Anyone else see these Skylake benchmarks?

it uses less power and has a higher TDP, so....

Maybe more power for the GPU.

clocked 5% down from Devil's Canyon? What's that about?

Isn't that just the turbo reduced from 4.4GHz to 4.2GHz. Although I would have liked to see that increased even if that means a higher TDP socket.
 
There are more and more stories corroborating 5GHz on air..

Oh, how I hope this is true.
 
Looks like the same thing we've been hearing about skylake for a while

5-10% IPC improvement.

BUT!!!

clocked 5% down from Devil's Canyon? What's that about?

BUT!!!

it uses less power and has a higher TDP, so....
More overclocking?
stock clocking doesn't really matter, what it tops out at does though.

Well I'd be coming from an i7 950 so the "IPC increase" for me will be huge. :D
 
Well I'd be coming from an i7 950 so the "IPC increase" for me will be huge. :D

Me too, I've got a i7 920 as my main rig. I've got a ES laptop with a skylake in it I've had a chance to benchmark, so I can vouch that it's pretty close to a 50% IPC improvement over nehalem.
 
Me too, I've got a i7 920 as my main rig. I've got a ES laptop with a skylake in it I've had a chance to benchmark, so I can vouch that it's pretty close to a 50% IPC improvement over nehalem.

... O/T Does anyone else read "Nehalem" as Nephalem now?
 
The retail chips we've heard about are normally topping out at 4.7GHz.

whomp waaaaa,

I wanted 4.8 as an easy to hit target with the golden chips clocking 5.2ghz.

I mean we were hitting 4.0ghz on the nehalems almost 7 years ago and we haven't mustered much more than roughly 20% boot in clock speed?

This is why we need AMD!
Zen, please don't suck.
 
whomp waaaaa,

I wanted 4.8 as an easy to hit target with the golden chips clocking 5.2ghz.

I mean we were hitting 4.0ghz on the nehalems almost 7 years ago and we haven't mustered much more than roughly 20% boot in clock speed?

This is why we need AMD!
Zen, please don't suck.

Kyle's article spells it out pretty well. I have been told there are some 4.9GHz chips out there but I have yet to see one myself. People in the industry talk and what we have heard is that most reviewers are seeing precisely the same clock speeds we are. I think Skylake is still a win for Intel overall because the clocks are nearly what we saw with Sandy Bridge in the real world. On open air test benches with triple radiators we saw 4.9GHz and 5.0GHz out of 2500K's routinely. In the real world people generally clocked their CPUs around 4.5GHz-4.6GHz with a few daring to go for more. What the golden samples could do is really moot and everyone acts like 5.0GHz air cooled Sandy Bridge chips were super common and they just weren't. Let's not forget that memory scaling is a huge factor for Skylake. When you look at benchmarks for it you need to keep in mind the resolutions being used and what clock speeds the memory is running at. There are some gains to be had. Obviously more higher resolution, real world testing needs to be done in order to fully understand the impact Skylake has on higher resolution gaming. In that instance, comparisons to X99 are inevitable. In that regard it might not stack up so well but against Sandy Bridge, overall I think Skylake does look pretty good in a lot of areas. Whether it justifies the cost is certainly a debatable matter.

One thing not in the article that you guys probably aren't aware of is just how well Skylake responds to memory clocking from a motherboard standpoint. While not every motherboard can do this, much of the time attaining high memory clocks is a matter of setting X.M.P. to on in the UEFI BIOS and walking away. Most of the time highly clocked memory modules are more fickle requiring extra hand tuning to achieve the desired or advertised clocks in some way with 24/7 stability. Memory clocking on X99 can flat out be a pain in the ass thanks to the strap settings and ratios that don't work as well for overclocking as others do.

The other piece to the puzzle is Z170-Express. This is especially true when you compare Z170-Express to P55, P67, Z68, X58 and to some extent X79 users running quad core CPUs. In gaming people are right in that the platform doesn't by itself translate to more performance. From a storage perspective you are almost guaranteed 4x PCIe lanes to the M.2 slot along with better NVMe compatibility and of course the option to RAID M.2 drives. There are also connectivity upgrades in the form of more SATA 6Gb/s ports, USB 3.0/3.1 etc. which are helpful if you are coming from those really old platforms.
 
Last edited:
I have been told there are some CPUs that are hitting 4.9GHz but I have yet to see it. Kyle's article spells it out pretty well and that's what we've been hearing about other reviewers getting.

Yeah I've been all over every review I can find, but Kyle's comments seem to be spot on. Obviously overclocking has a ton of variables with ambient conditions, cooler selection, mobo selection and variability, different ram and ram speeds. That's all before we get to actual CPU silicon.

I also so an LN2 setup hitting 6.8.

::sigh::

So 5820K or 6700K is the question, still. I wish the 6700K was faster or slower.
 
Yeah I've been all over every review I can find, but Kyle's comments seem to be spot on. Obviously overclocking has a ton of variables with ambient conditions, cooler selection, mobo selection and variability, different ram and ram speeds. That's all before we get to actual CPU silicon.

I also so an LN2 setup hitting 6.8.

::sigh::

So 5820K or 6700K is the question, still. I wish the 6700K was faster or slower.

RAM speeds don't seem to have shit to do with anything. Because Intel has done away with the base clock straps, how you clock your RAM seems to have zero impact on CPU overclocking. The system is just as happy at DDR4 3600 as it is at 2133MHz.
 
RAM speeds don't seem to have shit to do with anything. Because Intel has done away with the base clock straps, how you clock your RAM seems to have zero impact on CPU overclocking. The system is just as happy at DDR4 3600 as it is at 2133MHz.

Fair enough.

I was just trying to say that when you look around at reported overclocks you need to take the total system and test envrioment into account in order to mentally normalize the data.
 
Fair enough.

I was just trying to say that when you look around at reported overclocks you need to take the total system and test envrioment into account in order to mentally normalize the data.

A good point. Having to replace RAM, CPU and motherboard all at once is a tougher pill to swallow than just two out of the three components would be.
 
Back
Top