NEW CPU Benchmark for Gamers and Overclockers 3-D Mark

Comixbooks

Fully [H]
Joined
Jun 7, 2008
Messages
17,316
12b98653d20459b815e0564c3a023380e84abf70_960x311.jpg
MAJOR UPDATE
POSTED
Thu, June 24
New CPU benchmarks for gamers and overclockers

We’re expanding 3DMark today by adding a set of dedicated CPU benchmarks.

The 3DMark CPU Profile introduces a new approach to CPU benchmarking. Instead of producing a single score, the 3DMark CPU Profile shows you how CPU performance scales and changes with the number of cores and threads used.

The CPU Profile has six tests, each of which uses a different number of threads. The benchmark starts by using all available threads. It then repeats using 16 threads, 8 threads, 4 threads, 2 threads, and ends with a single-threaded test.

These six tests help you benchmark and compare CPU performance for a range of threading levels. They also provide a better way to compare different CPU models by looking at the results from thread levels they have in common.

The 3DMark CPU Profile shows you how your CPU scores compare with other results from the same processor. It’s a great way to check if your CPU is performing as expected.

9bd2b222be88cccb32c9d22eb02ab1ef07cd9499.png

Six levels of CPU performance
Max-threads test
— The Max-threads score represents the full performance potential of your CPU when using all available threads, the ideal benchmark for the latest high-core-count processors.

16-threads test
— The 16-threads score is a good measure of CPU performance for computationally intensive tasks, such as digital content creation and 3D rendering, that benefit from more threads.

8-threads test
— The performance of modern DirectX 12 games usually correlates most closely with the 8-threads score.

4-threads test
and
2-threads test
— The frame rates of older games developed for DirectX 9 usually correlate most closely with the 2-threads and 4-threads scores.

1 thread test
— The 1-thread score is a fundamental measure of your processor's performance.

Benchmark and compare CPU performance with 3DMark
The 3DMark CPU Profile shows you how your CPU scores compare with other results from the same processor.

713370db379c9591027a16628f159be5dd9b005b.png

On the 3DMark CPU Profile result screen, the green bars show you how your scores compare with the best scores for your CPU. The longer the green bar, the closer your score is to the best result.

The median score, shown by the marker, shows the performance level you should expect for your CPU. In most cases, the median represents performance with stock settings. If your score is below the median, it may indicate a problem with cooling or background processes. Check the hardware monitoring chart to see how the CPU temperature changed during the run.

The distance from the median marker to the end of the bar represents the overclocking potential of the CPU. For overclockers, the 3DMark CPU Profile gives you more ways to measure the effects of your overclocking and more ways to compete for the highest scores!

Please note that these features are powered by benchmark results from 3DMark users. These insights may be unavailable for some CPU models until enough results are submitted.

Available now
The 3DMark CPU Profile is available now as a free update. Download the latest version to get started.
 
Last edited:

zamardii12

2[H]4U
Joined
Jun 6, 2014
Messages
3,093
I haven't bothered overclocking a CPU or GPU since Ryzen 1st gen came out. Maybe I'm just getting old lol, as long as it works I'm fine with it.
Same here. My last PC was the first and last PC of mine I would ever overclock. CPUs are so damn fast now that overclocking literally has no negligible difference I feel in gaming at least if it's a recently released CPU within the last year or so. I always had boot and stability issues and was tired of tweaking it. I just want to turn a PC on and game for the 30 or so minutes I have of free time anymore.... not spending my time diagnosing problems.

Yeah... I agree... just getting older and caring less about that shit. :D
 
Last edited:

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
Seems like this place is becoming (S)oftforum. Yeah the newer chips auto boost pretty far on their own. If your past overclocks had problems, that was simply because it wasn't dialed in and tested correctly. My 8700k has ran overclocked since December 2017 without a single blue screen or crash. The only issues the rig had was when I tried getting into mining and I had not dialed in my settings correctly on my gpu yet.

