Intel Core i3 vs. Core i5 vs. Core i7: What do you get by spending more?

[Jad Chaar]

Great article. I was thinking of doing a budget build this summer and was wondering if I should go for an i3 or an i5. I may go for the i3 after looking at this! Thanks Steve.

 

[Mirek]

In Adobe Photoshop CC benchmark results there's an error:
i7 is 34% FASTER than i5 - not 55% as you state.
On the other hand in the same benchmark when comparison is reversed:
i5 is 55% SLOWER than i7
You guys at TECHSPOT should really learn to do the PERCENTAGE DIFFS math, when comparing values, really !!! Otherwise, the marks for your competence won't shine.

Just to clarify my original comment:

25.3 (i5) - 33% = 16.7 (i7) Faster

16.7 (i7) + 52% = 25.3 (i5) Slower

So it's not even 55% but mere 52%
Question is, where did the 55% come from ?

Now I know, why you'd gone completely wrong when talking percentages. You assume - erroneously that percentage value to illustrate the difference between two numericals is the same regardless wether you compare BIGGER to SMALLER or vice versa.
It is not ! Hence your error in the verbose.
You repeat the same error when referring to i5 which is only 29% FASTER than i3 not 42%. But i3 on the other hand is 42% SLOWER than i5
We can assume that the same error shows in the verbose throughout entire article as soon as percentages are mentioned.
You should really take care to verbalize results correctly.

 

[Steve]

In Adobe Photoshop CC benchmark results there's an error:

i7 is 34% FASTER than i5 - not 55% as you state.

On the other hand in the same benchmark when comparison is reversed:

i5 is 55% SLOWER than i7

You guys at TECHSPOT should really learn to do the PERCENTAGE DIFFS math, when comparing values, really !!! Otherwise, the marks for your competence won't shine.

Thanks for the advice but you’re not right, you’re not wrong either you just don’t seem to fully understand the percentages either.

Just to clarify my original comment:

25.3 (i5) - 33% = 16.7 (i7) Faster

16.7 (i7) + 52% = 25.3 (i5) Slower

So it's not even 55% but mere 52%

Question is, where did the 55% come from ?

Originally I had written the Core i7 is 34% faster than the Core i5, it was later updated.

Anyway you are just moving around the to and from.

The Core i7 is both 34% faster and 51% faster depending on how you look at it/word it.

The Core i7 completed the task in 34% ‘less time’.

The Core i7 offered 51% more ‘processing power’ (to complete the task)

Yes the wording is important and we won't make the mistake of just saying 'faster' in the future.

 

[amstech]

Phenom beats the i5 easily on modern platforms. Intel might be stronger on single threaded action, but on multithread and memory intensive programs (which most of the programs these days are) amd is positively ahead.

Lol.

 

[Mirek]

In Adobe Photoshop CC benchmark results there's an error:

i7 is 34% FASTER than i5 - not 55% as you state.

On the other hand in the same benchmark when comparison is reversed:

i5 is 55% SLOWER than i7

You guys at TECHSPOT should really learn to do the PERCENTAGE DIFFS math, when comparing values, really !!! Otherwise, the marks for your competence won't shine.

Thanks for the advice but you’re not right, you’re not wrong either you just don’t seem to fully understand the percentages either.

Just to clarify my original comment:

25.3 (i5) - 33% = 16.7 (i7) Faster

16.7 (i7) + 52% = 25.3 (i5) Slower

So it's not even 55% but mere 52%

Question is, where did the 55% come from ?

Originally I had written the Core i7 is 34% faster than the Core i5, it was later updated.

Anyway you are just moving around the to and from.

The Core i7 is both 34% faster and 51% faster depending on how you look at it/word it.

The Core i7 completed the task in 34% ‘less time’.

The Core i7 offered 51% more ‘processing power’ (to complete the task)

Yes the wording is important and we won't make the mistake of just saying 'faster' in the future.

Look, Steve. Your explanation that I use vague or imprecise wording in pointing out your errors doesn't hold water. The bars on the diagrams don't show processing power. They show TIMES to complete specific tasks. That's why using adjectives FASTER or SLOWER is perfecly adequate for comparison description. The i7 cannot be both 33% and 52% faster than i5, for the same benchmark - it defies logic. It has to do with math not wording. If you want the difference to be reflected in percentages you just have to decide wether you want to relate the slower cpu to the faster or the faster to the slower. Both calculating methods are valid but the percentages will differ depending on your preference. Once you've chosen comparison method, you have to be consistent throughout the article. Your error was to attribute percentage of second method to the first, as I showed in my equations.

