)
- compiling big projects
- running a lot of applications in parallel. (Including containerized apps)
- JS performance on heavy websites (miro, atlassian products, postman on web )
- Training ML models (usually NNs)
reply
EthosAtheos
For audio production software...
Pro Tools is probably the 8000lb gorilla, REAPER is an amazing DAW and is a solid option for people without unlimited budgets.
FL Studio is a very common and well liked application for beats. Cubase is a popular option good midi.
Audition is a good representation of what happens when Adobe buys an awesome company and then does almost nothing with it. It's worth a look because adobe...
O and for fun and profit a common business task Wave to MP3 conversion using ffdshow.
How you test their performance....Good luck building automation and measuring tools.
reply
For audio production software...
Pro Tools is probably the 8000lb gorilla, REAPER is an amazing DAW and is a solid option for people without unlimited budgets.
FL Studio is a very common and well liked application for beats. Cubase is a popular option good midi.
Audition is a good representation of what happens when Adobe buys an awesome company and then does almost nothing with it. It's worth a look because adobe...
O and for fun and profit a common business task Wave to MP3 conversion using ffdshow.
How you test their performance....Good luck building automation and measuring tools.
reply
blackbird42
12:40 fun fact - in engineering some units are purposefully left not simplified, cause the unit left as is better represents the parameter. This is one of those cases - you are interested in actual power usage rather than the energy that you'd have to recalculate for bills anyway ([kWh] btw, [J/s] [3600s] 1000, so essentially J3600000, other example would be an angular speed in [rad/s], where radians are esentially dimensionless and that gives us [1/s], which is [Hz]. While technically true, first one doesn't make any sense, the second one has more sense but it doesn't represent the parameter well.)
reply
12:40 fun fact - in engineering some units are purposefully left not simplified, cause the unit left as is better represents the parameter. This is one of those cases - you are interested in actual power usage rather than the energy that you'd have to recalculate for bills anyway ([kWh] btw, [J/s] [3600s] 1000, so essentially J3600000, other example would be an angular speed in [rad/s], where radians are esentially dimensionless and that gives us [1/s], which is [Hz]. While technically true, first one doesn't make any sense, the second one has more sense but it doesn't represent the parameter well.)
reply
meneldal
I do have a question for efficiency, is there any point to use only the cpu power when estimating costs You will typically get better results with tdps around 30-40W on the cpu, but other stuff also uses power, so if you were doing something like turning your pc on, building chromium then shutting it down, cpu power would only be a fraction of the total. It could be giving an unfair advantage for low power cpus that still have to draw power for ram, chipset, storage and so on.
You might want to run tests with different memory to see if faster memory is worth the extra power draw too.
reply
I do have a question for efficiency, is there any point to use only the cpu power when estimating costs You will typically get better results with tdps around 30-40W on the cpu, but other stuff also uses power, so if you were doing something like turning your pc on, building chromium then shutting it down, cpu power would only be a fraction of the total. It could be giving an unfair advantage for low power cpus that still have to draw power for ram, chipset, storage and so on.
You might want to run tests with different memory to see if faster memory is worth the extra power draw too.
reply
teku9691
Hey GN & Steve, I would really love to have a new test parameter on cpu intensive benchmarking via Autodesk maya’s Arnold rendering. Although you include blender, it’s not as widely used in profession settings from my experience. I’ve noticed that many if not most reviewers choose to not use maya and I know it’s a niche, but most of the production and animation industry uses it extensively. So for people like me who want systems that can help me get a foot into that industry, that parameter would help with differentiation between hardware.
Thank you
reply
Hey GN & Steve, I would really love to have a new test parameter on cpu intensive benchmarking via Autodesk maya’s Arnold rendering. Although you include blender, it’s not as widely used in profession settings from my experience. I’ve noticed that many if not most reviewers choose to not use maya and I know it’s a niche, but most of the production and animation industry uses it extensively. So for people like me who want systems that can help me get a foot into that industry, that parameter would help with differentiation between hardware.
Thank you
reply
BladeCrew
Steve there is something I want you to know, the windows issues also affect am4 cpu's. My ryzen 5 5500 clock speed jumps between 3GHz and 4.25GHz. I am using an ID-Cooling FrosFlow X 280 cpu liquid cooler and the max temp on my cpu is 48C. I am using an asrock B450M Steel Legend with the latest BIOS update and an Zotac RTX 3050 Twin Edge OC 8GB. All my games fluctuate between 3GHz to 4.25GH and even in cpu bound games it does the same. Dota 2, 3D Mark Firestrike benchmarks, etc. Could you look into why zen 4 is running like this.
reply
Steve there is something I want you to know, the windows issues also affect am4 cpu's. My ryzen 5 5500 clock speed jumps between 3GHz and 4.25GHz. I am using an ID-Cooling FrosFlow X 280 cpu liquid cooler and the max temp on my cpu is 48C. I am using an asrock B450M Steel Legend with the latest BIOS update and an Zotac RTX 3050 Twin Edge OC 8GB. All my games fluctuate between 3GHz to 4.25GH and even in cpu bound games it does the same. Dota 2, 3D Mark Firestrike benchmarks, etc. Could you look into why zen 4 is running like this.
reply
UsatiyNyan
I’m thinking, a thread-switching/IO heavy benchmark could be pretty good for perspective.
