Time to upgrade your Core i5 2500K?

[poofyhairguy]

Guess with these numbers,Pascals 970 replacement may be the last gpu i should consider for my i5 2500 non k and 1080p.

Yeah I thought the 970 was the last GPU for my 2600K but we will see.

 

[Head1985]

You clearly have a different view of HUGE to lots of us. My P4C@3.2 to E6600@3.2 was huge, my Q6600@3.2 to i2500K@4.3 was huge. Skylake is just a little boost in comparison to them.

I have been ready to replace my i2500K for years, but I just can't justify it - it still does > 60fps in everything, and really skylake isn't that much faster.

Sandy-ivy-2-5%
Sandy-Haswell 10-15%
Sandy-skylake 30-50%
Skylake delivered biggest IPC gain since sandy bridge(with 3000+Mhz DDR4)

 

[know of fence]

Sandy-ivy-2-5%
Sandy-Haswell 10-15%
Sandy-skylake 30-50%
Skylake delivered biggest IPC gain since sandy bridge(with 3000+Mhz DDR4)

Read the last page of the Anandtech review; this architecture even performs worse in games clock for clock.

Overall, Skylake is not an earth shattering leap in performance. In our IPC testing, with CPUs at 3 GHz, we saw a 5.7% increase in performance over a Haswell processor at the same clockspeed and ~ 25% gains over Sandy Bridge. That 5.7% value masks the fact that between Haswell and Skylake, we have Broadwell, marking a 5.7% increase for a two generation gap.

​

In our discrete gaming benchmarks, at 3GHz Skylake actually performs worse than Haswell at an equivalent clockspeed, giving up an average of 1.3% performance. We don’t have much from Intel as to analyze the architecture to see why this happens, and it is pretty arguable that it is noticeable, but it is there. Hopefully this is just a teething issue with the new platform.

 

[ShintaiDK]

Using DDR3 1866...

 

[Aristotelian]

I'll be going from a 2600K that is overclocked to 4.4 to a broadwell-e 10 core not because I think that the new cpu will obliterate the 2600K in every scenario, but rather because I'm building a completely new system. I can see why many sandybridge users don't feel particularly compelled to upgrade, but it'll eventually happen. I agree with other posters who point out the incremental improvements - it's very different from video card upgrades for example. I wonder if we will ever see a paradigm shift in cpu power again. If we are really waiting for software to catch up, how long would a 10 core broadwell-e adopter have to wait?

 

[Head1985]

Using DDR3 1866...

Yep super slow Ram and they didnt even test it in games.And that test is pretty old now.
No win10/Titanx OC GPU

 

[Sweepr]

Using DDR3 1866...

Also very poor selection of games (mostly under GPU-bound settings), which makes it almost useless for processor comparisons. And the minor performance regression was explained here.

I like PCLab's review because they tested some of the most CPU intensive titles and used fairly decent RAM (DDR3-2133 9-9-10-24 1N for Haswell/Broadwell, DDR4-2666 16-17-17-36 2N for Skylake). Skylake was 12.8% faster than Haswell @ equal clocks on average (14 titles @ 1080p Ultra). About what you would expect from an Intel tock.

 

[Head1985]

Pclabs have good Cpu tests, but they should use faster Ram with skylake.
2666Mhz with Cl16 is pretty (s)low compared 2133Mhz on haswell.I think 2133mhz with only CL9 are even faster...
They should Use like 3000-3200Mhz CL15 for skylake.That will add another 5-10% to IPC gain.

 

[Carfax83]

Sandy-ivy-2-5%
Sandy-Haswell 10-15%
Sandy-skylake 30-50%
Skylake delivered biggest IPC gain since sandy bridge(with 3000+Mhz DDR4)

Those numbers look too low for Haswell, and too high for Skylake.

It should look like this. Sandy to Ivy 4 > 10%. Sandy to Haswell 15 > 25%. Sandy to Skylake 20 > 35%..

The percentage on the right indicates CPU limited circumstances, and faster RAM speeds will likely add another 5 to 10% on these numbers.

 

[know of fence]

I like PCLab's review because they tested some of the most CPU intensive titles and used fairly decent RAM (DDR3-2133 9-9-10-24 1N for Haswell/Broadwell, DDR4-2666 16-17-17-36 2N for Skylake). Skylake was 12.8% faster than Haswell @ equal clocks on average (14 titles @ 1080p Ultra). About what you would expect from an Intel tock.

Great post, great info. I actually ran the numbers for all 14 gaming tests and Haswell to Skylake @ 4.5 GHz clock for clock comparison, offers ~12% increase avg. FPS increase for both the i7's (12.2%) and i5's(12.7%). Whereas upgrading from i5 to i7 within a generation offers about a 6.5% increase for Haswell, and 6.1% for Skylake.
Watchdogs (~20%) and Battlefield 4(~15%) benefit the most from the 8 threads and/or bigger L3 cache the i7 provides.

These are all stupid FPS tests of course, in frame time/frame drop tests Broadwell should have done much better @4.5 GHz. Also they used 1080p on a 980Ti, creating a CPU bottleneck, and titles like Total War, Starcart are known CPU hogs (also Far Cry 4 for some reason).

Also props to PClab.pl power testing that shows OC'ed skylake consuming much more power, despite DDR4 and 14nm process. Increased OC performance comes at pretty much proportionally increased power consumption, which curiously even applies to Sandy Bridge.