 

[Steve]

Look, Steve. Your explanation that I use vague or imprecise wording in pointing out your errors doesn't hold water. The bars on the diagrams don't show processing power. They show TIMES to complete specific tasks. That's why using adjectives FASTER or SLOWER is perfecly adequate for comparison description. The i7 cannot be both 33% and 52% faster than i5, for the same benchmark - it defies logic. It has to do with math not wording. If you want the difference to be reflected in percentages you just have to decide wether you want to relate the slower cpu to the faster or the faster to the slower. Both calculating methods are valid but the percentages will differ depending on your preference. Once you've chosen comparison method, you have to be consistent throughout the article. Your error was to attribute percentage of second method to the first, as I showed in my equations.

It doesn’t defy logic it just depends how you are determining it to be faster. Personally I agree with you, the graph is show in time and therefore the percentage should be based on time to completion (this is how the article was originally published and all my previous work).

However that isn’t to say doing it the other way is wrong, the math still works. When comparing the data from the other ‘high is better’ tests those percentages also work better when taking the average of all tests conducted.

A small group of readers had a problem with this and for whatever reason it was changed, we are still discussing this issue internally.

If you look at any of my past work I have always done it based on time...
https://www.techspot.com/review/943-best-value-desktop-cpu/

 

[Scorpus]

In Adobe Photoshop CC benchmark results there's an error:

i7 is 34% FASTER than i5 - not 55% as you state.

On the other hand in the same benchmark when comparison is reversed:

i5 is 55% SLOWER than i7

You guys at TECHSPOT should really learn to do the PERCENTAGE DIFFS math, when comparing values, really !!! Otherwise, the marks for your competence won't shine.

Thanks for the advice but you’re not right, you’re not wrong either you just don’t seem to fully understand the percentages either.

Just to clarify my original comment:

25.3 (i5) - 33% = 16.7 (i7) Faster

16.7 (i7) + 52% = 25.3 (i5) Slower

So it's not even 55% but mere 52%

Question is, where did the 55% come from ?

Originally I had written the Core i7 is 34% faster than the Core i5, it was later updated.

Anyway you are just moving around the to and from.

The Core i7 is both 34% faster and 51% faster depending on how you look at it/word it.

The Core i7 completed the task in 34% ‘less time’.

The Core i7 offered 51% more ‘processing power’ (to complete the task)

Yes the wording is important and we won't make the mistake of just saying 'faster' in the future.

Look, Steve. Your explanation that I use vague or imprecise wording in pointing out your errors doesn't hold water. The bars on the diagrams don't show processing power. They show TIMES to complete specific tasks. That's why using adjectives FASTER or SLOWER is perfecly adequate for comparison description. The i7 cannot be both 33% and 52% faster than i5, for the same benchmark - it defies logic. It has to do with math not wording. If you want the difference to be reflected in percentages you just have to decide wether you want to relate the slower cpu to the faster or the faster to the slower. Both calculating methods are valid but the percentages will differ depending on your preference. Once you've chosen comparison method, you have to be consistent throughout the article. Your error was to attribute percentage of second method to the first, as I showed in my equations.

The raw values for both CPUs is 16.7s for the 4790 and 25.3s for the 4690. Therefore the 4790 completed the task 51.497% faster than the 4690. If the 4790 completed the task in 6.325s it would be 300% faster, or four times as fast. This makes sense, as the completion time is one quarter of the 4690. The maths used to find both percentages is the same, and it also works for other values as expected.

The 4790 also completed the task in 33.992% less time than the 4690. Using the same example, if the 4790 completed the task in 6.325s, it would have taken 75% less time to complete the task. Again this is intuitive as the CPU completed the task in 1/4 the time.

In this case, for a "lower is better" scenario, a percentage faster is not equivalent to completing the task in a percentage less time. Therefore you cannot reduce the time of the slower CPU by the percentage faster the faster CPU is.

