95 C is Now Normal: AMD Ryzen 9 7950X CPU Review & Benchmarks
video description
Date: 2022-09-27
Related videos
Comments and reviews: 14
ModernOddity
Using MX-4 paste on a high-end CPU like this with higher TDP & TJM, I can appreciate wanting to show what a casual shopper would use. But if you're an enthusiast then every few degrees saved matters and you should opt to use Thermal Grizzly Kryonaut Extreme because of the better conductivity.
I've seen plenty of GN ads for it, so I don't know why you wouldn't suggest that (I do find it suggestive that MX-4 is still the baseline seeing it being used here), it would certainly help the more casual builders with their temps if they have an extra 20 since they can afford the top range of new CPUs. Like in my mind, if it's designed to run at 95C why not invest the pocket money in getting that to have it run a few degrees lower? One of the easiest and cheapest ways to get a quick temp improvement.
Like I said, I get the mentality of MX-4 being the baseline for years and it gets the job done, but with rising TDPs in CPUs and GPUs it may be time to adjust that baseline for testing and recommendations moving forward..
reply
Using MX-4 paste on a high-end CPU like this with higher TDP & TJM, I can appreciate wanting to show what a casual shopper would use. But if you're an enthusiast then every few degrees saved matters and you should opt to use Thermal Grizzly Kryonaut Extreme because of the better conductivity.
I've seen plenty of GN ads for it, so I don't know why you wouldn't suggest that (I do find it suggestive that MX-4 is still the baseline seeing it being used here), it would certainly help the more casual builders with their temps if they have an extra 20 since they can afford the top range of new CPUs. Like in my mind, if it's designed to run at 95C why not invest the pocket money in getting that to have it run a few degrees lower? One of the easiest and cheapest ways to get a quick temp improvement.
Like I said, I get the mentality of MX-4 being the baseline for years and it gets the job done, but with rising TDPs in CPUs and GPUs it may be time to adjust that baseline for testing and recommendations moving forward..
reply
Jordan
I think it's clear that CPU efficiency needs to be tested separately from maximum performance.
Any CPU at max performance will pull as much power as your thermal solution can handle.
If you tested the processor at the same performance , it would almost certainly be more efficient. I think if you want to make claims about CPU efficiency relative to other processors, you need to pick a power level, lock the processor at that power, and see how it performs. It's not useful to rate (especially this) processor performance when it's configured to run at maximum tdp, as it simply means a better cooler will perform better, but with worse efficiency.
If you want to meaningfully analyze efficiency, I'd lock the processor at, say, 100w, run the tests and compare that way. Your current efficiency numbers are meaningless, as all they really show is running a processor 500MHz below the ragged edge reduces power consumption by 50% ( cough 5950x )
reply
I think it's clear that CPU efficiency needs to be tested separately from maximum performance.
Any CPU at max performance will pull as much power as your thermal solution can handle.
If you tested the processor at the same performance , it would almost certainly be more efficient. I think if you want to make claims about CPU efficiency relative to other processors, you need to pick a power level, lock the processor at that power, and see how it performs. It's not useful to rate (especially this) processor performance when it's configured to run at maximum tdp, as it simply means a better cooler will perform better, but with worse efficiency.
If you want to meaningfully analyze efficiency, I'd lock the processor at, say, 100w, run the tests and compare that way. Your current efficiency numbers are meaningless, as all they really show is running a processor 500MHz below the ragged edge reduces power consumption by 50% ( cough 5950x )
reply
Profound
There is an important benchmark all reviewers are missing. If they would have done it for the 5950x I probably would not have bought this cpu. I'm talking about the multi-thread integer performance improvement going from 16 threads to 32 threads on a 16 core cpu. For example the ryzen 3900X has a good performance improvement from 12 threads to 24 threads on classic chess and checkers benchmarks. The 5950x has not in fact it hardly improves performance going from 16 threads to 32 threads in this type of integer benchmarks. I've seen absolutely no one making this type of benchmark.