It's kind of a self defeating argument saying there are limited gains with an overclock. Your own scale is whatever you set it at, what is negligible? 1%? 10%? 0.1%? It's up to everyone to make that decision. Same way we have to make the decision to even upgrade, why even do that? Wasn't your system working just fine before you upgraded? Are you really doing anything all that different from a few years ago? You just want that speed bump, so new system it is, let we leave some of that on the table. The new system was 10% better than the old one, but if you overclocked it could be 13%. Why was 10% worth it but putting in some effort was too much of a mountain to climb to bring you to 13%?

The worst is the first generation you do this. You go from an overclocked cpu to a stock cpu. Maybe you really only gained 5%, because you were comparing your old system's cpu score at stock.

Yeah the gains are less than they used to be. Still worth it, good enough isn't good enough. Stay [H]ard!
 

trandoanhung1991

[H]ard|Gawd
Joined
Aug 26, 2011
Messages
1,272
more like [O]ldForum.. we carefully choose our fights now. Overclocking with the added heat is low ROI. These days with the CPU choices, its a better play to upgrade if you value time and have $.

The more we age, the more our hourly wage rises, and at some point it's better to just shell out a couple hundred bucks more for that 5-10% than spending 2 weeks to a month stability testing...
 

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
The more we age, the more our hourly wage rises, and at some point it's better to just shell out a couple hundred bucks more for that 5-10% than spending 2 weeks to a month stability testing...
I don't know if that holds true. As I've aged and gained more income, I've skipped past the lesser chips and gone for the good stuff (my sig shows a 1080 ti which was basically the top in 2017, sorry I haven't updated it with my new build yet.) I've got a 5950x on an Asus Crosshair VIII Dark Hero mobo with an evga 3090 ftw3 and 32gb of 3800mhz cl14 ram. Outside of maybe the ram, there really isn't anywhere else I could have thrown any money to get a higher performing part. To get more overclocking is the only option I have, and while it does take some time to properly test your parts for stability I think a month might be a bit of an exaggeration. Besides, while you have the new pc testing you still have your current pc, no need to rush.
 

trandoanhung1991

[H]ard|Gawd
Joined
Aug 26, 2011
Messages
1,272
I don't know if that holds true. As I've aged and gained more income, I've skipped past the lesser chips and gone for the good stuff (my sig shows a 1080 ti which was basically the top in 2017, sorry I haven't updated it with my new build yet.) I've got a 5950x on an Asus Crosshair VIII Dark Hero mobo with an evga 3090 ftw3 and 32gb of 3800mhz cl14 ram. Outside of maybe the ram, there really isn't anywhere else I could have thrown any money to get a higher performing part. To get more overclocking is the only option I have, and while it does take some time to properly test your parts for stability I think a month might be a bit of an exaggeration. Besides, while you have the new pc testing you still have your current pc, no need to rush.

In my experience CPU OC takes about a week, then GPU core clock takes about 3-5 days, GPU memory a day or so... I have yet to properly OC RAM, it's such a PITA. I've wasted a good 2 weeks getting stable RAM OCs, never managed it and just said F it.

And then every so often BIOS and driver updates cause instability, which means I have to redo pretty much all of that again. The time wastage pile up pretty quick.

And I think yours is an edge case. I don't think most people would have a banger PC stability testing while having a good enough PC to use in the meanwhile. Personally I usually creep up with each part, not replace the whole darn thing.
 

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
In my experience CPU OC takes about a week, then GPU core clock takes about 3-5 days, GPU memory a day or so... I have yet to properly OC RAM, it's such a PITA. I've wasted a good 2 weeks getting stable RAM OCs, never managed it and just said F it.

And then every so often BIOS and driver updates cause instability, which means I have to redo pretty much all of that again. The time wastage pile up pretty quick.

And I think yours is an edge case. I don't think most people would have a banger PC stability testing while having a good enough PC to use in the meanwhile. Personally I usually creep up with each part, not replace the whole darn thing.
Ok some fair points. On the bios though, I will say I never touched it on my last pc, ran the December 2017 bios and never updated it once. One area where once I get it stable I refuse to touch it, and it served me well with no blue screens throughout its life. Driver updates, especially gpu drivers, yes those can be a pain in the ass. I do fondly recall the weeks of tweaking my ddr1 tccd ram, the joy of finally getting it stable running 580mhz 2.5-3-3-7 timings, but yeah I don't really relish spending that time anymore. Back then the cpu's memory controller held you back more than the ram, but the gains were awesome.