You can come to the conclusion that the 4790 is 51.497% faster by also thinking of a fixed-time workload. If the 4690 can complete 10 tasks in 25.3 seconds, and the 4790 is 51.497% faster, we can expect it to complete 15.15 tasks in 25.3 seconds. However as the benchmark is actually fixed-task, not fixed-time, we have to normalize both sides to the same amount of tasks, not the same length of time. Therefore the 4790 will complete 10 tasks in 16.69 seconds.

 

[melkiik]

Lol.

Oh dear. To make it clear, compairing 2500k, not any i5..
Well guess I have to live with the original statement

 

[Steve]

Oh dear. To make it clear, compairing 2500k, not any i5..
Well guess I have to live with the original statement

Yeah I think that is why ‘amstech’ found your comment so funny. The Core i5-2500K is still a beast today and certainly lays waste to the Phenom II X6 1100T in virtually every scenario, gaming or otherwise.

If you knew what we know today you would be mad to invest in a Phenom II X6 over a Core i5 back in early 2011.

In most modern games the Core i5-2500K can be found delivering similar performance to the FX-9590 and certainly the FX-8350.

Here is our 2500K review from 2011...

https://www.techspot.com/review/353-intel-sandy-bridge-corei5-2500k-corei7-2600k/page13.html

 

[Mirek]

The raw values for both CPUs is 16.7s for the 4790 and 25.3s for the 4690. Therefore the 4790 completed the task 51.497% faster than the 4690. If the 4790 completed the task in 6.325s it would be 300% faster, or four times as fast. This makes sense, as the completion time is one quarter of the 4690. The maths used to find both percentages is the same, and it also works for other values as expected.

The 4790 also completed the task in 33.992% less time than the 4690. Using the same example, if the 4790 completed the task in 6.325s, it would have taken 75% less time to complete the task. Again this is intuitive as the CPU completed the task in 1/4 the time.

In this case, for a "lower is better" scenario, a percentage faster is not equivalent to completing the task in a percentage less time. Therefore you cannot reduce the time of the slower CPU by the percentage faster the faster CPU is.

You can come to the conclusion that the 4790 is 51.497% faster by also thinking of a fixed-time workload. If the 4690 can complete 10 tasks in 25.3 seconds, and the 4790 is 51.497% faster, we can expect it to complete 15.15 tasks in 25.3 seconds. However as the benchmark is actually fixed-task, not fixed-time, we have to normalize both sides to the same amount of tasks, not the same length of time. Therefore the 4790 will complete 10 tasks in 16.69 seconds.

Scorpus,
When we compare two time values 16.7s and 25.3s and we want to express the difference in percentage values, and we are interested in what's FASTER and by how many percent, then we have to agree first that:
1. Lesser time means faster cpu for the benchmark
2. 16.7s is faster than 25.3s (for the same benchmark)
3. The time of 25.3s represents the percentage base of
100% to which we compare the lesser time of 16.7s.
4. The 4790 needed 8.6s less time (faster) than the
4690 to complete the benchmark
5. 8.6s is 34% of 25.3s (100%)
6. Meaning that the 4790 needed 34% less time than
the 4690 to complete the same benchmark, I.e was
34% faster than the 4690
7. Conversly the 4690 is 52% slower in time than the
4790, when 16.7s is the base of 100% and the time
of 25.3s represents a 52% increase I.e. slower

That's the correct way to mathematically relate smaller to bigger value (or conversely bigger to smaller) as percentage difference (decrease or increase) between the two. Ask a mathematician for confirmation.

 

[Cryio]

In most modern games the Core i5-2500K can be found delivering similar performance to the FX-9590 and certainly the FX-8350.

Here is our 2500K review from 2011...

https://www.techspot.com/review/353-intel-sandy-bridge-corei5-2500k-corei7-2600k/page13.html

I hated every reviewer who used Starcraft 2, Fallout 3 and Skyrim to point out how weak AMD CPUs were in games that knew to use only 2 freakin' CPU threads. In most proper threaded games (or at least games not so totally limited exclusively by the CPU AND limited to just 2 threads) there was nothing close to the sort of 30-40% gulf in performance between the Phenom II and i5s at the time, like we see in this Starcraft benchmark.

At the time of release, a Phenom II x4 950 or later an x6 1090T were great value for the money for the sort of performance/buck you got when these CPUs were overclocked to 3.8/4 GHz, especially since these were cheaper than Nehalem i5s.

