[SOLVED] What do CPU Benchmarks Actually Mean?

Dec 20, 2019
83
4
35
0
I'm trying to decide which CPU to buy, and looking at benchmarks and stats on the site userbenchmark.com. However, slightly confused now.
It's saying that an Intel Core i3-10320 is slightly better than a Ryzen 9 3900X . The i3 is a 4 core, 8 thread processor, the Ryzen 9 is a 12 core 24 thread processor! How can the i3 possibly be better?
What do these benchmarks actually mean and should I pay them any attention?
I was going to buy a Ryzen 7 3700X but there's a lot of CPUs that rank higher on that site, and some which are cheaper. For example the i3 10300.
Very confused now. How do I know which CPU to choose?
Also, Intel CPUS, even with lower cores and threads are rated higher than AMD Cpus. How come?
 
Ahh yes ok I see. Again, I don't really understand how cores and threads work so I was just kinda going with the scores on that site. Obviously it's not that simple! Thanks.
In the context of CPUs, "threads" refers to how many software threads a processor can run efficiently while still seeing significant performance gains. Processors with SMT (Simultaneous Multi-Threading, also called Hyperthreading on Intel's CPUs) can run two threads on each core simultaneously, utilizing parts of the core that would otherwise be sitting idle when only a single thread is being processed on that core. When two threads are running on a single core, they won't both be operating at full speed, but in certain software that can divide a task to be split across all available threads, one might see around a 50% speedup in some cases, without increasing power draw much. It's possible for CPUs to run more software threads simultaneously than they are designed to efficiently handle, but they generally won't see significant performance improvements from doing so, and may in fact see less-stable performance in realtime applications like games, since a given core needs to keep rapidly switching back and forth between which threads it is currently processing.

I can wait! That's a good idea. Will the 5800x be much more powerful than the 3700x? What about the Ryzen 5 5600X, how would it compare?
At least going by AMD's numbers, the 5000-series processors should offer somewhere around 19% more performance per clock on average compared to the 3000-series, while also featuring slightly higher clock rates. So, it's expected that performance at CPU-limited tasks should typically be over 20% faster for a given core count, though the exact performance difference may vary significantly from one application to the next.

That applies to the 5600X too, though that processor has fewer cores/threads than a 3700X or 5800X. Most applications are not going to be making heavy use of more than a handful of threads, so for most tasks, a 5600X should be faster than a 3700X, since the extra cores will be sitting mostly idle at those times. In software that can divide a task evenly between all available threads though, I suspect the 3700X might manage to be a little faster most of the time. That's because while the 5600X may offer over 20% more performance per thread, the 3700X has 33% more threads to go around.

Most software doesn't fit into that category though, as many tasks can't easily be divided like that, but some things can. For example, relevant to video editing, a CPU-based video encoder can generally split its workload evenly across many threads, as each thread can be assigned to process a different frame of the video at the same time, with minimal communication between them. Again though, unless you are encoding video all the time and are primarily concerned about cutting down encode times above all else, lots of other software will benefit more from having more performance available per core rather than more cores.

While we'll have to wait for reviews to see exactly how performance fares in different software, I suspect the 5600X will probably be a better choice than a 3700X, at least if they are priced about the same. It seems likely that the 5600X's higher performance per core should allow it to come relatively close to the 3700X at heavily-multithreaded tasks like video encoding, while being notably faster at the more common lightly-threaded tasks. The 5800X offers both the higher performance per core compared to the 3700X, as well as the higher core count compared to the 5600X, though there's a rather large price jump over those models. While the 5800X does offer 33% more cores/threads than the 5600X, it's set to cost $450, or about 50% more. And again, if we look at performance in most common software, which won't be utilizing all of its cores, it's performance should be about the same as the 5600X, probably no more than about 2% faster judging by clock rates.
 
Don’t use userbenchmark for comparing CPU’s. It is known to have an unfair bias towards Intel. Use reviews from trusted places like Toms.