I used to replace a gpu out of cycle, but very rarely any other parts. All too often the new gen of cpu needs a different socket, this isn't as common as it was in years past. Ram often can't be reused either, or it's so slow now you want to upgrade anyway.

A pc from five years ago is still entirely usable today. We aren't making the leaps and bounds of the past. I would argue that outside of a gpu upgrade, the piecemeal approach is more edge case now. I wanted to upgrade my server, which had a 1700x on an asus x370 prime with 16gb of 2400mhz ram. No bios update to allow zen 3 on there, so to support the 5600x I moved to I had to get a new mobo and I got faster ram too. Even when I was willing to suffer the older and inferior parts, it was not possible, on that four year old board even though it was the same socket. That said, the gf's pc has a b450 board that I tested my 5950x out in, so it is possible to make some upgrades, but just not every time.

Anyway, I think it's only ever first time builders or people who have had disasters that can't use their existing pc while they build the new one. I've been doing it this way since 2007. Just hold off on some games until you can get the better system, they won't go rotten or anything (I did this with Cyberpunk and Half Life Alyx.) Those without a banger pc then have to decide what their time is worth, as we always have. The thrill of the possible free gains, the agony of losing the silicone lottery.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,548
Comparing SMT On vs Off on TR 3960x, stock CPU bios settings:

3dCPUmark.png

For this test it is virtually the same performance. Looks like my CPU OC higher with SMT On, why that would be I don't know.
 
Joined
Nov 18, 2011
Messages
694
I haven't bothered overclocking a CPU or GPU since Ryzen 1st gen came out. Maybe I'm just getting old lol, as long as it works I'm fine with it.

Everything from the 9900K and later there is little return. Ryzen is even worse. A few hundred MHZ for hours of work to get it Prime stable.

But I still like a challenge so I went SFF. Right now I have a 5950X undervolted being cooled by a Noctua L12S in a 6L case (CCD MI-6). It’s not the fastest 5950X, but it’s been a fun challenge.
 

LigTasm

Supreme [H]ardness
Joined
Jul 29, 2011
Messages
5,151
Everything from the 9900K and later there is little return. Ryzen is even worse. A few hundred MHZ for hours of work to get it Prime stable.

But I still like a challenge so I went SFF. Right now I have a 5950X undervolted being cooled by a Noctua L12S in a 6L case (CCD MI-6). It’s not the fastest 5950X, but it’s been a fun challenge.

My desires kinda changed to just want the system to be quiet these days, but I did spend a lot of time and money on serious overclocking water cooled builds in the past. I agree SFF is a lot of fun though, I've been tempted to pick up some parts and build in one of the ITX cases I have sitting around. I have an unused 3700X that would be perfect for that build.
 
Joined
Apr 9, 2021
Messages
733
Overclocking will always be a thing because selling fully maxed out hardware is not something that Intel and AMD like doing. The binning process becomes very expensive, and the failure rates go up. They also have to assume that the user has an entry level motherboard, an entry level PSU, and entry level cooling. If you have enthusiast-grade supporting hardware, there's going to be at least some headroom. Of course, the key to overclocking will always be to capitalize on a segment in which there is no competition. Right now as of today, that's the 16-core segment. The 5950X. Intel has no real answer for it right now, and as a result AMD is happy to sell it "detuned." That's the one chip you can buy today that easily gains 20% or more from overclocking.
 
Last edited:

Criticalhitkoala

[H]ard|Gawd
Joined
Apr 17, 2015
Messages
1,982
I haven't bothered overclocking a CPU or GPU since Ryzen 1st gen came out. Maybe I'm just getting old lol, as long as it works I'm fine with it.

I stopped oc'ing because of heat dissapation to the room. Even with Central healing and a window unit, having all the machines blasting including my wifes make it either loud room but comfy temp, or hot room with lower sound. Oh and also cause I'm old.
 