 

[Steve]

I hated every reviewer who used Starcraft 2, Fallout 3 and Skyrim to point out how weak AMD CPUs were in games that knew to use only 2 freakin' CPU threads. In most proper threaded games (or at least games not so totally limited exclusively by the CPU AND limited to just 2 threads) there was nothing close to the sort of 30-40% gulf in performance between the Phenom II and i5s at the time, like we see in this Starcraft benchmark.

At the time of release, a Phenom II x4 950 or later an x6 1090T were great value for the money for the sort of performance/buck you got when these CPUs were overclocked to 3.8/4 GHz, especially since these were cheaper than Nehalem i5s.

Hate all you want, you are still in denial.

Are you saying if I compare the 2500K against the 1100T today in modern game that utilize 4 or more threads the 2500K isn’t going to crush the 1100T?

Also the SC2 benchmark shows that the 1100T has similar core efficiency to the older and lower clocked Core i5 750, so it is a valid and interesting test.

 

[melkiik]

Yeah I think that is why ‘amstech’ found your comment so funny. The Core i5-2500K is still a beast today and certainly lays waste to the Phenom II X6 1100T in virtually every scenario, gaming or otherwise.

If you knew what we know today you would be mad to invest in a Phenom II X6 over a Core i5 back in early 2011.

In most modern games the Core i5-2500K can be found delivering similar performance to the FX-9590 and certainly the FX-8350.

Here is our 2500K review from 2011...

https://www.techspot.com/review/353-intel-sandy-bridge-corei5-2500k-corei7-2600k/page13.html

These charts prove that on paper it looks good. What about the experience itself? Ofcourse you might be right about the performance per core, but in an earlier comment I pointed out that these days as the games and programs can utilize more than the 1 or 2 cores -> phenom is guaranteed to provide more power. And to backup my fellow AMD user Cryio, especially when overcloked to 3,8 ~4,0ghz, it should perform better than the stock 2500k.
Taking account of the costs, that it takes to build A overclokable AMD platform versus overclockable intel platform, I think the AMD is way ahead. Not maybe with the latest processor series, but at that time we still had good competition.

Steve, I could run the benchmarks at the workshop, would you like me to?
for example 1100t clocked at 3,8 vs stock 2500k, on MODERN titles.

btw starcraft is and intel game
http://en.intelextrememasters.com/season9/worldchampionship/

 

[melkiik]

I hated every reviewer who used Starcraft 2, Fallout 3 and Skyrim to point out how weak AMD CPUs were in games that knew to use only 2 freakin' CPU threads. In most proper threaded games (or at least games not so totally limited exclusively by the CPU AND limited to just 2 threads) there was nothing close to the sort of 30-40% gulf in performance between the Phenom II and i5s at the time, like we see in this Starcraft benchmark.

At the time of release, a Phenom II x4 950 or later an x6 1090T were great value for the money for the sort of performance/buck you got when these CPUs were overclocked to 3.8/4 GHz, especially since these were cheaper than Nehalem i5s.

And actually when you take a AMD friendly game (like starcraft is for intel) you get equal scores even on the stock speeds. This never ending battle:

https://www.techspot.com/review/458-battlefield-3-performance/page7.html

Pretty good that the title is even about the same age as SCii.

 

[Steve]

These charts prove that on paper it looks good. What about the experience itself?

That’s not what the term on paper means these aren't paper specs, rather reality. The Core i5-2500K beats the Phenom II X6 1100T in virtually every test because it is faster and yes it feels faster (because it is).

Ofcourse you might be right about the performance per core, but in an earlier comment I pointed out that these days as the games and programs can utilize more than the 1 or 2 cores -> phenom is guaranteed to provide more power.

That’s a pretty rubbish guarantee. You keep telling us how the Phenom II X6 1100T is much more powerful than the Core i5-2500K in modern games, can we have some proof?

Here are a few recently tested games by gamegpu.ru and they included the Core i5-2500K and the FX-8350/FX-9590 (I hope we can agree that these AMD processors are faster than the Phenom II X6 1100T).

So if the Core i5-2500K is smoking the FX-8350 and sometimes even the FX-9590 in modern games what do you suppose it will do to the Phenom II X6 1100T?

Virtually every game tested in the last 2 years shows the game thing so you will have a hard time fishing for tests that show otherwise.