Edit: Also don’t make any decision on hardware based on a single review. I’d always try to look at 3 or more different reviews to try and build a balanced picture.
 
Reactions: kurdtnz
Userbenchmark is hopelessly biased towards Intel, that's the first and foremost thing to keep in mind. That said, that site is useless for comparing different products, especially across brands.

So...for THAT benchmark site at least...it's only use is to find out how your system is performing compared to identical or similar systems. But even then it's limited utility because the results are skewed by people who've made multiple runs with highly over-clocked systems, often on LN2 and therefore impractical for real-world useage. So now it's really only useful for you to compare results with your own system as you upgrade or tweak settings.
 
I'm trying to decide which CPU to buy, and looking at benchmarks and stats on the site userbenchmark.com.
It's saying that an Intel Core i3-10320 is slightly better than a Ryzen 9 3900X .
It's obviously BS. Their cpu score calculation algorithm is flawed.
Don't use userbenchmark as only way for comparing cpu performance.

Userbenchmark site can be used for diagnosing system deficiencies quickly.
But other than this feature rest of it is rather questionable.
 

kanewolf

Titan
Moderator
I'm trying to decide which CPU to buy, and looking at benchmarks and stats on the site userbenchmark.com. However, slightly confused now.
It's saying that an Intel Core i3-10320 is slightly better than a Ryzen 9 3900X . The i3 is a 4 core, 8 thread processor, the Ryzen 9 is a 12 core 24 thread processor! How can the i3 possibly be better?
What do these benchmarks actually mean and should I pay them any attention?
I was going to buy a Ryzen 7 3700X but there's a lot of CPUs that rank higher on that site, and some which are cheaper. For example the i3 10300.
Very confused now. How do I know which CPU to choose?
Also, Intel CPUS, even with lower cores and threads are rated higher than AMD Cpus. How come?
There is no single "best" CPU. There are best CPUs at specific price points. The cost of a 3900X is very different than an i3-10320. There are "best" CPUs for different applications also. The target applications for a 3900X is different than the i3. Telling us what you can spend, and what your use is will give you the best crowdsourced recommendations.
 

DMAN999

Reputable
Ambassador
The 3700x is an excellent CPU for gaming and for video encoding/multi core loads.
Read some of these:
https://www.pcgamesn.com/amd/ryzen-7-3700x-review-benchmarks-performance#nn-graph-sheet-cinebench-r20
 

PrinceTexasLoaf

Respectable
Sep 28, 2018
739
81
2,140
29
I'm trying to decide which CPU to buy, and looking at benchmarks and stats on the site userbenchmark.com. However, slightly confused now.
It's saying that an Intel Core i3-10320 is slightly better than a Ryzen 9 3900X . The i3 is a 4 core, 8 thread processor, the Ryzen 9 is a 12 core 24 thread processor! How can the i3 possibly be better?
What do these benchmarks actually mean and should I pay them any attention?
I was going to buy a Ryzen 7 3700X but there's a lot of CPUs that rank higher on that site, and some which are cheaper. For example the i3 10300.
Very confused now. How do I know which CPU to choose?
Also, Intel CPUS, even with lower cores and threads are rated higher than AMD Cpus. How come?
the i3 is only better is games like minecraft where practically the only thing that matters is single core performance. So super minimal low-end games that require a potato to run is where the i3 shreds. any game that's cpu demanding the i3 would be easily brought to it's knees. Userbenchmark rates gaming performance almost solely on single core performance hence why it gets those confusing results ;o scroll down to the bottom of this page https://cpu.userbenchmark.com/Compare/Intel-Core-i3-10320-vs-AMD-Ryzen-9-3900X/m1180543vs4044
and there you can see the comparison of them multi-core wise. It's in the "Nice to haves" section.
 
Dec 20, 2019
83
4
35
0
There is no single "best" CPU. There are best CPUs at specific price points. The cost of a 3900X is very different than an i3-10320. There are "best" CPUs for different applications also. The target applications for a 3900X is different than the i3. Telling us what you can spend, and what your use is will give you the best crowdsourced recommendations.
Thanks. What I need is a CPU for video editing, capable of handling 4k with ease, primarily. Price would be preferably under $500. I was considering the 3700x.
 
