Monday, July 8th 2024

AMD Ryzen 9 9900X Benchmarked in Geekbench 6, Beats Intel's Best in Single-Core Score

by
AleksandarK
 Discuss (105 Comments)
As AMD prepares to roll out its next-generation Ryzen 9000 series of CPUs based on Zen 5 architecture, we are starting to see some systems being tested by third-party OEMs and system integrators. Today, we have Geekbench 6 scores of the Ryzen 9 9900X CPU, and the 12-core, 24-thread processor that has demonstrated impressive performance gains. Boasting a base clock of 4.4 GHz and a boost clock of up to 5.6 GHz, the CPU features only 120 W TDP, a significant reduction from the previous 170 W of the previous generation. In Geekbench 6 tests, the Ryzen 9 9900X achieved a single-core score of 3,401 and a multicore score of 19,756.

These results place it ahead of Intel's current flagship Core i9-14900KS, which scored 3,189 points in single-core performance. Regarding multicore tasks, the i9-14900K scored 21,890 points, still higher than AMD's upcoming 12-core SKU. The benchmark of AMD's CPU was conducted on an ASUS ROG Crosshair X670E Gene motherboard with 32 GB of DDR5 memory. As anticipation builds for the official release, these early benchmarks suggest that AMD will deliver a compelling product that balances high performance with improved energy efficiency. The top tier models will still carry a 170 W TDP, while some high-end and middle-end SKUs get a TDP reduction like the Ryzen 7 9700X and Ryzen 5 9600X dial down to 65 W, decreased from 105 W in their previous iterations.
Sources: Geekbench v6, via Wccftech

Related News

Add your own comment

105 Comments on AMD Ryzen 9 9900X Benchmarked in Geekbench 6, Beats Intel's Best in Single-Core Score

#76
AnotherReader
Vya DomusI don't know what "mostly single threaded" is supposed to mean, python like virtually all languages has support for threading.
The most common version of Python, CPython, uses a global interpreter lock. Essentially, this means that it is single threaded.
Posted on Reply
#77
dir_d
RootinTootinPootinUh huh..

Zen 5 performs that way thanks to the beta testers for Zen 4..now, Zen 5 buyers, you beta test for Zen 6.. :roll:
What are you on about?? I really don't understand what you mean by this.... Then you are laughing, what am i missing?
Posted on Reply
#78
Vya Domus
atomsymbolBy that logic, even assembly language (such as: *.S files in Linux) has support for threading
Correct. You're assertion that python is single threaded is a nothing burger, everything in almost every language is single threaded by default until you use multiple threads (duh?).
AnotherReaderThe most common version of Python, CPython, uses a global interpreter lock. Essentially, this means that it is single threaded.
You people are aware that if you want to do any kind of heavy processing within python you usually use something like numpy which calls routines in C, which are multithreaded by the way. Everyone knows python is dog slow, multithreaded or not, you don't do any real work within python, you offload that to something much faster not written in python.

No offense you guys are seriously out of the loop, you think software is still written like it's the 90s, almost nothing that is intended to process a lot of data is purely single threaded nowadays no matter what language you're using. This myth that software is still primarily single threaded has to die, there probably isn't a single piece of software you use day to day that isn't multithreaded to some extent.
Posted on Reply
#79
atomsymbol
Vya Domus..., everything in almost every language is single threaded by default until you use multiple threads.
This post contradicts your previous post in which you claimed that "single thread performance is becoming more and more inconsequential, it's just a metric these companies keep clinging on for marketing purposes like clock speed used to be".
Posted on Reply
#80
Vya Domus
atomsymbolThis post contradicts your previous post in which claimed that
No it just doesn't lol. What are you talking about.

Most software is multithreaded, from game engines to basic things like excel or the browser that you are on right now. Some of you have been living under a rock, If you don't believe me boot within windows with just 1 core, forget things like gaming or rendering, you'll see even basic programs will run like crap. Multi core performance is critical.
Posted on Reply
#81
atomsymbol
Vya DomusNo it just doesn't lol. What are you talking about.
I am talking about the fact that what you are posting here is an obvious over-simplification and about the fact that you aren't citing any kind of scientific study which would support your claim that "single thread performance doesn't matter in 2024".
Posted on Reply
#82
Vya Domus
atomsymbol"single thread performance doesn't matter in 2024".
I didn't say it didn't matter you genius, I said it's importance is less and less significant.