And to backup my fellow AMD user Cryio, especially when overcloked to 3,8 ~4,0ghz, it should perform better than the stock 2500k.

Taking account of the costs, that it takes to build A overclokable AMD platform versus overclockable intel platform, I think the AMD is way ahead.

What a load of fan boy rubbish that is. Why can you overclock the Phenom II X6 1100T but not the Core i5-2500K? The Core i5-2500K has no trouble reaching 4.5GHz, it is a considerably better overclocker than the Phenom II X6 1100T.

You have to be joking about the cost argument. I wouldn’t argue one way or the other, they are both about the same. That said the 2500K does much better using an affordable cooler than the 1100T does.

Steve, I could run the benchmarks at the workshop, would you like me to?

for example 1100t clocked at 3,8 vs stock 2500k, on MODERN titles.

btw starcraft is and intel game

http://en.intelextrememasters.com/season9/worldchampionship/

Why would I want you to do that? I can do exactly the same thing and I already know the results. If you want to try and make your point by testing the Phenom II X6 1100T clocked at 3.8GHz vs. the 2550K at stock then no I don’t want to see your results.

Ohh my mistake, StarCraft II is clearly an Intel game

And actually when you take a AMD friendly game (like starcraft is for intel) you get equal scores even on the stock speeds. This never ending battle:

https://www.techspot.com/review/458-battlefield-3-performance/page7.html

Pretty good that the title is even about the same age as SCii.

Why not post a test that is clearly heavily GPU bound :S LOL

 

[Giody]

We have done plenty of AMD vs. Intel reviews in the past, this one was titled

"Intel Core i3 vs. Core i5 vs. Core i7: What do you get by spending more?"

For a reason



It's been done

https://www.techspot.com/review/943-best-value-desktop-cpu/
https://www.techspot.com/review/816-intel-haswell-refresh-and-z97-platform/
https://www.techspot.com/review/875-intel-core-i7-5960x-haswell-e/
https://www.techspot.com/review/837-intel-core-i7-4790k-devils-canyon/

nice ty but still missing a thing
amd vs intel cpu at dx12 star swarm

 

[melkiik]

That’s not what the term on paper means these aren't paper specs, rather reality. The Core i5-2500K beats the Phenom II X6 1100T in virtually every test because it is faster and yes it feels faster (because it is).



That’s a pretty rubbish guarantee. You keep telling us how the Phenom II X6 1100T is much more powerful than the Core i5-2500K in modern games, can we have some proof?

Here are a few recently tested games by gamegpu.ru and they included the Core i5-2500K and the FX-8350/FX-9590 (I hope we can agree that these AMD processors are faster than the Phenom II X6 1100T).

So if the Core i5-2500K is smoking the FX-8350 and sometimes even the FX-9590 in modern games what do you suppose it will do to the Phenom II X6 1100T?

Virtually every game tested in the last 2 years shows the game thing so you will have a hard time fishing for tests that show otherwise.



What a load of fan boy rubbish that is. Why can you overclock the Phenom II X6 1100T but not the Core i5-2500K? The Core i5-2500K has no trouble reaching 4.5GHz, it is a considerably better overclocker than the Phenom II X6 1100T.

You have to be joking about the cost argument. I wouldn’t argue one way or the other, they are both about the same. That said the 2500K does much better using an affordable cooler than the 1100T does.



Why would I want you to do that? I can do exactly the same thing and I already know the results. If you want to try and make your point by testing the Phenom II X6 1100T clocked at 3.8GHz vs. the 2550K at stock then no I don’t want to see your results.

Ohh my mistake, StarCraft II is clearly an Intel game



Why not post a test that is clearly heavily GPU bound :S LOL

Oh dear. I've lost. You guys are serious with this intel, do they pay you marketing money?

 

[SuperVeloce]

You really are convinced, aren't you? The only AMD left in my family is my old overclocked 8320 given to my brother for some rendering jobs. And even then my haswell i7 is much faster even on this front

 

[Cryio]

That’s not what the term on paper means these aren't paper specs, rather reality. The Core i5-2500K beats the Phenom II X6 1100T in virtually every test because it is faster and yes it feels faster (because it is).



That’s a pretty rubbish guarantee. You keep telling us how the Phenom II X6 1100T is much more powerful than the Core i5-2500K in modern games, can we have some proof?