It's as if the AMD 3000 series were meant for multithreading workloads and the 5000 series were meant for gaming workloads.
And this is a big difference for me. This type of test is a good way to see how good those hyperthreading cores are. They are very good on the 3000 series but on the 5000 series are a disgrace.
reply
There is an important benchmark all reviewers are missing. If they would have done it for the 5950x I probably would not have bought this cpu. I'm talking about the multi-thread integer performance improvement going from 16 threads to 32 threads on a 16 core cpu. For example the ryzen 3900X has a good performance improvement from 12 threads to 24 threads on classic chess and checkers benchmarks. The 5950x has not in fact it hardly improves performance going from 16 threads to 32 threads in this type of integer benchmarks. I've seen absolutely no one making this type of benchmark.
It's as if the AMD 3000 series were meant for multithreading workloads and the 5000 series were meant for gaming workloads.
And this is a big difference for me. This type of test is a good way to see how good those hyperthreading cores are. They are very good on the 3000 series but on the 5000 series are a disgrace.
reply
Ultranerd
Thanks for the review, informative as always! Very impressive numbers (particularly as I already have an overkill custom loop setup and would be able to re-use most of it). The level of uplift in production workloads at matched corecount is nice, but not game-changing, and similarly with the gaming results. Overall, about as good as could've reasonably been hoped, but not enough to make it worth it for me to eat the upgrade cost (including platform upgrade cost of new mobo/ram) over my current 5950x system given my usecases (I have significant nongame all-core workloads but only for hobby purposes, not moneymakers, and my main games are minecraft-with-mods-and-shaders, which is 100% GPU bound, and dwarf fortress, which is 100% CPU bound and 100% singlethreaded, which would benefit but not hugely).
reply
Thanks for the review, informative as always! Very impressive numbers (particularly as I already have an overkill custom loop setup and would be able to re-use most of it). The level of uplift in production workloads at matched corecount is nice, but not game-changing, and similarly with the gaming results. Overall, about as good as could've reasonably been hoped, but not enough to make it worth it for me to eat the upgrade cost (including platform upgrade cost of new mobo/ram) over my current 5950x system given my usecases (I have significant nongame all-core workloads but only for hobby purposes, not moneymakers, and my main games are minecraft-with-mods-and-shaders, which is 100% GPU bound, and dwarf fortress, which is 100% CPU bound and 100% singlethreaded, which would benefit but not hugely).
reply
MrMcGreed
1: Can't wait to see XOC content on this: The boost tech + lapping will be hella-interesting (here's to hoping Jay and Steve will get into another battle when the new GPUs come out)
2: Wasn't intending on buying and really not doing it now with that heat-output
3: Would love to see the 7950X and 7900X vs 5950X and 5900X at similar powerdraw - wondering how much of the leap of the 7000-series is due to the powerdraw
4: Holy hell, GN has the next review out already, nice job! - very happy to be a Patreon-supporter.
5: I don't play games at less than 600 FPS - up until now I could only play pong, so happy to see R6S is finally near playable...
6: Hucking fell, this beast with 3D V-cache... thermal nightmare, but what a monster...
reply
1: Can't wait to see XOC content on this: The boost tech + lapping will be hella-interesting (here's to hoping Jay and Steve will get into another battle when the new GPUs come out)
2: Wasn't intending on buying and really not doing it now with that heat-output
3: Would love to see the 7950X and 7900X vs 5950X and 5900X at similar powerdraw - wondering how much of the leap of the 7000-series is due to the powerdraw
4: Holy hell, GN has the next review out already, nice job! - very happy to be a Patreon-supporter.
5: I don't play games at less than 600 FPS - up until now I could only play pong, so happy to see R6S is finally near playable...