Joined
Apr 9, 2021
Messages
733
I would like to run this benchmark, but I guess I'd have to buy 3DMark first. I can't think of a good reason to buy a benchmark when there are so many free ones.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
I Really dont see a lot of value in this TBH is this a generic synthetic CPUtest ?
There is "a million" CPU tests on real world code I think would be better to use.

or am i missing something ? in such case let me know
 

Bigbacon

Fully [H]
Joined
Jul 12, 2007
Messages
19,275
I haven't bothered overclocking a CPU or GPU since Ryzen 1st gen came out. Maybe I'm just getting old lol, as long as it works I'm fine with it.

same here. they do a good enough job boosting themselves and the saving power when not needed. Ryzen is the first CPU I have not overclocked.
 

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
I Really dont see a lot of value in this TBH is this a generic synthetic CPUtest ?
There is "a million" CPU tests on real world code I think would be better to use.

or am i missing something ? in such case let me know
Benefit here is the online database and being able to compare against other similar systems, and share scores through their site easily. Perhaps others do this too, I don't know. They're also probably trying to make new products to keep themselves relevant, I think its an ok move vs just staying stagnant.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
Benefit here is the online database and being able to compare against other similar systems, and share scores through their site easily. Perhaps others do this too, I don't know. They're also probably trying to make new products to keep themselves relevant, I think its an OK move vs just staying stagnant.
my problem is that a synthetic CPU test is close to useless especial considering the highly difference in CPU architecture. So despite a CPU A coming out ahead, it is still pretty likely that is it slower than CPU B, but just the benchmarks is optimized for for this giving architecture. You can compare and say OK this is better for this kind of load types because the load is synthetic and not relevant for real world load.

The ease of comparing is nice, but ease of comparing close to useless data just seem close to useless to me.
 

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
my problem is that a synthetic CPU test is close to useless especial considering the highly difference in CPU architecture. So despite a CPU A coming out ahead, it is still pretty likely that is it slower than CPU B, but just the benchmarks is optimized for for this giving architecture. You can compare and say OK this is better for this kind of load types because the load is synthetic and not relevant for real world load.

The ease of comparing is nice, but ease of comparing close to useless data just seem close to useless to me.
So what would you use instead? without standardized tests there's no metric that would ever tell us that x is better than y, for anything. At the very least here you can compare a 5950x to other 5950x's in the world, so you know its absolutely the same architecture. Sure maybe comparing it to an older amd or Intel's chips isn't 1:1 but you have to use something.
 

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,548
So what would you use instead? without standardized tests there's no metric that would ever tell us that x is better than y, for anything. At the very least here you can compare a 5950x to other 5950x's in the world, so you know its absolutely the same architecture. Sure maybe comparing it to an older amd or Intel's chips isn't 1:1 but you have to use something.
X is better than Y at what? Just because in this one test X scores higher than Y does not automatically indicate X is better than Y in something you actually use.

I find these tests useful for optimizing your own hardware, somewhat useful to troubleshoot your performance with similar system or component. Totally useless to low value in comparing two different CPUs. In this test, 48 threads is virtually the same as 24 threads on a 3960x, that does not remotely indicate the performance of my 3960x in programs rendering 3d as in like Modo.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
So what would you use instead? without standardized tests there's no metric that would ever tell us that x is better than y, for anything. At the very least here you can compare a 5950x to other 5950x's in the world, so you know its absolutely the same architecture. Sure maybe comparing it to an older amd or Intel's chips isn't 1:1 but you have to use something.
I'm not sure in which way you are using the word standardized. as in can be executed on two different platform and compare of if you are using standardized as in one test to rule them all the everyone should be using,

If it the former you can use tons of tools to get a close to realworld workload benchmark
Geekbench
Winsat
Prime95
DnetC
Lame
7zip
Winrar
Linpack
Qaac
CineBench
Blender
x.264
x.265
ECT
PNGout
- Pretty much any command line tools you use that is CPU dependant
- Any Game in very low resolution/AF/AA (Albeit with some challenges with cross platform test)

These are just on top of my head
All of these will have a real world application so the benchmark can better indicate a real effective difference in architecture


"Least here you can compare a 5950x to other 5950x's"
Correct but that is very poor bar for a benchmark and something you can do with all of the above as well. this is like the minimum ability of a benchmark

If I want to know if my I9 XXXX vs my Thread ripper YYYY is better for a encoding movies/audio a 10% better score in a synthetic test is useless as I cant be sure that actually does something in a realworld software of if it just becuase the benchmark is more code optimized for a given CPU.
However actually running real world application code and test it give you exactly the performance gain indication you are looking for.