Dec 20, 2019
83
4
35
0
The 3700x is an excellent CPU for gaming and for video encoding/multi core loads.
Read some of these:
https://www.pcgamesn.com/amd/ryzen-7-3700x-review-benchmarks-performance#nn-graph-sheet-cinebench-r20
Thanks! Yeah that was the one I'd kind of decided on, I'd heard it was a kind of "sweet spot". Then I just got confused by that site, but you guys have helped clear that up!
 
Dec 20, 2019
83
4
35
0
the i3 is only better is games like minecraft where practically the only thing that matters is single core performance. So super minimal low-end games that require a potato to run is where the i3 shreds. any game that's cpu demanding the i3 would be easily brought to it's knees. Userbenchmark rates gaming performance almost solely on single core performance hence why it gets those confusing results ;o scroll down to the bottom of this page https://cpu.userbenchmark.com/Compare/Intel-Core-i3-10320-vs-AMD-Ryzen-9-3900X/m1180543vs4044
and there you can see the comparison of them multi-core wise. It's in the "Nice to haves" section.
Ahh yes ok I see. Again, I don't really understand how cores and threads work so I was just kinda going with the scores on that site. Obviously it's not that simple! Thanks.
 

kanewolf

Titan
Moderator
Thanks. What I need is a CPU for video editing, capable of handling 4k with ease, primarily. Price would be preferably under $500. I was considering the 3700x.
The 3700X would be good for video editing. BUT, if you can wait a month, and allow the new AMD CPUs to be released and see what that does to the prices of 3700X or get the 5700X.
For video editing, you want 32GB RAM (2x16GB), and a 1 or 2TB SSD to work from.
 
Reactions: DMAN999
Dec 20, 2019
83
4
35
0
The 3700X would be good for video editing. BUT, if you can wait a month, and allow the new AMD CPUs to be released and see what that does to the prices of 3700X or get the 5700X.
For video editing, you want 32GB RAM (2x16GB), and a 1 or 2TB SSD to work from.
I can wait! That's a good idea. Will the 5800x be much more powerful than the 3700x? What about the Ryzen 5 5600X, how would it compare?
 
Ahh yes ok I see. Again, I don't really understand how cores and threads work so I was just kinda going with the scores on that site. Obviously it's not that simple! Thanks.
In the context of CPUs, "threads" refers to how many software threads a processor can run efficiently while still seeing significant performance gains. Processors with SMT (Simultaneous Multi-Threading, also called Hyperthreading on Intel's CPUs) can run two threads on each core simultaneously, utilizing parts of the core that would otherwise be sitting idle when only a single thread is being processed on that core. When two threads are running on a single core, they won't both be operating at full speed, but in certain software that can divide a task to be split across all available threads, one might see around a 50% speedup in some cases, without increasing power draw much. It's possible for CPUs to run more software threads simultaneously than they are designed to efficiently handle, but they generally won't see significant performance improvements from doing so, and may in fact see less-stable performance in realtime applications like games, since a given core needs to keep rapidly switching back and forth between which threads it is currently processing.

I can wait! That's a good idea. Will the 5800x be much more powerful than the 3700x? What about the Ryzen 5 5600X, how would it compare?
At least going by AMD's numbers, the 5000-series processors should offer somewhere around 19% more performance per clock on average compared to the 3000-series, while also featuring slightly higher clock rates. So, it's expected that performance at CPU-limited tasks should typically be over 20% faster for a given core count, though the exact performance difference may vary significantly from one application to the next.

That applies to the 5600X too, though that processor has fewer cores/threads than a 3700X or 5800X. Most applications are not going to be making heavy use of more than a handful of threads, so for most tasks, a 5600X should be faster than a 3700X, since the extra cores will be sitting mostly idle at those times. In software that can divide a task evenly between all available threads though, I suspect the 3700X might manage to be a little faster most of the time. That's because while the 5600X may offer over 20% more performance per thread, the 3700X has 33% more threads to go around.

