Intel Core i7-4790K Devil's Canyon Review

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
PC Perspective has posted a review of the Intel Core i7-4790K Devil's Canyon processor. For more on Devil's Canyon, you should read our Intel Devil's Canyon Core Processor Presentation and our Devil's Canyon: The Good, the Bad, the Ugly article.

Performance of the Core i7-4790K is impressive, even if you choose to not overclock at all. The 500 MHz clock speed increase (at both base and Turbo speeds) when compared to the Core i7-4790K makes the choice between the two parts pretty much a no-brainer. Benchmarks clearly showed the advantage of the Devil's Canyon CPU and it ranges from 7-15% depending on the application and the threaded efficiency of the application.
Click to expand...
 
Hmmm. I'm running a i5-2500K at 4.3 GHz on air...so about a 20-25% improvement UNLESS the app is heavily threaded, where the i7 would show a bigger improvement.

Meh for now. Rather spend the money on a 1TB SSD.
 
%7-%15 is impressive? It's great and all, more performance, but that's not the word I'd use for a small incremental change. I think the impressive part is the 500mhz bump, it's like P4 territory without the accompanying low IPC and high power consumption.

But... imho seems like we're either starting to struggle to eek out more performance out of CPUs, or lack of competition is making things slow down significantly.
 
Could we be hitting the limit of the capabilities of the x86 architecture? When are we moving to ternary instruction sets :D?
 
The HUGE + on this is vt-d and tsx-ni on k series.

I hope to see more of this in the future.
 
Robstar said:
The HUGE + on this is vt-d and tsx-ni on k series.

I hope to see more of this in the future.
Click to expand...

^ this right here is really the major temptation for me. I've yet to need Vt-d for anything in my guests but, it would be nice to have.
 
Armenius said:
Could we be hitting the limit of the capabilities of the x86 architecture? When are we moving to ternary instruction sets :D?
Click to expand...

Nope. It's strictly an materials/fab problem. Today's "x86" cpus bear almost no resemblance at all to the '386/'486--and this has been true ever since AMD shipped the K7--or, for about the last 15 years or so...;)

Intel thinks "enthusiasts" are fundamentally idiots, and of course make up only a teeny-tiny portion of Intel's cpu business--so somebody in some Intel cpu department does a cherry-pick bin and *posts* @ 5GHz on air--and then the "Devil's Canyon" news is broadcast from the top of the Intel PR stack--even though it is almost completely pure fabrication (great word, here--fab-rication...;)) People really, really need to stop being slaves to MHz...it's 2014, after all.
 
Nope. It's strictly an materials/fab problem.
Click to expand...

I believe the problem is Intel can not yet get their 22nm process to have a performance improvement versus their 32nm process. 22nm tri-gate does well at low frequencies but does not seem to be any better than 32nm was at the high end.
 
Wierdo said:
But... imho seems like we're either starting to struggle to eek out more performance out of CPUs, or lack of competition is making things slow down significantly.
Click to expand...

Both of those scare me. Lack of competition means we pay more for less improvement. Getting less performance out of CPU's means that we're doing great now, but in 5 years we'll be doing just a little better... I wonder if the good thing would be programmers now start taking more advantage of it and optimizing things better... We have no much power and space that they just cobble things together. Before, when they were limited, they'd find tricks and such to make things work with less.

I think my biggest gains if I were to upgrade wouldn't be the actual CPU performance (it'd be a step up from my 2600K @4.4, but not huge), but the added features of the motherboard... :)
 
drescherjm said:
I believe the problem is Intel can not yet get their 22nm process to have a performance improvement versus their 32nm process. 22nm tri-gate does well at low frequencies but does not seem to be any better than 32nm was at the high end.
Click to expand...

Sounds about right--it's a typical fab problem. Often when we hear that this or that cpu is "xxnm"--say "22nm," that's really not true at all; generally it's actually a blend of 22nm & 28nm &32nm, and so on, in a precise formula known only to the manufacturer. The lower numbers are, of course, used for simplicity-in-marketing purposes. If yields at a certain process size improve then generally so does performance, etc.

What I don't like about this kind of marketing--"Devil's Canyon"--which is yet another barometer of the low opinion of enthusiasts that Intel generally has always had (basically because they want to do things with their cpus that Intel would rather they didn't), is that it is weird and condescending (Can you imagine Intel having called "Sandy Bridge" "Devil's Bridge" or "Satan's Sirocco"?....;) )
 
Intel doesn't really seem to make much effort to cater to enthusiasts, but that's probably because that market segment is a very tiny part of their business and, aside from some vocal members and minor press involvement, there's just not much to be gained from investing a lot of effort into it. They sell network cards, SSDs, and a big range of processors and chipsets along with doing other work. The high end gamer desktop CPU market is super teeny tiny and pretty unimportant when their agenda with CPUs is competing in power consumption and heat output as people who purchase their parts are just as concerned with battery life, power, and cooling as they are with performance. Hardly anyone is really worried about getting 108 instead of 102 FPS in a computer game thanks to overclocking. The days of soldering on a new timing crystal or playing around with voltage jumpers and prying off heat spreaders are pretty much long gone in favor of business, warranty, industry, and mobility.
 
You must log in or register to reply here.
Back
Top