Synthetic test on highly different architectures becomes more about code optimization for a giving architecture than, performance of the hardware and that is a bad way to go if you want to know about the hardware performance.
In real world code at least i know the difference is there, when i have measured it
What the value in being 10% higher perfomance in this benchmark if it is 5% slower in everything you use you computer for?


Or to put in short and clear: Using a synthetic test shows performance in something you don't use you computer for.
 
Last edited:

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
X is better than Y at what? Just because in this one test X scores higher than Y does not automatically indicate X is better than Y in something you actually use.

I find these tests useful for optimizing your own hardware, somewhat useful to troubleshoot your performance with similar system or component. Totally useless to low value in comparing two different CPUs. In this test, 48 threads is virtually the same as 24 threads on a 3960x, that does not remotely indicate the performance of my 3960x in programs rendering 3d as in like Modo.
Then it seems to me all your argument is I don't want to look at a test like this, I will look at tests of people running Modo, because that is the one and only program I care about. That's fine, but sometimes you want to just have a more all around comparison. This on in particular can show the relationship of thread counts compared to your own cpu, you can see how at lower thread counts your cpu boosts higher. This, not any test that has ever been developed or ever will be, will 100% say this one is always better than that one. I never said this was the one test to rule them all or anything.
I'm not sure in which way you are using the word standardized. as in can be executed on two different platform and compare of if you are using standardized as in one test to rule them all the everyone should be using,

If it the former you can use tons of tools to get a close to realworld workload benchmark
Geekbench
Winsat
Prime95
DnetC
Lame
7zip
Winrar
Linpack
Qaac
CineBench
Blender
x.264
x.265
ECT
PNGout
- Pretty much any command line tools you use that is CPU dependant
- Any Game in very low resolution/AF/AA (Albeit with some challenges with cross platform test)

These are just on top of my head
All of these will have a real world application so the benchmark can better indicate a real effective difference in architecture


"Least here you can compare a 5950x to other 5950x's"
Correct but that is very poor bar for a benchmark and something you can do with all of the above as well. this is like the minimum ability of a benchmark

If I want to know if my I9 XXXX vs my Thread ripper YYYY is better for a encoding movies/audio a 10% better score in a synthetic test is useless as I cant be sure that actually does something in a realworld software of if it just becuase the benchmark is more code optimized for a given CPU.
However actually running real world application code and test it give you exactly the performance gain indication you are looking for.


Synthetic test on highly different architectures becomes more about code optimization for a giving architecture than, performance of the hardware and that is a bad way to go if you want to know about the hardware performance.
In real world code at least i know the difference is there, when i have measured it
What the value in being 10% higher perfomance in this benchmark if it is 5% slower in everything you use you computer for?


Or to put in short and clear: Using a synthetic test shows performance in something you don't use you computer for.
I never said this was the one test to rule them all! By standardized I simply mean it's the same test, if 100 people with different configurations run the test, they can compare their results to them and see how the stack up against their own model and other models.

Yes there are tons of others available, I don't understand the point. You should always use multiple tests to get a view of how something performs. Just as basing all your decisions on how 7zip performs would make no sense to me if all I cared about was blender performance, for example, everyone is different.


Seems all that matters here is synthetic benchmarks are dumb, and they have no value so they should be deleted from the world. You see no value to this product so there's nothing more to be discussed here.
 

SvenBent

2[H]4U
Joined
Sep 13, 2008
Messages
3,308
Then it seems to me all your argument is I don't want to look at a test like this, I will look at tests of people running Modo, because that is the one and only program I care about. That's fine, but sometimes you want to just have a more all around comparison. This on in particular can show the relationship of thread counts compared to your own cpu, you can see how at lower thread counts your cpu boosts higher. This, not any test that has ever been developed or ever will be, will 100% say this one is always better than that one. I never said this was the one test to rule them all or anything.