Here are a few recently tested games by gamegpu.ru and they included the Core i5-2500K and the FX-8350/FX-9590 (I hope we can agree that these AMD processors are faster than the Phenom II X6 1100T).

So if the Core i5-2500K is smoking the FX-8350 and sometimes even the FX-9590 in modern games what do you suppose it will do to the Phenom II X6 1100T?

Virtually every game tested in the last 2 years shows the game thing so you will have a hard time fishing for tests that show otherwise.

Since Windows 7 implemented Core Parking, I think I'll say this 'till the end of time. *No reviewer* takes into consideration Core Parking when they are benchmarking AMD CPUs. And I don't mean disabling the Core Parking function in Windows 7/8.1 with a registry hack. No, I mean start the game, Alt-Tab into Task Manager, change Core Affinity to 2 threads, Alt-Tab back into the game (the performance should've dipped significantly) then Alt-Tab into Task Manager to set the game's affinity back too all available threads. Sometimes the performance boosts are pretty massive just because of this simple trick.

This trick of course works bests on FX 6xxx and up. Quad Core CPUs don't benefit much from this. Even for an i7, this trick does wonders in Crysis 3 for example.

I'm not saying Haswel isn't faster than Vishera at same or lower clocks. It simply is. But in no way is an i5 at 3.4 GHz 40% faster than an FX 9590 at 4.7 Ghz in a properly multithreaded game like Attila (or it should be properly mulithreaded).

 

[Steve]

Since Windows 7 implemented Core Parking, I think I'll say this 'till the end of time. *No reviewer* takes into consideration Core Parking when they are benchmarking AMD CPUs. And I don't mean disabling the Core Parking function in Windows 7/8.1 with a registry hack. No, I mean start the game, Alt-Tab into Task Manager, change Core Affinity to 2 threads, Alt-Tab back into the game (the performance should've dipped significantly) then Alt-Tab into Task Manager to set the game's affinity back too all available threads. Sometimes the performance boosts are pretty massive just because of this simple trick.

This trick of course works bests on FX 6xxx and up. Quad Core CPUs don't benefit much from this. Even for an i7, this trick does wonders in Crysis 3 for example.

I'm not saying Haswel isn't faster than Vishera at same or lower clocks. It simply is. But in no way is an i5 at 3.4 GHz 40% faster than an FX 9590 at 4.7 Ghz in a properly multithreaded game like Attila (or it should be properly mulithreaded).

I also find it strange that no reviewer as you put it looks into this since you say it makes such a difference. Hell you would think AMD would be making more noise about this as well, perhaps even developing a program or solution for the work around. Complete madness!

 

[Cryio]

I also find it strange that no reviewer as you put it looks into this since you say it makes such a difference. Hell you would think AMD would be making more noise about this as well, perhaps even developing a program or solution for the work around. Complete madness!

All joking aside, I can guarantee this makes a difference in my PC. Regardless of the game I play, I can for example always get 5-7 fps more just by messing with the game's affinity. Which doesn't make any sense from my point of view, but the real world improvement is there.

Owner of a 560 Ti 1 GB and an FX 6300 clocked at 4.5 GHz, for reference.

 

[SuperVeloce]

Whoa, most games from last two years are heavily gpu bottlenecked on your set-up and you get ~5fps more?

 

[20 goto 10]

Detailed and accurate, but the scope of the games tested is ridiculous, infuriating.

They are all of the same type - first person games have almost always been entirely dependant on gpu power.

In fact, this seems to be true of almost anything designed to run on console as well.

And it's also recieved wisdom that single thread performance heavily outweighs overall CPU capability.

However, GTA V and other upcoming multi-platform games will require quad-core, and as a whole all games are slowly becoming more demanding of cpu.

And it seems that the recent complicated RTS (Total War, etc), and high-end sims (Assetto Corsa, X-plane) are heavily CPU dependent, and benefit from extra cores and threads. MMO does benefit, but I'm not sure how taxing the popular games are. Spot the theme? Simple game mechanics and npc behaviour doesn't tax a CPU. Complicated mechanics and AI absolutely do.

Example - the size of the grid of AI cars in Assetto Corsa scales almost exactly with overall cpu score. yet, if we disable the stupid graphics post-processing, at 1080p, maximum details and 4*msaa, fps all but identical from a 750ti to a 980.