Even when it comes to things like games, most people assumed single core performance was the most important thing. X3D chips showed single thread performance wasn't the limiting factor a lot of the time, memory access speed was, game engines are heavily multithreaded.
Posted on Reply
#83
atomsymbol
Vya DomusEven when it comes to things like games, most people assumed single core performance was the most important thing. X3D chips showed single thread performance wasn't the limiting factor a lot of the time, memory access speed was, game engines are heavily multithreaded.
This is just pseudo-evidence or some religious belief. Provide actual evidence. The fact is that it is very rare for gaming websites (including TechPowerUp) to measure IPC (instructions per clock) when conducting gaming benchmarks, it is very rare for gaming websites (including TechPowerUp) to measure CPU core utilization when conducting benchmarks - and it is even more rare (as in: ultra rare) for such websites to measure memory controller utilization, L3 cache miss rate or the number of task migrations between cores.
Posted on Reply
#84
AnotherReader
Vya DomusCorrect. You're assertion that python is single threaded is a nothing burger, everything in almost every language is single threaded by default until you use multiple threads (duh?).

You people are aware that if you want to do any kind of heavy processing within python you usually use something like numpy which calls routines in C, which are multithreaded by the way. Everyone knows python is dog slow, multithreaded or not, you don't do any real work within python, you offload that to something much faster not written in python.

No offense you guys are seriously out of the loop, you think software is still written like it's the 90s, almost nothing that is intended to process a lot of data is purely single threaded nowadays no matter what language you're using. This myth that software is still primarily single threaded has to die, there probably isn't a single piece of software you use day to day that isn't multithreaded to some extent.
I'm not unaware of modern programming paradigms. Let's use another common example; JavaScript is single threaded.
Posted on Reply
#85
Super Firm Tofu
Vya DomusI didn't say it didn't matter you genius, I said it's importance is less and less significant.

Even when it comes to things like games, most people assumed single core performance was the most important thing. X3D chips showed single thread performance wasn't the limiting factor a lot of the time, memory access speed was, game engines are heavily multithreaded.
First, can you stop with the personal insults/name calling? Thanks.

If single threaded performance is 'less and less significant', what, exactly, makes up multi-threaded performance? 16 faster threads are better than 16 slower threads, no?
Posted on Reply
#86
Vya Domus
atomsymbolsome religious belief.
This is cringe and you're taking about a bunch of stuff that's not relevant to the discussion.

I don't need to know what the cache miss rate is to see that games run faster on X3D chips despite the lower single thread performance, it means that wasn't the limiting factor, it's very simple.
AnotherReaderI'm not unaware of modern programming paradigms. Let's use another common example; JavaScript is single threaded.
Same story, there are ways to speed things up with multiple threads with javascript even if it's not natively supported but once again everyone knows javascript is slow and not because it doesn't easily support multithreading but because it's slow period, just like python.

This is a very bad argument, if your problem is that you are writing software in something that doesn't easily support multithreading or if said multithreading performance sucks the solution is to offload that segment of your program to something where that's not an issue and if you can't figure that out then that's on you and not on the hardware or language.
Super Firm TofuIf single threaded performance is 'less and less significant', what, exactly, makes up multi-threaded performance? 16 faster threads are better than 16 slower threads, no?
Like I pointed out above just do the following experiment, severely downclock your CPU vs have just 1 core enabled, see which scenario yields a more tolerable experience. You don't have to believe anyone or anything, just see for yourself what is more important at this moment in time.
Posted on Reply
#87
atomsymbol
Vya DomusI don't need to know what the cache miss rate is to see that games run faster on X3D chips despite the lower single thread performance, it means that wasn't the limiting factor, it's very simple.
This doesn't make sense in context of your previous claims about single-threaded performance being less and less significant over time.