I never said this was the one test to rule them all! By standardized I simply mean it's the same test, if 100 people with different configurations run the test, they can compare their results to them and see how the stack up against their own model and other models.

Yes there are tons of others available, I don't understand the point. You should always use multiple tests to get a view of how something performs. Just as basing all your decisions on how 7zip performs would make no sense to me if all I cared about was blender performance, for example, everyone is different.


Seems all that matters here is synthetic benchmarks are dumb, and they have no value so they should be deleted from the world. You see no value to this product so there's nothing more to be discussed here.


You didn say it was but then proceed to explain what is exactly the definition I used it for. so its a no but a yes ?

If you by now do not understand the lack of value in a benchmark that will show performance totally unrelated to any usage, that I'm not sure how to explain it any further.

Being able to uniformally compared in unimportant metrics does not make the metric better. its still unusable metric.

Your concept of all around benchmark does not exist. That is not how this works
Also its not an all around benchmark. Its a CPU benchmark

-- edit --
Its seems to me you might not understand how code optimization works and how big an impact it has and thins code is just code so somthing running code faster runs all code faster. that is not how htings worsk
Code path A might be faster on CPU Y than on X and code path B might be reversed.

A perfect examples is the old Intel compiler that would only use optimal code path ofr intel CPU's and avidng SSE on AMD cpu's depite them being able to do it a benchmark compiled with this compile would show and unfavor to intel CPU's but you dont know if you pieces of real world code works the same way so this boost is 100% uselss

Unrolling loops might be better for a CPU with strong prefecth perferomance but bad branch prediction.
but on a CPU with a strong branch prediction and bad prefetch performance it will hurt performance.

This optimization choice ( or lack thereof) will push you code to performance better on oen cpu vs the other. but again totally unrelated for usage as its synthethic code.


"You see no value to this product so there's nothing more to be discussed here."
That fine just remember that you reached out to me to debate it so so that is all on you.
I just put in my opionm that you for some reasong had a need to try to correct.
 
Last edited:

arnemetis

2[H]4U
Joined
Aug 2, 2004
Messages
3,545
You didn say it was but then proceed to explain what is exactly the definition I used it for. so its a no but a yes ?

If you by now do not understand the lack of value in a benchmark that will show performance totally unrelated to any usage, that I'm not sure how to explain it any further.

Being able to uniformally compared in unimportant metrics does not make the metric better. its still unusable metric.

Your concept of all around benchmark does not exist. That is not how this works
Also its not an all around benchmark. Its a CPU benchmark

-- edit --
Its seems to me you might not understand how code optimization works and how big an impact it has and thins code is just code so somthing running code faster runs all code faster. that is not how htings worsk
Code path A might be faster on CPU Y than on X and code path B might be reversed.

A perfect examples is the old Intel compiler that would only use optimal code path ofr intel CPU's and avidng SSE on AMD cpu's depite them being able to do it a benchmark compiled with this compile would show and unfavor to intel CPU's but you dont know if you pieces of real world code works the same way so this boost is 100% uselss

Unrolling loops might be better for a CPU with strong prefecth perferomance but bad branch prediction.
but on a CPU with a strong branch prediction and bad prefetch performance it will hurt performance.

This optimization choice ( or lack thereof) will push you code to performance better on oen cpu vs the other. but again totally unrelated for usage as its synthethic code.


"You see no value to this product so there's nothing more to be discussed here."
That fine just remember that you reached out to me to debate it so so that is all on you.
I just put in my opionm that you for some reasong had a need to try to correct.
Ok sure, let's go around again. Any metric, whether it represents a real life workload or not, can be useful. In this case, you can compare it to other same model processors. If yours is coming in way under the average, it can indicate a potential problem with yours. Or perhaps it comes in near the top, it indicates your overclock is showing meaningful gains, even against your own previous scores. Yes this will not necessarily translate to 1:1 gains in other applications, but it is a way to compare the performance of the processor to itself. Let's forget using it to compare to other models. Perhaps the repository of scores you can compare to is nothing new either, so sure let's take that off the table too.