Please, please do a more exhaustive test. It's what the market needs to know, rather than this restatement of the obvious. Like the motherboard group tests that show less than 1fps difference in a two year old game over twenty different boards. Sheesh.

Personally, I am looking RTS and Sims. Am toying with a 4790k, or waiting for Skylake. The 4790k will do for a few years, but I ought to at least have a look at the Skylake offerings. I will either settle for a 960 for now (prices falling nicely, and will be fine for 1080p), or, if I have a good summer a fast 970 (some of which approach the stock 980 in many tests) and a nice monitor.

 

[Steve]

Detailed and accurate, but the scope of the games tested is ridiculous, infuriating.

They are all of the same type - first person games have almost always been entirely dependant on gpu power.

In fact, this seems to be true of almost anything designed to run on console as well.

And it's also recieved wisdom that single thread performance heavily outweighs overall CPU capability.

However, GTA V and other upcoming multi-platform games will require quad-core, and as a whole all games are slowly becoming more demanding of cpu.

And it seems that the recent complicated RTS (Total War, etc), and high-end sims (Assetto Corsa, X-plane) are heavily CPU dependent, and benefit from extra cores and threads. MMO does benefit, but I'm not sure how taxing the popular games are. Spot the theme? Simple game mechanics and npc behaviour doesn't tax a CPU. Complicated mechanics and AI absolutely do.

Example - the size of the grid of AI cars in Assetto Corsa scales almost exactly with overall cpu score. yet, if we disable the stupid graphics post-processing, at 1080p, maximum details and 4*msaa, fps all but identical from a 750ti to a 980.

Please, please do a more exhaustive test. It's what the market needs to know, rather than this restatement of the obvious. Like the motherboard group tests that show less than 1fps difference in a two year old game over twenty different boards. Sheesh.

Personally, I am looking RTS and Sims. Am toying with a 4790k, or waiting for Skylake. The 4790k will do for a few years, but I ought to at least have a look at the Skylake offerings. I will either settle for a 960 for now (prices falling nicely, and will be fine for 1080p), or, if I have a good summer a fast 970 (some of which approach the stock 980 in many tests) and a nice monitor.

We have already tested GTA V and scaling between the Core i3 and Core i5 were similar to what we showed in this review when testing Metro Redux.

The games selected for testing in this article are not ridiculous. We picked eight very popular modern games. Obviously we can't test MMO games and most RTS games while being CPU dependent don't take advantage of more than 4 threads and therefore don't heavily favor the Core i5 over the Core i3.

Fact is this article is as you said 'detailed and accurate'. Our conclusion is accurate and I have never seen any evidence to suggest otherwise.

 

[20 goto 10]

We have already tested GTA V and scaling between the Core i3 and Core i5 were similar to what we showed in this review when testing Metro Redux.

Fact is this article is as you said 'detailed and accurate'. Our conclusion is accurate and I have never seen any evidence to suggest otherwise.

Maybe ridiculous is the wrong word, but I absolutely stand by the criticism that all the games are essentially of the same type.
The scope is 1st person / close up third person AAA titles with relatively few NPC's and very simple mechanics. Console games - further than that, I think I'm right in saying all but one game is available for the last generation. It seems that to me that the limitated CPU power available to developers working on those systems is inherently likely to mean they don't tax PC cpu's at all.

It's interesting that in the games the only game not available for the old gen is the one that scales the most. That would seem to justify concern about the future trend for games not available on last gen consoles.

So yes, accurate within it's own scope. But the scope seems needless narrow, and hardly reflective of the diverse nature of PC gaming.

Thus question is strongly begged - what about other games. Non AAA, non first-person/quasi first person. Non-console? There must be a sizable section of the PC gaming market who are primarily concerned with PC exclusive titles.

Anyway, I'll have to take your word that you haven't seen any evidence to the contrary. I'm not sure that excuses not testing for it!

http://forums.totalwar.com/showthre...2-graphics-engine-visualized-with-Afterburner

I've seen a lot of discussion about CPU bottlenecking on the Assetto Corsa forums, and have also heard more than once that X-Plane also eats every drop of CPU power (especially with scenery add-ons).

Personally, I'd love to read the broad kind of test I'm talking about, and I think you'd be doing the market a big service.

Cheers!