Why does i5 6400 get better marks than i5 7400 and i5 7500 ?

[Bruno Vincent]

Seems according to benchmarks it's quite a bit faster on multiple core tests:

https://browser.geekbench.com/v4/cpu/search?dir=desc&q=i5+6400&sort=multicore_score

Any reason for this?

 

[feelinfroggy777]

Geekbench is not a very reliable or consistent benchmark. I would go with Cinebench. Considering there are millions of different combinations of RAM and Motherboards, there will be anomalies in any bench.

The 6500/7500 is faster than the 6400. That is a fact.

 

[TJ Hooker]

In addition to what was said above, it is possible (or at least used to be possible) to overclock locked (non -k) Skylake CPUs with the right motherboard and BIOS. I would guess that the highest results are from overclocked 6400s, and the frequency isn't being reported correctly. Overclocking locked CPUs isn't possible with Kaby Lake CPUs, hence the highest results for those CPUs being lower.

 

[theyeti87]

That's not what I am seeing.

http://cpu.userbenchmark.com/Compare/Intel-Core-i5-7400-vs-Intel-Core-i5-6400/3886vs3512

http://cpu.userbenchmark.com/Compare/Intel-Core-i5-7500-vs-Intel-Core-i5-6400/3648vs3512

 

[TMTOWTSAC]

What TJ said. The first page for the 6400 results are all MSI z170 7984 or 7971 motherboards which allow overclocking. You wouldn't normally pair a more expensive OC board with a locked chip. But that generation allowed BCLK on specific boards running older BIOS versions. BCLK is much more hit and miss than frequency multiplier OC though, since it affects the speed of multiple components, not just the CPU.

 

[Bruno Vincent]

In addition to what was said above, it is possible (or at least used to be possible) to overclock locked (non -k) Skylake CPUs with the right motherboard and BIOS. I would guess that the highest results are from overclocked 6400s, and the frequency isn't being reported correctly. Overclocking locked CPUs isn't possible with Kaby Lake CPUs, hence the highest results for those CPUs being lower.

Wow! Does it mean an older overclockrd cpu can be faster than newer models?

Also, the ryzen 1600
gets a 25000 score, are those overclocked ryzens? I5 6400 = 18000 and i5 7500 16000

 

[TMTOWTSAC]

It depends on the IPC (instructions per clock) difference between generations, the clockspeeds you're able to hit with your OC, and platform enhancements such as DDR4 vs DDR3 RAM and RAM speed. For Skylake (6000 series) and Kabylake (7000 series) they're virtually identical in everything but clockspeed and some video enhancements. So core for core, whichever reaches higher clocks wins. For older generations, clock for clock they're slower so they need a greater clockspeed advantage in order to pull ahead.

As far as that 25000 number, for one it's definitely overclocked. But the bigger difference is that it's a multithreaded score. i5's have 4 physical cores each running 1 thread. Ryzen 1600 and 1600x have 6 physical cores, each able to run 2 concurrent threads. Ryzen is slower core for core, but it has 50% more cores and 300% more threads than the i5. That more than makes up for the slower cores when you're running heavily multithreaded workloads.

 

[mdd1963]

Many of the synthetic benchmarks scale much better ( nearly linearly) with click speed bumps, more cores, more threads, etc., but, to me anyway, the 1080p gaming benches more accurately tell the tale. If a CPU is near the top in BF1, it's not going to be a slouch in much of anything else. (whereas Cinebench scales great with core increases, great for rendering, etc.)

 

[Bruno Vincent]

It depends on the IPC (instructions per clock) difference between generations, the clockspeeds you're able to hit with your OC, and platform enhancements such as DDR4 vs DDR3 RAM and RAM speed. For Skylake (6000 series) and Kabylake (7000 series) they're virtually identical in everything but clockspeed and some video enhancements. So core for core, whichever reaches higher clocks wins. For older generations, clock for clock they're slower so they need a greater clockspeed advantage in order to pull ahead.

As far as that 25000 number, for one it's definitely overclocked. But the bigger difference is that it's a multithreaded score. i5's have 4 physical cores each running 1 thread. Ryzen 1600 and 1600x have 6 physical cores, each able to run 2 concurrent threads. Ryzen is slower core for core, but it has 50% more cores and 300% more threads than the i5. That more than makes up for the slower cores when you're running heavily multithreaded workloads.

 

[Bruno Vincent]

The plot thickens!

I have no intention of gaming, only productivity.

So the geekbench score include overclocking?

I won't overclock my Ryzen, so how do I know what performs better for my usecase?

Debating i5 7600 versus Ryzen 1600, wondering if Ryzen "really" is any better?

 

[feelinfroggy777]

Ryzen is better for multithreaded workloads. If you are using the CPU as a workstation, then go for Ryzen and dont think twice. The 7600 best for gaming. So if your not going to game, then the choice is easy.

As I said earlier, Geekbench is not a reliable benchmark. Especially for AMD CPUs. If you want reliable benchmarks for CPU look at Cinebench. It will give you a much more accurate portrayal of the CPUs and where they stand.

 

[TMTOWTSAC]

Well, you can get lost in the infinite variety of theoretical workloads. Focus in on your specific needs. What sort of computing do you do? What programs do you use while doing it? Compare processors with how they perform using those programs. Figure out the tradeoffs and buy accordingly.

 

[Bruno Vincent]

Thank you everybody for the excellent advice

Where can I see cinebench results?

And additionally, where can I see [non gaming] results?

Only interested in productivity and multitasking

 

[Bruno Vincent]

I downloaded cinebench and my laptop now gets:

7.66 fps
127 cb

On what website can I view the Ryzen and i5 stats for the same test?