I think you are unaware of the following: Single-threaded performance can be expected to increase 10-20% per CPU generation throughout the next 20 years. This is because the IPC of CPUs in year 2024 (approximately 1.75 instructions per clock in case of Zen4 CPUs, depending on application) is nowhere near the theoretical limits (something like 40 instructions per clock, depending on application) and because branch prediction in current state-of-the-art CPUs is still quite simple compared to what is theoretically possible. Zen5 is [supposedly] the 1st x86 CPU ever to be able to fetch 2 basic blocks per cycle, while Intel CPU architectures will have to go in the same direction. At some point in the near future, a high-performance CPU will be able to fetch 3 basic blocks in a single clock cycle, which will enable the CPU to execute approximately 3-4 instructions per clock in a single thread. There is correlation between [single-threaded performance] and [number of basic blocks executed per clock cycle].
Posted on Reply
#88
Vya Domus
atomsymbolThis doesn't make sense in context
It does, the facts are core counts have increased dramatically and software has become increasingly multithreaded to the point where lack of multi core support comes at a great cost in terms of performance.
atomsymbolThis is because the IPC of CPUs in year 2024 (approximately 1.75 instructions per clock in case of Zen4 CPUs, depending on application) is nowhere near the theoretical limits (something like 40 instructions per clock, depending on application) and because branch prediction in current state-of-the-art CPUs is still quite simple compared to what is theoretically possible.
What a bizarre claim, you are free to get a job at AMD/Intel/ARM etc and fix what who knows many people couldn't for decades. Branch prediction has being studied to death, no matter how elaborate the scheme is it barely beats the 50% success rate of a random guess reliably, if there was a way to improve this significantly it would have been done by now. One of the best strategies is to simply pick the branch that was correct previously, hardly anything beats that at the cost of significant power/silicon.

Extrapolating performance when it doesn't scale linearly is a rookie mistake.
Posted on Reply
#89
atomsymbol
Vya DomusIt does, the facts are core counts have increased dramatically and software has become increasingly multithreaded to the point where lack of multi core support comes at a great cost in terms of performance.
You are predicting that the number of cores in notebook/desktop "gaming" CPUs will be increasing past 16-24 cores. I am not making such a prediction - because there is a good chance that such prediction might turn out to be false.
Vya DomusWhat a bizarre claim, you are free to get a job at AMD/Intel/ARM etc and fix what who knows many people couldn't for decades.
Your response assumes that AMD/ARM/Apple/Intel is completely unaware of what I wrote in my previous post. This is a false assumption. They are aware.
Vya DomusBranch prediction has being studied to death, no matter how elaborate the scheme is they barely beat the 50% of random guess reliably,
OK, now I see what the base problem here is: you don't even know what branch prediction is!
Vya Domusif there was a way to improve this significantly it would have been done by now.
They are in fact doing it. You are just unaware of them doing it.
Vya DomusExtrapolating performance when it doesn't scale linearly is a rookie mistake.
Written by a person who doesn't even know what branch prediction is.
Posted on Reply
#90
Vya Domus
atomsymbolOK, now I see what the base problem here is: you don't even know what branch prediction is!
You're a total clown, I am done talking to you.
Posted on Reply
#91
atomsymbol
Vya DomusYou're a total clown, I am done talking to you.
It will take 10-20 years for you to realize that you were wrong.
Posted on Reply
#92
notanin
I missed the point of discussion here.

Both single core and multicore performance are important. Your mileage will vary (and you generally know what you need in advance!).
Single core performance is no longer the king (but again, could be very important for your usage).

Threading was around for quite some time for developers to pick it up and start using it to a certain degree. You can't do any processing without it, at least in sensible way.

X3D indeed a good example on how higher frequencies are not the only thing which affects real performance.
Posted on Reply
#93
AnotherReader
Vya DomusThis is cringe and you're taking about a bunch of stuff that's not relevant to the discussion.

I don't need to know what the cache miss rate is to see that games run faster on X3D chips despite the lower single thread performance, it means that wasn't the limiting factor, it's very simple.

Same story, there are ways to speed things up with multiple threads with javascript even if it's not natively supported but once again everyone knows javascript is slow and not because it doesn't easily support multithreading but because it's slow period, just like python.