6: Hucking fell, this beast with 3D V-cache... thermal nightmare, but what a monster...
reply
DaLeadBull
Good luck cooling these things. These new CPU's are impressive from a performance point of view but I feel like the pendulum has swung too far in the performance direction leaving the efficiency behind. Though the performance/efficiency ratio might have improved over the last generation, these new CPU's (and GPU's) are consuming a lot of power and producing a lot of heat which means you need to spend a bunch of money to power and cool these things. This means new PSU's and new coolers/fans etc., this just means the consumer has to factor in all of these new expenses into their budget when deciding to upgrade there systems to these new CPU's (GPU's). I like the performance uplift but not sure I like all these other things that come with it.
reply
Good luck cooling these things. These new CPU's are impressive from a performance point of view but I feel like the pendulum has swung too far in the performance direction leaving the efficiency behind. Though the performance/efficiency ratio might have improved over the last generation, these new CPU's (and GPU's) are consuming a lot of power and producing a lot of heat which means you need to spend a bunch of money to power and cool these things. This means new PSU's and new coolers/fans etc., this just means the consumer has to factor in all of these new expenses into their budget when deciding to upgrade there systems to these new CPU's (GPU's). I like the performance uplift but not sure I like all these other things that come with it.
reply
Anders
... sometimes at more power requirement and extreme cooling solution ... er, no. You just explained it yourself: if you slap a gigantic cooler on it, it WILL suck down more juice. It will keep slurping up that juice until it reaches 95C. If you had tested this on a more reasonable tower cooler it would have performed a little worse but used a lot less power. Those last 100-200MHz are costly as fsck in wattage.
I would LOVE to see a cooler vs performance vs power draw benchmark of Zen 4. Personally I think its the right move by AMD: Even boneheads can figure out bigger cooler = more performance = higher power bill . But I'd really like to see the scaling! :D
reply
... sometimes at more power requirement and extreme cooling solution ... er, no. You just explained it yourself: if you slap a gigantic cooler on it, it WILL suck down more juice. It will keep slurping up that juice until it reaches 95C. If you had tested this on a more reasonable tower cooler it would have performed a little worse but used a lot less power. Those last 100-200MHz are costly as fsck in wattage.
I would LOVE to see a cooler vs performance vs power draw benchmark of Zen 4. Personally I think its the right move by AMD: Even boneheads can figure out bigger cooler = more performance = higher power bill . But I'd really like to see the scaling! :D
reply
Scott
Anyone else notice that Steve presented very high frame rates in screen after screen, but waited until the one test where intel was still on top to mock the whole idea of chasing high frame rates at this level? Suddenly, not only does it not matter, but it's silly to think it does. I think GN is neutral with their testing and results, but show a bias when presenting new CPUs and GPUs on the comparison charts. It's comes out in wording they use, what they mock, what they emphasize, what order they present games in, etc. The numbers are fair, but there's always a tint in the presentation.
reply
Anyone else notice that Steve presented very high frame rates in screen after screen, but waited until the one test where intel was still on top to mock the whole idea of chasing high frame rates at this level? Suddenly, not only does it not matter, but it's silly to think it does. I think GN is neutral with their testing and results, but show a bias when presenting new CPUs and GPUs on the comparison charts. It's comes out in wording they use, what they mock, what they emphasize, what order they present games in, etc. The numbers are fair, but there's always a tint in the presentation.
reply
Kevin
Steve Please look into what is going on with the Secretlab Magnus Pro electrical issue. Apparently, the desk is a fire hazard and Secretlab is asking everyone to unplug their power strips from the desk. It has been a week since they contacted me with no information on how they plan on resolving this concerning issue. Spending 1000k plus on this desk and not happy with the fact i cannot use a major function of the desk. Also, just look at all the complaints on desks being destroyed by FedEx when delivering. This is not the best launch from a company that is plugged for its quality.
reply
Steve Please look into what is going on with the Secretlab Magnus Pro electrical issue. Apparently, the desk is a fire hazard and Secretlab is asking everyone to unplug their power strips from the desk. It has been a week since they contacted me with no information on how they plan on resolving this concerning issue. Spending 1000k plus on this desk and not happy with the fact i cannot use a major function of the desk. Also, just look at all the complaints on desks being destroyed by FedEx when delivering. This is not the best launch from a company that is plugged for its quality.
reply
Sean
I know its not as exciting but do these chips have an eco mode like my 3700x does? I run mine at its default eco mode 45w and I have no issue with the performance of it. Its getting a little old now and I'm wanting to upgrade I'm just hesitant due to the new platform adoption costs and insane ideology of lets just throw electricity at the problem to get more performance and be dammed with the planet and running costs.