Most software doesn't fit into that category though, as many tasks can't easily be divided like that, but some things can. For example, relevant to video editing, a CPU-based video encoder can generally split its workload evenly across many threads, as each thread can be assigned to process a different frame of the video at the same time, with minimal communication between them. Again though, unless you are encoding video all the time and are primarily concerned about cutting down encode times above all else, lots of other software will benefit more from having more performance available per core rather than more cores.

While we'll have to wait for reviews to see exactly how performance fares in different software, I suspect the 5600X will probably be a better choice than a 3700X, at least if they are priced about the same. It seems likely that the 5600X's higher performance per core should allow it to come relatively close to the 3700X at heavily-multithreaded tasks like video encoding, while being notably faster at the more common lightly-threaded tasks. The 5800X offers both the higher performance per core compared to the 3700X, as well as the higher core count compared to the 5600X, though there's a rather large price jump over those models. While the 5800X does offer 33% more cores/threads than the 5600X, it's set to cost $450, or about 50% more. And again, if we look at performance in most common software, which won't be utilizing all of its cores, it's performance should be about the same as the 5600X, probably no more than about 2% faster judging by clock rates.
 
I use two resources to evaluate cpus--passmark and cpuworld.

I only really look at a few things--single thread performance, the amount of actual cores it has, and how fast those cores connect to the memory bus. Biggest numbers across all 3 will be the fastest overall.
 
I can wait! That's a good idea. Will the 5800x be much more powerful than the 3700x? What about the Ryzen 5 5600X, how would it compare?
AMD claims 19% IPC uplift...and it's looking pretty good if these Passmark single thread performance BM numbers are to be believed. Reviews should be really interesting to see how this translates to the type of gaming performance that was Intel's last remaining category for claiming leadership.
 

lga1156_ftw

Notable
Feb 25, 2020
1,124
124
990
35
Userbenchmark is only good if you are competitive gamer and want to see what get the most fps with 1080p low settings. Im been using older gen cpu:s (westmere and lynnfield) and even when i get more fps than modern ryzen 3 and 5 in similar situations (tested VERY well) when i go to sites like passmark im getting very bad results. So it can also go the other way around.
This is true for games like Overwatch/ Fortnite and CSGO.

The numbers you see on synthetic benchmarks just doesnt convert to real life competitive gaming. The new amd 5000 series might change it we will see later this year. Ive seen it so many times when people change from 3700x to 9900k for example theyre fps boost up substantially ingame on competitive settings.
 
Last edited:
Userbenchmark is only good if you are competitive gamer and want to see what get the most fps with 1080p low settings. Im been using older gen cpu:s (westmere and lynnfield) and even when i get more fps than modern ryzen 3 and 5 in similar situations (tested VERY well) when i go to sites like passmark im getting very bad results. So it can also go the other way around.
This is true for games like Overwatch/ Fortnite and CSGO.

The numbers you see on synthetic benchmarks just doesnt convert to real life competitive gaming. The new amd 5000 series might change it we will see later this year. Ive seen it so many times when people change from 3700x to 9900k for example theyre fps boost up substantially ingame on competitive settings.
Using benchmarks for a particular application will never tell you what the particular application will do because of the different ways to optimize each application by the coders. The only way to know for a particular application is to test with that application.
 
Reactions: lga1156_ftw

lga1156_ftw

Notable
Feb 25, 2020
1,124
124
990
35
Well said.

I never had any problems using userbenchmark for example quickly check if particular system have any problems, people thrash the site too much everywhere. I agree its biased towards intel but some sites are the other way around. I might have tunnel vision since i mostly benchmarked pc:s which where used for competitive gaming ( always low graphic settings and well optimized ) then looking at the fps results they are quite accurate.

But yeah, everyone whos not competitive gamer its better to use other sites.
 

ASK THE COMMUNITY

TRENDING THREADS