I thought this ARTICLE was interesting as it confirms the anecdotal evidence of many SimHQers like myself that there really doesn't seem to be much benefit to upgrading if you have an i5 2500K or newer.

It's not immediately obvious from the article because the synthetic benchmarks muddy the waters, but all the practical benchmarks show that newer than Ivy Bridge CPUs only give marginal improvement over the Ivy Bridge generation. In that regard, it should be noted that the author did not benchmark the Ivy Bridge chips overclocked. Given that both the i5 2500k and i7 2600k are legendarily good overclockers (my own 2500k went to 4.8 GhZ on air without bumping voltage), the marginal improvement is almost sure to disappear with overclocking.

The main improvements have been in energy efficiency and features. Since these aren't a big concern to me, I think I'll stick with my CPU/MoBo even though it's been five years already!

Hmph. Yeah, gonna upgrade my video card this year, but maybe I'll wait on the CPU till my 2600k dies.



The Jedi Master

Yep, as I posted... Even my lical PC shop guy advised against selling me an upgrade for the six year old i7-860...

Yeah same here my i5-2500k has been running like a champ (and overclocked to 4.5ghz) since 2011, I think AMD is not pushing Intel at all on the gamer/performance CPU segment and so Intel is fine with maintaining the status quo.

Hopefully it stays this way for another couple of years, I'd rather spend money on GPUs upgrades than having to upgrade CPU + MB + Memory.

The motherboard might be a good upgrade so you can use M.2 SSDs. USB 3.1 might be nice to... CPU wouldn't be much better, and DDR4 won't be noticeable at all. I wish I bought a 2500K some years back, but I am running a 4670K. If it lasts a long time I won't complain.

Agree! CPUs have not gotten "significantly" faster -- just more power efficient. And, agree that no-one (including AMD) has been "pushing" Intel recently (though we AMD fans hope "next year will be different").

Another part of the story may be that we've reached the "speed limit" on just how fast a CPU's basic circuitry can work.

Long, long ago (when I was a Materials Scientist briefly working on potential CPU materials), we thought the limit of "silicon" was going to be far below 1GHz. That was wrong -- smart people figured out a way. However, the "practical limit" may be the approximately 5GHz we have today. Both Intel and AMD shot for 10GHz at points in the last decade. But, they both fell way short.

Now, the practical solution is more cores running on less power per core (so we can have more cores). But for the customer to "see" a difference, that solution requires well written "threaded" software -- that is not easy to write.

To me, that doesn't mean performance has stagnated, it just means that Intels performance priorities aren't in alignment with ours, and for good reason. It's a small contingent of titles these days that require such heavy CPU resources, with most games taxing the GPU more heavily. Combine that with the explosive growth of mobile devices over the past few years, and it makes way more sense to keep the "horsepower" the same while focusing on better "fuel economy". Also factor in middling competitive response from AMD, and the motivation to build high output CPUs shrinks even further.

That said, Laptop users are likely in a position to me more motivated to upgrade, as the higher power reqs/Heat output of the previous CPUs are a greater pain point for them.

It's true that for many games the developer push has been on things that are GPU intensive and not so much CPU.

However, while the clock speeds have not really progressed much in the last decade (hence the switch from clock speeds to model numbers as far back as the Core 2 Duo/Quad for Intel) the quest for more IPC has been the focus.
This seems to be where Intel has slowed their push. AMD isn't threatening them, but they've been a victim of their own success.

If your new product is only a single digit percentage improved over your previous one, why buy the new one? If AMD could at least match them then some people might switch away from Intel, leaving them a shrinking market share and an incentive to innovate. But with it either stagnant or growing there's little point.
Their problem isn't that they're losing sales to AMD, it's that they're losing PC CPU sales to people buying other devices. Market share maintained, but market shrinking.

Their comfortable lead over their direct competition blinded them to losing the lead over their indirect competition of tablets, smart phones, etc. Why upgrade that 5 yr old home PC (which will likely run Win 10 just as well or better than it did the Vista it came with) when you don't need it for anything other than PC games anymore?




The Jedi Master

+1 Jedi

Unfortunately for Intel, they don't provide the chips that power IPhones, IPads and other mobile devices.

Beyond gamers and engineers who use Auto-Cad, I really don't see much demand for the latest and fastest CPU's in the consumer market. CPU's that are 2 or even 3 generations old are "good enough" for the average user.

Worse for Intel and AMD, they're too much.

Who needs a big old monitor, a box under their desk, and a kb/mouse to check email, surf a few sites, watch some cat videos/pictures, and play a couple of casual games? Instead of being tied to one place in the house to do this stuff, sit on the couch! In bed! At the table! On the porch! In the basement! On the toilet! Anywhere!

PCs need a new killer app beyond PC games for the masses to reengage with it.




The Jedi Master

TBH, we are at the point in Processing Development,

Where,

Big Cores, High Frequency, High Single Thread Performance, Lots of Power

are being phased out in favor of:

Small Cores, Lower Frequency, High Single Thread Performance, Power Efficient

Smaller the Cores, the less Power you can subject it too..

The Issue is, Software Developers dont wanna spend the money to develop Multi Threaded Apps.

The First CPU Developer to Integrate a Hardware Layer that Makes all the Cores of their CPU appear to the Software as ONE there fore converting Serial Tasks to Parallel tasks, will see a huge performance boost in Performance.

One Company also had the experimental architecture of 4 small cores at 2.5 GHz and one Large Core at 5GHz, Single Threaded Apps would be set to use the big core while MT Apps use the 4.


The same for everything else, Software is the bottleneck.

There's only so much you can do with parallelization across multi-cores; see Amdahl's Law.

Ultimately, you're going to be bottlenecked by the serial parts of the workload, and that's why single-thread performance is still very important even in parallel applications.

To make matters more difficult for us, games and simulations are inherently difficult to parallelize efficiently due to their real-time nature.

If done wrongly, it can REDUCE performance compared to just single-threading because some threads get left hanging waiting for other threads to finish with the resources they're working on, among other things. It's not always as simple as "let's make this one single thread into a bunch of threads that use all my cores!"

Abstracting multiple cores into one virtual core sounds good in theory, but in practice, it will likely wind up with unexpected slow paths that are difficult to tune and debug for reasons mentioned above.

It's why the next-gen graphics APIs (Direct3D 12, Vulkan) are stepping back on the abstraction and giving developers more direct access to the hardware's workings, since the abstraction that's supposed to make things easy for developers (hey, it sure beats coding a renderer for every specific GPU architecture!) is now just getting in the way of people who know what they are doing.

With all that said, I'd really like to see a high-performance CPU that can manhandle its way through DCS, FSX, PlanetSide 2, ArmA 3, Star Citizen, all sorts of emulators, etc., but the technology just doesn't exist to allow that right now. The most we can hope for is for devs to optimize their software as best as they can with what we have today, and that's admittedly much easier said than done with the variety of hardware and software configurations out there.