If they do have an eco mode not just manual undervolting, do you have any data with impact to performance and power data in this mode?
reply
I know its not as exciting but do these chips have an eco mode like my 3700x does? I run mine at its default eco mode 45w and I have no issue with the performance of it. Its getting a little old now and I'm wanting to upgrade I'm just hesitant due to the new platform adoption costs and insane ideology of lets just throw electricity at the problem to get more performance and be dammed with the planet and running costs.
If they do have an eco mode not just manual undervolting, do you have any data with impact to performance and power data in this mode?
reply
John
You could cook an EGG on that CPU! Great so now I can cook breakfast and game at the same time!
LOL. 95c is 203f.
I guess delidding is going to be a must now on these 7000 series cpus. Delidding should drop those temps by at lest 20c. I Guarantee Thermal grizzly or somebody comes out with a delidding solution for this! I would if I had the funding!
Great job Steve & team! As always thank you for all your hard work and long hours! You guy's are always the go too if you want the truth!
reply
You could cook an EGG on that CPU! Great so now I can cook breakfast and game at the same time!
LOL. 95c is 203f.
I guess delidding is going to be a must now on these 7000 series cpus. Delidding should drop those temps by at lest 20c. I Guarantee Thermal grizzly or somebody comes out with a delidding solution for this! I would if I had the funding!
Great job Steve & team! As always thank you for all your hard work and long hours! You guy's are always the go too if you want the truth!
reply
glimm3r
I have a feeling people will find a way to limit the temperature, lose like 10% performance in production but 30% power consumption, and make this a way more attractive product.
On another note: isn't it strange that they increased the clocks a bit over last gen, had some minor IPC improvements, die shrinkage, DDR5 and the performance uplift is nothing to write home about? Maybe Zen4 is hitting a bottleneck that AMD couldn't figure out?
reply
I have a feeling people will find a way to limit the temperature, lose like 10% performance in production but 30% power consumption, and make this a way more attractive product.
On another note: isn't it strange that they increased the clocks a bit over last gen, had some minor IPC improvements, die shrinkage, DDR5 and the performance uplift is nothing to write home about? Maybe Zen4 is hitting a bottleneck that AMD couldn't figure out?
reply
rworhatch
Since AMD and Intel are both forcing us to explore the extreme power draw regime, I would be interested in a video that explores just how high you can get power efficiency for different sorts of parts by engaging in 'extreme underclocking.' Is there a limit to how much extra efficiency you can get by dropping power consumption? Does a flagship part do better or worse than a midrange part when used that way?
reply
Since AMD and Intel are both forcing us to explore the extreme power draw regime, I would be interested in a video that explores just how high you can get power efficiency for different sorts of parts by engaging in 'extreme underclocking.' Is there a limit to how much extra efficiency you can get by dropping power consumption? Does a flagship part do better or worse than a midrange part when used that way?
reply
Ale
it's incredible how much squeezed are cpus and gpus today, great for getting top performance, but efficiency goes out of the window (at least on the stock config, that most users will use) It would be nice to advise customers to use a less agressive (and much more efficient) mode. I'm now really excited to see intels 13 gen (Who would have though I would be so excited for an intel release). Competition is great.
reply
it's incredible how much squeezed are cpus and gpus today, great for getting top performance, but efficiency goes out of the window (at least on the stock config, that most users will use) It would be nice to advise customers to use a less agressive (and much more efficient) mode. I'm now really excited to see intels 13 gen (Who would have though I would be so excited for an intel release). Competition is great.
reply
Add a review, comment
Other channel videos
