• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

SuperPi puts me off Skylake

Associate
Joined
18 Mar 2003
Posts
1,129
(Before I post some findings let's not dig up this old argument of why SuperPi is not needed, or why single threaded benchmarks is Zzzzzzzzz. It is needed by some as a way of determining custom application performance; it is a time tested benchmark; it is used by major review sites :D )

SuperPi in the Sky
Looking at some of the benchmarks for Skylake and Haswell I noticed that clock for clock SuperPi scores are worse on Skylake.

This is a big disappointment, as usually the tick-tock performance increases as quoted by Intel have generally translated into those performance gains in SuperPi (from Sandybridge to Haswell +15% etc.). But on one site I see this:

SuperPi 1M
i7 6700K 8.643
i7 4790K 8.159

I appreciate the 4790K will ramp up to 4.4Ghz, whilst the 6700K will ramp up to 4.2Ghz. But equalising the clock speeds the 6700K is about 1% behind the 4790K.

Reading Stulid's GA-Z170XP-SLI Review where the 6700K is overclocked to 4.6Ghz, and comparing that to my 4770K clocked at 4.6Ghz reveals this:

SuperPi 1M
i7 6700K @4.6Ghz 8.003
i7 4770K @4.6Ghz 7.846

That puts Skylake as being 2% worse than my 4770K. So much for a "20% improvement" as quote by Intel.

So why is Skylake so bad for this benchmark?

CPU or Chipset Architecture
Is there something different in the processor or chipset that causes this?

Memory
My 4770K is DDR3 at 2133Mhz, whilst Stulid's test was DDR4 at 2800Mhz. I believe SuperPi performance is dependent on mem speed so you would expect the 2800 to be better.

Maybe this is a latency issue?

My timings are: 9-11-10-27-278-2T

SuperFlawed
Maybe SuperPi is misreading the results because of the architecture? Like on some mobos there is an option to correct a 3D gaming benchmark or something.

I am using SuperPi Mod 1.5 XS
 
Last edited:
^ In most of those they are level pegging, and in some the 4790K is out in front.

What sort of a test is that supposed to be anyway? It isn't even clock for clock!
 
You know. I was in two minds to put the disclaimer (Before I ... at the end of the post but assumed starting off with it would prevent those very replies I did not want :D

Can anyone answer the questions?

Is it the memory? Is it the timings? What are the tightest timings you can get for DDR4 2133Mhz? Would that make SuperPi results faster for Skylake?

Edit. Just saw P4Clocks reply.
 
Just checked some SuperPi results for OCd 5820K.

5820K @4.584Ghz 16GB Ripjaws @2716MHz = 7.846

That is exactly the same (spookily exactly the same) as my 4770K

Another test

5820K @4.7Ghz CL14.0 14-14-35 (speed not quoted) = 7.840

I guess it is the case that DDR4 timings is the problem for me here. Hmm. May have to hold off on the upgrade.
 
^ In most of those they are level pegging, and in some the 4790K is out in front.

What sort of a test is that supposed to be anyway? It isn't even clock for clock!

You must need glasses, Skylake is around 5-10% faster 90% of the time in the video, despite being clocked 300Mhz less.
 
Personally Skylake on its own is not something I would consider an upgrade too though.

Maybe Skylake E, from x79, but even then I doubt it is worth the outlay.
 
there is no reason not to get skylake, single core performance of skylake is very fast nothing on the market can keep up with it.
everyone keep banking on about how 6 cores are a must and skylake is already over the hill.


OP - What is your current system and what do you use it for?
 
there is no reason not to get skylake, single core performance of skylake is very fast nothing on the market can keep up with it.
everyone keep banking on about how 6 cores are a must and skylake is already over the hill.

There is if you use your pc for programs that use multiple cores/threads. A 5820k is the same price yet will be much quicker in multithreaded applications clock for clock.

Just for gaming, yes a 6700k would probably be best but it is hardly a huge leap from Haswell so with Haswell E you get the best of both worlds for the same price. That is why most suggest getting a 5820k over a 6700k.

You also then have the x99 platform and a seemingly pretty powerful upgrade path as rumours are that Broadwell E will have an 8 core 16 thread cpu that won't be an "extreme" (and therefore crazy overpriced) part - the "extreme" chip and price is seemingly reserved for a 10 core 20 thread chip.
 
Last edited:
There is if you use your pc for programs that use multiple cores/threads. A 5820k is the same price yet will be much quicker in multithreaded applications clock for clock.

Just for gaming, yes a 6700k would probably be best but it is hardly a huge leap from Haswell so with Haswell E you get the best of both worlds for the same price. That is why most suggest getting a 5820k over a 6700k.

You also then have the x99 platform and a seemingly pretty powerful upgrade path as rumours are that Broadwell E will have an 8 core 16 thread cpu that won't be an "extreme" (and therefore crazy overpriced) part - the "extreme" chip and price is seemingly reserved for a 10 core 20 thread chip.

NO! for gaming you need a i5 not the i7, why do ppl keep looking at the cost of i7's and saying it better to have have the 5820k you dont need an i7 for gaming.
if your gaming you need an i5 16gb of ram and a GPU then your set for the next 5 years.

