What is currently the fastest Single Thread CPU?

[SpeedyVV]

I'm not talking about GHz or any other spec.

Just any generic single threaded application performance.

 

[Araxie]

i7 5775C due 128mb eDRAM, at same clocks than a 6700K for example should perform better in any kind of workload.. however is very rare to find and pricey, also very low clocked and don't oc as high as a 6700K, so out of the box there's nothing better than the i7 6700K.

 

[PontiacGTX]

i7 6700k with DDR4-3000MHz+

 

[Araxie]

i7 6700k with DDR4-3000MHz+

Not exactly.. that depend entirely on RAM timings, mine with 10-12-12-28 at 2400mhz perform much better than any other typical 3000mhz+..

 

[Killerxp100]

How about 3200mhz @ 13-13-13-33...

 

[Araxie]

How about 3200mhz @ 13-13-13-33...

I haven't actually be able to reach that high at those timings.. very good indeed.. what kit are you using? those are almost DDR3 matching timings which is very good.. regular timings for that speed are about 15-17-17-35 normally..

 

[Killerxp100]

I haven't actually be able to reach that high at those timings.. very good indeed.. what kit are you using? those are almost DDR3 matching timings which is very good.. regular timings for that speed are about 15-17-17-35 normally..

It's a Gskill Trident Z kit 3200 Cas 14, great sticks. I have seen some reviews done that show higher clocks with slower timings is better then the other way around when it comes to gaming. I think it was 2400 cas 12 compared with 3000 cas 16.

 

[noko]

Large amounts of data transfers the faster ram will be faster, short but many accesses the better timing ram can beat it out. It really depends what is going on now with the above speed and ram timings that would ideal.

 

[Deimos]

Not exactly.. that depend entirely on RAM timings, mine with 10-12-12-28 at 2400mhz perform much better than any other typical 3000mhz+..

Not quite true. Timing counts for something but memory latency is a function of both timing and clock speed. As a very simple/rudimentary example, 3200mhz is a 33% increase over 2400Mhz. Timings would have to be 33% worse at 3200Mhz for the latency to be the same, but there would still be an increase in raw bandwidth which would offer a performance increase in certain situations.

There is a simple formula for calculating the latency in nanoseconds; CL/(frequency) * 1000 = timing in ns.
CL10 @ 2400 = 8.32ns
CL14 @ 3200 = 8.74ns

As you can see, 40% worse timings at the higher clock speed nets around 5% worse latency at the higher clock speed. The trade off is a 33% increase in raw bandwidth.

There is not much in it latency wise, but the faster memory speed has been proven to provide a tangible performance advantage (gaming) on Skylake.

 

[noko]

I'm not talking about GHz or any other spec.

Just any generic single threaded application performance.

An I5 6600K OC to to a max stable OC. You could also just OC an I5 6400, 6500 or 6600 (recommend the 6500 due to cpu multiplier is higher you don't have to go as high on the BLCK clock rate which can give you a potential higher clock). Downside is you loose the power saving function when OCing a non-K processor or lock cpu so a I5 6600K you may like better plus is easier just to change the multiplier and not be stuck with a continuous high speed OC while reading the news . Also the hypothreaded versions I7 consume a significant amount more power -> heat which can limit the OC depending upon cooling used and can actually reduce a single application performance.

The I3's are hypothreaded dual core chips, the Pentiums like the G5420 (Skylake core) is a dual core as well and for single thread it should do as well and OC the same and maybe better due to less heat then an I5 6500. Except unless you are running Doss, other system threads will be competing for those two cores so having extra cores to handle other threads will allow one core to be dedicated to a single thread task/program which can be set using processor affinity in task manager.