Skylake - whaddya hear, whaddya say

[Gman]

I've been waiting a while now for a CPU/platform to come along to make it worthwhile to upgrade the MB/CPU on a couple systems.  I had a really cheap used 5960x chip fall in my lap, so I built up a reasonable x99 platform with a single 980.  My x79 system with the same video card is within a couple % usually in most common games I've found, very little reason to upgrade IMO for bang/$.

This has been the case for the last couple years IMO, since the 4770k and other i5 similar generation chips and the x79 platform stuff, there has been very little increase in overall gaming performance other than video card/GPU upgrades IMO.  IE just upgrading the GPU is 90%+ of the ball game.

I'm hoping that the Skylake 67xxk stuff coming in the next quarter or two will finally be a big enough jump in performance to make upgrading a worthwhile effort so far as the MB and CPU go.  More stuff is leaking out weekly, but still nothing really mind blowing or concrete. 

[Skuzzy]

Games are moving more and more of their operations to the video card.  It is doubtful you will see much on an improvement in most games with a better CPU.

Now for other applications, such as audio or video work, a faster CPU will be very welcome.

[Gman]

See that's just it - there HAS been a marked improvement in applications performance as you've said, my 5960x will crush stuff compared to 4770k or 3820 x79 CPU, which is great, if you do that sort of thing.  I don't, so it matters little to me (for now).  I think you're correct, even Skylake won't make much of a difference. 

I suppose it's just a matter of accepting that, and upgrading anyway at this point, it just seems...odd..I don't know, strange, that as you say SO much of the performance is based on video card performance, and any old cpu from the 2500k up to 5960x will do, and show little change in performance when tested using the same GPU.

If anyone would know why that is, I'm sure you do Skuzzy, since you're building a new game and all.  It's systemic across the entire game universe now IMO, CPU matters little, GPU matters much.  10 years ago, even not that long, it certainly wasn't the case, it was a more balanced deal, in terms of new hardware, you sort of had to upgrade both CPU/MB and the GPU.

[Bizman]

What interests me the most is will the socket architecture be a long living one like the current LGA 1150 with a truckload of different processors to choose from. That would enable choosing a budget processor at first and get a top tier one later when the prices come down.

[Ratsy]

What interests me the most is will the socket architecture be a long living one like the current LGA 1150 with a truckload of different processors to choose from. That would enable choosing a budget processor at first and get a top tier one later when the prices come down.

This 

[Skuzzy]

GMan, this venue is not the best for talking about a topic as complex as the one you want addressed.

The simple explanation. 

In the last 5 years, or so, video card performance has not only increased for all operations, they have gotten more sophisticated in what they can do to generate the graphics and how they can do it.

On the graphics card we can run multiple operations, at the same time.  We cannot do that on the CPU side.  For instance, if we have 8 layers to process to get one pixel drawn, we can do that in one basic operation on the video card due to the massive parallelism available.  On the CPU side it would take, at least, 8 serial functions to accomplish the same task. 

Regardless of the number of cores on the CPU side, in a worst case scenario (you have to plan that way as you have no control over it) all the cores have to wait in line to do anything as only one core can access anything external (i.e. PCI, PCIe, SATA, Memory buses and so on).

The other advantage the GPU has is in its math performance.  GPU floating point math is exponentially faster than the CPU floating point performance.

Again, this is a very simple explanation.

[Gman]

Simple to you perhaps, not so much for most of the end users - I read PC gaming sites every day, and I've never had that spelled out before, not from at least a dozen various hardware/software sites.  Thanks for the explanation - surprisingly there has never been an article addressing this anywhere, not that I've found or read, and it's an important issue IMO, considering how expensive and yet popular PC hardware is, that end users who aren't coders/creators/whatever know these things, even if it is dumbed down for us.  I've know the "what", just not the "why", even if it is broad strokes.

I found the one point especially interesting, "you have to plan that way as you have no control over it", as that sort of clears up for me at least, some of the issues regarding use of cores on many core CPU chips.

[Skuzzy]

Most WEB sites do not have a true systems engineer available to help with technical writing, nor one who has seen the evolution from simple time slicing in UNIX V6 which made one CPU core appear to run hundreds of applications at the same time, to the all out hardware controlled threading available today.

Most WEB sites may have a knowledgeable enthusiast who efforts conclusions based on a simple set of parameters. Many buy into marketing as their fact base.  For the most part people like marketing driven sites as it is easy to understand, regardless of the accuracy.

I tend to try and stay away from technical writing as most people find it boring and are happier with "shiny" talk.  That is why I did not break down and write a 50 page white paper to answer your question.

[Bizman]

Skuzzy, thanks for the simplified explanation. Simple is usually better, even with the cost of some inaccuracy. It helps people understand things and understanding is the key and gateway to deeper knowledge. 

It's not only the web sites that evaluate new stuff using only a few parameters, magazines do the same. Could it even be some sort of hidden marketing? The reviews tell how much faster a multi core CPU can do certain tasks. Then again the same sites and magazines tell how much better graphics cards are in multiple layer calculations such as Bitcoin mining compared to CPU's. IIRC there is even graphics cards without any monitor connector, just for running multiple operations faster than a CPU. Combining these fragments of information has been left to the consumer who for the most part doesn't have what it takes.

[Pudgie]

GMan, this venue is not the best for talking about a topic as complex as the one you want addressed.

The simple explanation. 

In the last 5 years, or so, video card performance has not only increased for all operations, they have gotten more sophisticated in what they can do to generate the graphics and how they can do it.

On the graphics card we can run multiple operations, at the same time.  We cannot do that on the CPU side.  For instance, if we have 8 layers to process to get one pixel drawn, we can do that in one basic operation on the video card due to the massive parallelism available.  On the CPU side it would take, at least, 8 serial functions to accomplish the same task. 

Regardless of the number of cores on the CPU side, in a worst case scenario (you have to plan that way as you have no control over it) all the cores have to wait in line to do anything as only one core can access anything external (i.e. PCI, PCIe, SATA, Memory buses and so on).

The other advantage the GPU has is in its math performance.  GPU floating point math is exponentially faster than the CPU floating point performance.

Again, this is a very simple explanation.

Just to let you know Skuzzy, I wouldn't mind at all reading a 50 page white paper from you on this subject..................
 

Just from reading what you did post is making me smile.........