My boy runs bf4, titanfall and fifa16 at max setting with an i3 4150.


Edit: i just sold off an 5930K rig because i got an 6600k rig for very cheap off a friend and in ever thing i do that is is faster.
ppl keep saying the extra cores are a must. if you an average pc user you have no need for the cores or threads.

:rolleyes: BUT Dave... console games use 8 cores.... 8 AMD cores. a 2500k is better than the AMD 8 core cpu's
 
Last edited:
NO! for gaming you need a i5 not the i7, why do ppl keep looking at the cost of i7's and saying it better to have have the 5820k you dont need an i7 for gaming.
if your gaming you need an i5 16gb of ram and a GPU then your set for the next 5 years.

My boy runs bf4, titanfall and fifa16 at max setting with an i3 4150.


Edit: i just sold off an 5930K rig because i got an 6600k rig for very cheap off a friend and in ever thing i do that is is faster.
ppl keep saying the extra cores are a must. if you an average pc user you have no need for the cores or threads.

:rolleyes: BUT Dave... console games use 8 cores.... 8 AMD cores. a 2500k is better than the AMD 8 core cpu's

What? I was only comparing a 6700k to a 5820k.

Yes an i5 will do for gaming. However there are some games where an i7 will help hugely (as an example, try the grass areas in Crysis 3 - i got a massive framerate boost going from a 2500k to a 3770k in that part). For the most part though, yes you won't notice the difference between a similarly clocked i5 and i7 in a lot of games.

However, you said "there is no reason not to get skylake" - There are many reasons not to get skylake depending on what you use your pc for.
 
^ In most of those they are level pegging, and in some the 4790K is out in front.

What sort of a test is that supposed to be anyway? It isn't even clock for clock!

In terms of what can be expected out of most samples, it's like for like enough. Most if not all 6700K should be able to manage 4.6 - if you look across the board you'll struggle to find one that won't (there is good reason for this)
 
However, you said "there is no reason not to get skylake" - There are many reasons not to get skylake depending on what you use your pc for.

I mean for the day to day user. i have seen ppl open a spec me post asking for a gaming rig spec and ever post say dont get skylake get 5820k...
hell i have seen ppl ask for day to day systems and have spec'ed up 5820k systems. its gotten to some stupid stage where its x99 or nothing.

you need the CPU that fits your needs.
my 5930k 16 2666Mhz ram and 290x plaid games very well. but now my 6600k 8gb 2400mhz ram and 290x plays games better. i get better FPS in all games i play.
however if i spent all day video coding i have have an 8 core xeon.
 
There is more to life than gaming. Some use rigs for:

Financial modelling, medical research, weather forecasting, simulations, a whole host of number crunching solutions.

In some instances multiple cores may not be the answer. Faster Ghz and lower latencies is the key to being overall faster than a previous rig.

Think of it like a vehicle that you you use to get across the county. Sure you can have your 3.2L four cylinder Jag that gets you, your nan, grandad and the kids to the shops quick enough. You also have plenty of boot space for the shopping. Then someone can have a V6 Cayenne or something that has more space and will transport your goods quicker.

But imagine you are in a city that is constant stop-start. Traffic lights, traffic, roundabouts. What is needed here is a vehicle that can hit 155MPH on the open road (as fast as the V6) but more importantly has the fastest standing start getaway speed.

In order to do that you strip out the seats, change the wheels, upgrade the suspension. Sure it's now no good for transporting your Nan to Asda but it sure is the fastest way to do letter drops.

So every two years FordTel produce a new leaner engine. They say "it's 20% faster than our 2013 model" and history shows this to be correct. But in 2015 their new Skylake engine runs into a problem for the letter drop guys. Yes it now hits 165MPH, yes it's got plenty of room for Nan, Grandad and little Tarquin. But they did something to the wheels. The standing start 0-60 speed has now got 1% WORSE rather than 20% better. It seems that the tyres are causing wheel spin (re: higher latency) and for someone who needs to zip around town this is a big problem.

Memory Latency
This is the issue for me. I think a solution could be this:

White Dragon "SUPER Low-Latency" 16GB (2x8GB) DDR4 PC4-24000C13 3000MHz Dual Channel Kit - Blue

That is 3000MHz RAM Speed, CAS 13-13-13-35 Timings

or then there is this:

Savage Black 8GB (2x4GB) DDR4 PC4-19200C12 2400MHz Dual Channel Kit - Black (HX424C12SBK2/8

2400MHz RAM Speed, CAS 12-14-14-28 Timings

I guess I need to purchase both and see the trade off between wheel spin and high cruising speed; to see which is the best at doing all those letter drops around town.
 
The thing with each newer generation of intel processors, the largest performance improvements come through the means of Extensions which are going unused in the majority of software. Such as AVX2 etc.

There might only be 5% in each generation in normal operation but larger performance improvements are seen when the newer extensions are used. But the software needs to support it. We might wee wider adoption once Zen drops with support for AVX2 etc.
 
Back
Top Bottom