Examples:
- Raid of TigerBeetle databases, they have pretty good simulation tests (measure RPS)
- a Go server under apache benchmark with expected 100k RPS (measure RPS)
RPS - requests per second (fulfilled)
These benchmarks are ultimately IO bound, but they would require program and the processor to switch between multiple threads constantly. This would measure thread-switching overhead and atomic operation overhead.
reply
I’m thinking, a thread-switching/IO heavy benchmark could be pretty good for perspective.
Examples:
- Raid of TigerBeetle databases, they have pretty good simulation tests (measure RPS)
- a Go server under apache benchmark with expected 100k RPS (measure RPS)
RPS - requests per second (fulfilled)
These benchmarks are ultimately IO bound, but they would require program and the processor to switch between multiple threads constantly. This would measure thread-switching overhead and atomic operation overhead.
reply
NWMcCabe11
Not sure if this has been requested already, but I have recently gotten into FPV drone Flying and basically instead of crashing your hundreds of dollars worth of parts and countless hours of soldering into the concrete or brick wall you use a few different sims And so far, I have found two different Sims that make my current set up just immediately and absolutely take a shit. They are both on steam Tryp fpv 2.0 and Liftoff.uncrashed is better on pc’s but those are top 3 IMO
reply
Not sure if this has been requested already, but I have recently gotten into FPV drone Flying and basically instead of crashing your hundreds of dollars worth of parts and countless hours of soldering into the concrete or brick wall you use a few different sims And so far, I have found two different Sims that make my current set up just immediately and absolutely take a shit. They are both on steam Tryp fpv 2.0 and Liftoff.uncrashed is better on pc’s but those are top 3 IMO
reply
Nabruno
This request might be a bit too niche, but I would love to see FME Form included in the benchmark list. It's a software used for spatial transformation, automation and data integration in a bunch of industries. Depending on the script and the data that is being processed it can take anywhere from less than a second to hours to run. And since it is script based out of the box every run is exactly the same, so it would test well with a shorter time per run = better performance.
reply
This request might be a bit too niche, but I would love to see FME Form included in the benchmark list. It's a software used for spatial transformation, automation and data integration in a bunch of industries. Depending on the script and the data that is being processed it can take anywhere from less than a second to hours to run. And since it is script based out of the box every run is exactly the same, so it would test well with a shorter time per run = better performance.
reply
woobilicious.
DAW's are heavily constrained by drivers and context switching, I do wonder if there's going to be nothing here to test as context switching between the application, and the audio driver will completely flush the pipeline, and thus completely eliminate IPC improvements, on the other hand it could be the opposite where older CPUs perform better because their pipelines aren't nearly as deep. Running DPC Monitoring on AMD and Nvidia drivers maybe interesting too.
reply
DAW's are heavily constrained by drivers and context switching, I do wonder if there's going to be nothing here to test as context switching between the application, and the audio driver will completely flush the pipeline, and thus completely eliminate IPC improvements, on the other hand it could be the opposite where older CPUs perform better because their pipelines aren't nearly as deep. Running DPC Monitoring on AMD and Nvidia drivers maybe interesting too.
reply
RayJohnson1980
well file decompression between 7950x vs 9950x is only 1.5% uplift. but suggest break the generations down on this down to MIPS per core so to compare efficency of the 9950x vs 9900x vs 9700x vs R5 77600x because i see an issue. 17,525 mips/core on 9950x , 18,127mips/core on 9900x and 16,951mips/core on 9700x and lastly 18,203mips/core which is better than all of current generation per core efficiency from a previous generation. where did amd go wrong
reply
well file decompression between 7950x vs 9950x is only 1.5% uplift. but suggest break the generations down on this down to MIPS per core so to compare efficency of the 9950x vs 9900x vs 9700x vs R5 77600x because i see an issue. 17,525 mips/core on 9950x , 18,127mips/core on 9900x and 16,951mips/core on 9700x and lastly 18,203mips/core which is better than all of current generation per core efficiency from a previous generation. where did amd go wrong
reply
Add a review, comment
Other channel videos