This is a very bad argument, if your problem is that you are writing software in something that doesn't easily support multithreading or if said multithreading performance sucks the solution is to offload that segment of your program to something where that's not an issue and if you can't figure that out then that's on you and not on the hardware or language.


Like I pointed out above just do the following experiment, severely downclock your CPU vs have just 1 core enabled, see which scenario yields a more tolerable experience. You don't have to believe anyone or anything, just see for yourself what is more important at this moment in time.
I agree that there are ways around that issue, and I'm not arguing for naive coding in JavaScript. Perhaps you would prefer to redefine the term as "per core performance" which is influenced by single threaded performance and TDP. Parallelization has its limits; not all algorithms have a parallel equivalent.
Posted on Reply
#94
atomsymbol
notaninI missed the point of discussion here.
One important point in this discussion is that in order to be able to make statements like "single-threaded performance is becoming less and less important" the person making the statement should first understand what branch prediction is and have at least some idea about how branch prediction is implemented in a CPU. The efficiency of branch predictions, and the width of the branch predictor (that is: the number of branch instructions predicted in 1 clock cycle), are among the most important factors determining single-threaded performance.
Posted on Reply
#95
Vya Domus
AnotherReadernot all algorithms have a parallel equivalent.
It's very rare that no portion of a non trivial piece of software can be multithreaded or vectorized. Software isn't just "let's write 1 algorithm, oops, can't be parallelized time to pack it up".
Posted on Reply
#96
atomsymbol
Vya DomusIt's very rare that no portion of a non trivial piece of software can be multithreaded or vectorized. Software isn't just "let's write 1 algorithm, oops, can't be parallelized time to pack it up".
You forgot to account for the fact that, in a non-negligible number of cases, if part of program is converted to a multi-threaded version then the multi-threaded version runs slower than the single-threaded version. In such a case, it actually ends up being "oops, can't be parallelized - time to pack it up".
Posted on Reply
#97
WhenMusicAttacks
fevgatos2 posts ago you said the most important metric for consumers was ST. Now you are saying it's gaming performance....make up your mind bud.
The DIY market is not the average consumer, they are mostly gamer that care about a specific kind of single core performance that is related to cache size. The way you twist what i write mixing and matching is similar to how you mix and match tdp, power limit and wall power measurement.
Vya DomusI didn't say it didn't matter you genius, I said it's importance is less and less significant.

Even when it comes to things like games, most people assumed single core performance was the most important thing. X3D chips showed single thread performance wasn't the limiting factor a lot of the time, memory access speed was, game engines are heavily multithreaded.
Not more than 8 cores, otherwise the 7950x3d would kill the 7800x3d and the 5600x3d would be 25% slower than the 5800x3d.
Today, 6 core is still more than enough for consumer applications, i can tune memory on 12400f and 12500 to match gaming fps of i9 cpus, because the latter FPS advantage is mostly because of its larger cache.
notaninX3D indeed a good example on how higher frequencies are not the only thing which affects real performance.
No, it shows that memory access can be the bottleneck in many user case, and that the gaming advantage of multicore cpus was mostly because of their larger cache andn ot the number of threads. Wich is also evident when you see that it always scaled for intel (that made monolithic i9s) while it didn't for AMD that made chiplets / ccds / ccx structures that made 16 cores perform worst in games despite the larger cache.
Posted on Reply
#98
Minus Infinity
LostSwede, please turn off comments for any cpu story. This is tedious in the extreme. No, I don't have to read the comments, but some people have good insights that I enjoy. Just not in these articles, where it turns into a shit fight. It's like I'm on that joke site wccftech.
Posted on Reply
#99
RootinTootinPootin
dir_dWhat are you on about?? I really don't understand what you mean by this.... Then you are laughing, what am i missing?
simply dumbfounded on every AMD release its always a hype at things that AMD innovated, when purely its a mere result of users/consumers beta testing the platform for them, and fixes them bubu's on the next product iteration.
Posted on Reply
Add your own comment
Nov 21st, 2024 09:19 EST change timezone

Latest GPU Drivers

New Forum Posts

Popular Reviews

Controversial News Posts