An all around benchmark does exist, it may just not be one you approve of. I don't feel this program fits in that category, it is in fact cpu specific. Check out userbenchmark.net, it's doign this same thing by offering up these metrics between different processors, graphics cards, you name it. It's a general metric people can use without digging down and getting results specific to every single program they want to run. Since that site exists, this 3dmark cpu benchmark is rather redundant, but it's just another tool.

The optimization discussion simply spins us in circles. Even synthetic benchmarks will probably favor one side or another, every single program in existence is this way. I acknowledge code differences can make one better than another in different scenarios. So it seems in summary you believe that real life workload tests are the only useful ones, synthetics are entirely worthless, ok. So if someone is looking to upgrade, they should compare results of their specific programs (and versions! Gotta make sure everyone is using the same version of the software or the analysis is null and void as the code changes and supports newer instruction sets) in order to compare processors. Do you honestly look at dozens or hundreds of program comparisons in order to learn about cpus? Certainly one shouldn't make their decision based on how a cpu performs in winrar, and winrar alone because it represents a real world workload. But I also can't believe most people spend the time hunting down reviews of every processor and how it performs in every program they use, and then somehow take that mix of data (because of course some will be better on this processor than that one) to make a decision on what to buy? God help them if they ever change what they do and introduce a new program! Perhaps you do do all this, my argument is simply that it is unreasonable for everyone to do this and synthetics do still hold some value.

In my opinion the value of these general benchmarks is to get an approximation, a general idea of where to start. If all you do is 7zip all day, by all means look those numbers up. Is this machine dedicated to video editing only? Ok then that is your metric. Most people have just a few conditions they worry about, so they do look up those particular benchmarks in addition to general ones like these. This is very common with game comparisons for gpus, you look at games you play or want to play, you look at resolutions you use, etc. I'm sure plenty of people do things this way, they have a machine that is dedicated to a specific task, and that is all that matters. But
 

cybereality

[H]F Junkie
Joined
Mar 22, 2008
Messages
8,317
Yeah, I don't bother with 3DMark anymore. It's completely detached from reality making it basically useless.

I will read reviews, sure, I do like to see games *I actually play* or apps/code I would use. Something like compile times on Chromium or the Linux kernel (even though I compile neither) would at least give me an idea of real world performance.

Beyond that, I will just use my computer, run my own benchmarks for things that matter to me and decide for myself. The numbers don't matter.
 
  • Like
Reactions: noko
like this

noko

Supreme [H]ardness
Joined
Apr 14, 2010
Messages
6,548
Then it seems to me all your argument is I don't want to look at a test like this, I will look at tests of people running Modo, because that is the one and only program I care about. That's fine, but sometimes you want to just have a more all around comparison. This on in particular can show the relationship of thread counts compared to your own cpu, you can see how at lower thread counts your cpu boosts higher. This, not any test that has ever been developed or ever will be, will 100% say this one is always better than that one. I never said this was the one test to rule them all or anything.
. . .
Reading comprehension may help you, :D. Just kidding, I said exactly what I find beneficial or not in this benchmark and will use it:
"I find these tests useful for optimizing your own hardware, somewhat useful to troubleshoot your performance with similar system or component. Totally useless to low value in comparing two different CPUs."​
 

ThreeDee

[H]F Junkie
Joined
Sep 5, 2001
Messages
11,282
"OMG! Your ePenis is slightly larger than mine! ...darn it!"
I used to overclock everything and it's mother years ago .. I even had a Vapochill™ and got my old P4 2.4c up to 3.6ghz .. I might run this new "Benchmark" for poops and giggles though

I only tweak my RAM a little bit now (5800x, x570 Taichi, 2 x 16GB RGB HyperX Predator 3200 CAS16 stuff @ 3733 CAS16, w/fclk 1:1, 6700xt)
As stated .. I just want to turn my stuff on and game, encode some videos .. watch a movie .. whatever... and not run benchmarks

My go to stability test is just play some Hunt:Showdown .. if my memory tweaks aren't stable .. Hunt will let me know where other stress tests have given me a false sense of stability.
 
Top