Intel CPU's for 2016

[Howellerman 104]

Looks like Kaby Lake is not to end of year and Broardwell E, expensive and does FSX/P3D really use more than four cores?

 

gb.

Heya Gboz,

 

When I am flying GA aircraft around PNW and similar on my OC 3770K with hyper-threading enabled, P3D v3.2, I see all eight threads running pretty much full tilt, with "system" busy in the 80+ percent busy range. It is the terrain processing that loads up the cores, and the terrain engine is definitely multi-threaded. You can see this yourself by turning on Task Manager, Performance, and setting refresh rate to Slow. Let everything settle. Start up P3D and what those cores/threads become saturated during terrain loading. I have P3D installed on an SSD, so that may allow better utilization/CPU saturation, but it is clear that terrain loading is multi-threaded.

 

JKH

 

JKH 

[jabloomf1230 2,445]

You're correct John, P3d will use as many cores as available to load ground texture tiles.

[Slayer 190]

Still , there's only so much to load and you are limited to what your disk or SSD can feed it. Ive always thought there is a diminishing return with 8 or more core chips. Id rather have less cores and a faster mhz.

[Howellerman 104]

I don't know that for sure, Steve. I remember Rob Ainscough on his hyper-threaded 6 core/12 thread system running all those logical processors pretty hard!

 

That being said, I think the next system I put together will use the upcoming Broadwell 6900K with 8 cores, 16 threads. At roughly a thousand bucks, it is a LOT less expensive than the 10 core/20-thread version! And if I can get the same 4.2GHz that I am seeing today on the outdated 3770K I think it *should* be best of both worlds. 

 

A project for when I retire next Spring!

[martin-w 3,234]

That being said, I think the next system I put together will use the upcoming Broadwell 6900K with 8 cores, 16 threads. At roughly a thousand bucks, it is a LOT less expensive than the 10 core/20-thread version! And if I can get the same 4.2GHz that I am seeing today on the outdated 3770K I think it *should* be best of both worlds. 

 

 

 

6900K = A massive £887.

6950X = An enormous £1398.

6700K = A very reasonable £259.

 

I would then ask myself a question. If Broadwell E and it's numerous cores is indeed advantageous in the sim [and I still doubt it] HOW advantageous is it?

 

Does it really make sense to spend a HUGE £798 more or an enormous £1139 more for subjective opinions like "oh that looks a bit smoother, or "guess what guys I get 3% more frames than with Skylake"? 

 

Is that good value for money, a decent bang for the buck? Is it really such a huge advantage to have more cores and a bigger cache? A big enough advantage that it's worth spending mega bucks?True, for some people money is no object, £1000, £2000, £3000 means nothing to them. But for 99.9% of us it would be a crazy investment. 

 

I'm playing devils advocate here to a degree, but the point is, that before those of us replete with cash spend tons of money on a hugely expensive CPU, for what most would deem very small gains, we first need to see REAL numbers. Real numbers generated by objective means. That means some kind of objective metric that definitively proves that there's a worthwhile gain. That means comparing systems with the proper utility [that I don't believe exists] and elimination all variables between the competing Skylake and Broadwell E systems. 

 

In another thread, someone posted that in another forum Broadwell E owners are posting "better numbers".This statement means nothing for the reasons above. Subjective opinion, no objective metric and failure to eliminate variables just isn't good enough when it comes to HUGE additional expenditure on an insanely expensive CPU.

 

Believe it or not, I could afford the 6950X. But no, I went for the 6700K. No way would I pay such a huge sum for, at this stage hypothetical, and ultimately if confirmed, small gains. Just my opinion, if others wish to spend their money that way, then of course they are free to do so. I'm just trying to apply a little perspective.

[Howellerman 104]

 

 

subjective opinions like "oh that looks a bit smoother, or "guess what guys I get 3% more frames than with Skylake"?

 

You do know your audience, yes?    

 

 

Definitely agree that the two additional cores for the 6950 are not worth the uplift! That would be like buying a Koenigsegg Agera-RS instead of a Enzo Ferrari. Oh... wait... never mind...

[martin-w 3,234]

You do know your audience, yes?

 

 

 

I do, and even for we crazy simmers, who spend a lot of dosh chasing performance that gamers would expect from their systems out of the box... 798 quid or 1139 EXTRA is a real eye opener for mediocre gains.

[VeryBumpy 272]

Just dreaming here but it would be cool if AMD came out with a FSX specific CPU for us simmers that somehow fooled FSX that it was an equivalent to 7GHz. 

[SpiritFlyer 987]

but the point is, that before those of us replete with cash spend tons of money on a hugely expensive CPU, for what most would deem very small gains, we first need to see REAL numbers. Real numbers generated by objective means. That means some kind of objective metric that definitively proves that there's a worthwhile gain.

 

 

Exactly, and I would even include (with a stretch) the continuing viability in FS of super fast Sandy Bridges, as being shown in other threads...

 

"Hyper-jumps" in CPU technology will come BUT insofar as flight driven performance in the past few years is concerned, it is pretty much "Impulse" driven.

 

It would be good to have "real change" (for a change) on the core clock count to equal the realized dramatic jumps on the GPU side.

 

Kind regards,

[magnetite 38]

Having a 7 GHz CPU only gets you so far. You still need a GPU that's fast enough to render all the stuff the CPU is sending it. As well as a PCIe bus that's fast enough to send things from the CPU to the GPU. 

[martin-w 3,234]

Exactly, and I would even include (with a stretch) the continuing viability in FS of super fast Sandy Bridges, as being shown in other threads...

 

"Hyper-jumps" in CPU technology will come BUT insofar as flight driven performance in the past few years is concerned, it is pretty much "Impulse" driven.

 

It would be good to have "real change" (for a change) on the core clock count to equal the realized dramatic jumps on the GPU side.

 

Kind regards,

 

 

 

When I mention a "worthwhile gain" that is of course a subjective  term. And given the following in regard to sandy Bridge and Skylake...

 

Average User Bench +42%

Peaked Overclock Bench +33%

 

Some would regard the above as a worthwhile gain, and I'm sure some wouldn't.  Dependent of course on the specific performance gain realized in their favored application.

 

http://cpu.userbenchmark.com/Compare/Intel-Core-i7-6700K-vs-Intel-Core-i5-2500K/3502vs619

 

 

It would be good to have "real change" (for a change) on the core clock count to equal the realized dramatic jumps on the GPU side.

 

It would, but that would require a change in philosophy at Intel and AMD. Not to mention technological breakthrough. It's getting harder and harder to increase frequency. The transistor count isn't in the millions any more, it's in the billions. 

 

The more quickly we switch transistors on and off the more heat generated of course, and therefore lower clock speeds generate less heat. Not forgetting of course that an increase in clock speed necessitates an increase in voltage, and the cost of power and environmental impact is an issue these days, particularly in data centers.