Seeking help understanding significance of benchmark differences

The CPUs in my old but very dependable office PCs are as follows: i7-3770 @ 3.40MHz, benchmark 6372 (from cpubenchmark.net) and i5-4570 @ 3.20MHz, benchmark 5151. They are now 9 and 8 years old, respectively, and I bought both as inexpensive refurbs. Both run W10 Pro 21H1 really well. Both are, of course, still running on DDR3 RAM. I'm not a gamer, and for my computing needs, I have not experienced any booting or operating lags with either. Interestingly, though, the i5 creates new Restore Points and boots (from button push to useable desktop) somewhat faster than the i7.

I'm thinking about replacing one or the other with a 4-year-old i7-7700 @ 3.60MHZ, benchmark 8618, that I've found at an attractive refurb price. It runs on DDR4 RAM.

I'm not sure what to make of these benchmark differences. Here's my question, then: Given benchmarks 5151, 6372, and 8618, can I expect to notice any performance difference (in everyday computing) that will reflect a two- or three-thousand -point difference in benchmarks?

[Oh, and for the sake of this discussion, let's ignore the question of Windows 11-capable (atm) hardware.]

Wisewiz said:

The CPUs in my old but very dependable office PCs are as follows: i7-3770 @ 3.40MHz, benchmark 6372 (from cpubenchmark.net) and i5-4570 @ 3.20MHz, benchmark 5151. They are now 9 and 8 years old, respectively, and I bought both as inexpensive refurbs. Both run W10 Pro 21H1 really well. Both are, of course, still running on DDR3 RAM. I'm not a gamer, and for my computing needs, I have not experienced any booting or operating lags with either. Interestingly, though, the i5 creates new Restore Points and boots (from button push to useable desktop) somewhat faster than the i7.

I'm thinking about replacing one or the other with a 4-year-old i7-7700 @ 3.60MHZ, benchmark 8618, that I've found at an attractive refurb price. It runs on DDR4 RAM.

I'm not sure what to make of these benchmark differences. Here's my question, then: Given benchmarks 5151, 6372, and 8618, can I expect to notice any performance difference (in everyday computing) that will reflect a two- or three-thousand -point difference in benchmarks?

[Oh, and for the sake of this discussion, let's ignore the question of Windows 11-capable (atm) hardware.]

Don't go by cpubenchmark.net Run Cinebench R23 on them to see the actual difference.

As for how the different benchmarks will show themselves... it'll be in weird things, like zipping files faster, resizing images, faster searches, pretty much anything that runs mostly on the CPU.


You can find the Cinebench offline installer, here... Cinebench Release 23 Installers




CPU bound examples: What are some examples of CPU bound and non-CPU bound problems? What would be the best programming language to tackle each situation? - Quora

Ghot said:

Don't go by cpubenchmark.net Run Cinebench R23 on them to see the actual difference.

That's fine if you have the CPU installed so you can test it, but what if you are assessing the performance of prospective purchases? Do Cinebench have a comparison table of test results? Cpubenchmark do, and that's why I still go there before buying any new machine (well, used/refurb in my case).

In my experience their benchmarks are a reasonable guide to the performance you'd expect, a higher benchmark will be a faster Windows machine. But the benchmark alone is not the only factor. Windows seems to thrive on multiple cores and/or threads - the more the better. In my experience Windows will feel 'slicker' on a cpu with more cores/threads than on one with the same (or slightly higher) benchmark, but less cores/threads.

RAM speed and SSD vs HDD speeds are also a significant factor in the apparent 'speed' of Windows.

To answer @Wisewiz's question...

: Given benchmarks 5151, 6372, and 8618, can I expect to notice any performance difference (in everyday computing) that will reflect a two- or three-thousand -point difference in benchmarks?

Yes, but probably not by as much as you'd expect. On my machines a difference of 500-1000 make a hardly noticeable difference to how Windows 'feels'. With a 3000 difference it will be more noticeable, but other factors besides cpu speed may make that appear to be only a small improvement.

@Ghot @Bree

Thank you both very much. Bree, my old machines are running on 16GB of RAM at about 800MHz, so stepping up to 16GB of DDR4 @ 2400MHz should make whopper of a difference. For some things, at least.

Ghot, is there anything special I should know before I trigger Cinebench.exe? Gonna try it on the main (i5) machine first.

Wisewiz said:

@Ghot @Bree

Thank you both very much. Bree, my old machines are running on 16GB of RAM at about 800MHz, so stepping up to 16GB of DDR4 @ 2400MHz should make whopper of a difference. For some things, at least.

Ghot, is there anything special I should know before I trigger Cinebench.exe? Gonna try it on the main (i5) machine first.

Not really. I wouldn't do the single core bench... that one takes forever.

My Cinebench score. Everything is stock.




@Bree

Cinebench Leaderboard

Cinebench R23 (Multi-Core) CPU benchmark list

Ghot said:

@Bree
Cinebench Leaderboard
Cinebench R23 (Multi-Core) CPU benchmark list

So nothing from Cinebench themselves, just from third parties? And patchy coverage for older CPUs? For my eclectic collection of laptops I can only find the most recent processor (an i7-6600U) on cpu-monkey, while cpubenchmark has them all (and allows comparison of up to 5 CPUs).

Intel Celeron T1600 @ 1.66GHz vs Intel Pentium B950 @ 2.10GHz vs Intel Core i5-520M @ 2.40GHz vs Intel Core i7-4600U @ 2.10GHz vs Intel Core i7-6600U @ 2.60GHz [cpubenchmark.net] by PassMark Software

(BTW, the one with the 1st gen i5-520M is currently doing very well at testing the W11 Dev build)

Time for you two to agree to disagree

Okay, so I have a multi-core score 0f 2978 for my i5, and the board that the utility displays shows (are they spying on me? ) an i7-7700 at 6302. That's the machine I'm considering. But it says single-core for the result of 6302. Does that destroy the comparability?

The multi-core took about 9.5 mins to run. That significant?

ThrashZone said:

Hi,
Budget always rules whether it's time or not to upgrade but if your older systems which I have a few still meet your needs then keep it simple [...] but forking out atm 2k.us on a older system wouldn't be advisable.

2k.us? I'm looking at about USD600. Oh, and I'm a dust fanatic. These SFF PCs are so easy to open and service that I crack all three of mine (counting the one that belongs to the missus) every 60 days or so and give 'em the no-heat hair-dryer treatment.