Kaby Lake 7700K vs Sandy Bridge 2600K IPC Review @ [H]

FrgMstr

Just Plain Mean
Staff member
Joined
May 18, 1997
Messages
55,534
Kaby Lake 7700K vs Sandy Bridge 2600K IPC Review - There are many HardOCP readers that are still running Sandy Bridge CPUs and have been waiting with anticipation of one day upgrading to a new system. One of the biggest things asked in the last month is just how the 2600K stacks up against the new 7700K processor. So we got hold of one of our readers 2600K systems and put that question to the test.
 
interesting review and thanks nimisys for helping make this happen.. i'm actually quite surprised how small the gap is between sandy bridge and kaby lake is. it'll be interesting to see if ryzen reaches the potential AMD is hyping about and lights a fire under intels ass or if maybe intels starting to hit the wall on the cpu design they've been using since the i7 920.
 
This may not be very [H] of me, but what are the power draw comparisons?
 
This may not be very [H] of me, but what are the power draw comparisons?
I started to do those and I really just figured those really just did mean jack shit looking at the big picture.

My tools would not work on the 2600K to actually look at package and chip wattage and I did not want to compare two totally different motherboards.
 
Very nice work Kyle. You really hit the nail on the head by showing us that Intel has only delivered 20-25% in nearly 6 years of millions, possibly billion dollars worth of alleged Intel R&D time. It tells me two things, we NEED AMD to be competitive, and it also shows us the possibility that the current "Core" architecture is simply maxed out, and maybe we will get 5% IPC increase when the "next gen" get released.

Right now, it's AMD's time to make it, or break it, for all of us.

Another interesting point I would like to rase, is one that nobody, not even you have mentioned (to the best of my knowledge) Kyle, is that despite 6 years of architecture optimisation, as well as process and feature improvements/shrinks, Intel's CPU's are still just as hot and voltage hungry when moderately overclocking as they have always been. I have always wondered why this is the case? It used to be that when we had a process shrink, we could overclock the same but at lower voltage and then get lower temperatures, or overclock higher, and get the same sort of temperatures and voltages as before. Yet now, you get a process shrink or even two, and you get maybe 200MHz more overclocking at the same temps as before, and almost the same voltage, even Intel's new transistors (which I think have been changed 3 times since the 2600K) have not really changed this behaviour. We are still running just about any high end Intel CPU at or very close to 4.5GHz, and still getting the same kind of temps generation after generation. Sometimes I wonder if Intel really is doing any kind of work on these CPU's...
 
Another interesting point I would like to rase, is one that nobody, not even you have mentioned (to the best of my knowledge) Kyle, is that despite 6 years of architecture optimisation, as well as process and feature improvements/shrinks, Intel's CPU's are still just as hot and voltage hungry when moderately overclocking as they have always been. I have always wondered why this is the case? It used to be that when we had a process shrink, we could overclock the same but at lower voltage and then get lower temperatures, or overclock higher, and get the same sort of temperatures and voltages as before. Yet now, you get a process shrink or even two, and you get maybe 200MHz more overclocking at the same temps as before, and almost the same voltage, even Intel's new transistors (which I think have been changed 3 times since the 2600K) have not really changed this behaviour. We are still running just about any high end Intel CPU at or very close to 4.5GHz, and still getting the same kind of temps generation after generation. Sometimes I wonder if Intel really is doing any kind of work on these CPU's...
That is 100% true till you start looking at mobile CPUs. I really do not think there is a need to pull that in any more for high end CPUs on the desktop.
 
Great read. Makes me feel not so bad that I'm still on Lynnfield. Still pretty great for gaming alongside my R9 390 but I'll spring on Ryzen if Kyle shows that AMD is back in the fight!
 
Nice to have some more data to back up what we've all known for a while. The only reason I would tell anyone to shift from Sandy is for I/O reasons. If anything I'd be more interested in a 3770K comparison because that's where native PCI-E 3.0 came in and the Intel chipset dropped in native SATA3, on top of that Ivy had the extensions to AVX so I'm willing to bet most of that 20% would evaporate in the synthetics.
 
Hi Kyle - long time lurker here since coppermine/K7 days. Very excited about the upcoming VR CPU tests...

I think Project Cars is one game that a [email protected] will show a benefit over a [email protected] if using a 980ti OC or faster card.. at lowest settings/no weather I'm seeing 70-80% GPU usage but the CPU latency in oculus debug is showing right around 11ms with some spikes above causing the 45 fps fallback at times.

(Elite Dangerous stays well below 11ms always unless my 2600k is clocked to below ~4.3ghz).
 
Intel certainly isn't resting on their laurels.
...they've made a burrow, picked out some fluffy pillows, and have begun extensive hibernation.

Understatement.

I'm sitting on Haswell and not feeling compelled to do one thing about it. Talk about a generational sweet spot...

Given that we had several generations between Sandy (Ivy, Haswell, Braswell, Sky) and Kaby Lake, I'm kind of shocked to see that the single threaded gain is still in the single digits (9%) for some tests, multiply that by 4 (36%) and we're still not seeing equivalent scaling (closer to 25% on average) on a multiprocess level for the other tests...

Which really leads me to wonder that if all the improvements are not from architectural changes, but just physics: shrinking the process and rearranging the die. (Not to mention the slight RAM speed difference between the setups)

Yeah, this really is AMD's chance to shine. Hope they don't fuck it up.
 
Yeah, this really is AMD's chance to shine. Hope they don't fuck it up.

Indeed. After well over ten years since I last used an AMD processor in a build (at least for a system of mine) I'm really stoked to see what they bring to the table.

I'm trying to remember my last AMD based build, I want to say it was some flavor of Athlon 64 X2... I know it was Socket 939 and on a DFI Lanparty board. It was after that I turned to Intel's Core 2 and never looked back. It wasn't a brand loyalty thing, it was just that every single time Intel handed AMD's ass to them clock for clock.

Seeing all of us get their hopes up for Ryzen makes me remember the tech landscape ten to fifteen years ago and how there really was stiff competition. I'm excited to get a taste of that once again!
 
Very nice work Kyle. You really hit the nail on the head by showing us that Intel has only delivered 20-25% in nearly 6 years of millions, possibly billion dollars worth of alleged Intel R&D time. It tells me two things, we NEED AMD to be competitive, and it also shows us the possibility that the current "Core" architecture is simply maxed out, and maybe we will get 5% IPC increase when the "next gen" get released.

Right now, it's AMD's time to make it, or break it, for all of us.

Another interesting point I would like to rase, is one that nobody, not even you have mentioned (to the best of my knowledge) Kyle, is that despite 6 years of architecture optimisation, as well as process and feature improvements/shrinks, Intel's CPU's are still just as hot and voltage hungry when moderately overclocking as they have always been. I have always wondered why this is the case? It used to be that when we had a process shrink, we could overclock the same but at lower voltage and then get lower temperatures, or overclock higher, and get the same sort of temperatures and voltages as before. Yet now, you get a process shrink or even two, and you get maybe 200MHz more overclocking at the same temps as before, and almost the same voltage, even Intel's new transistors (which I think have been changed 3 times since the 2600K) have not really changed this behaviour. We are still running just about any high end Intel CPU at or very close to 4.5GHz, and still getting the same kind of temps generation after generation. Sometimes I wonder if Intel really is doing any kind of work on these CPU's...

i think the big change was when intel started combining their IGP and CPU's. while the processor it's self is shrinking the IGP isn't and is getting more complex which counters anything productive on the processor side. it never really did make sense on high end cpu's and am not sure why intel continues to do it other than to save money by not manufacturing 2 different sku's with and without the IGP.
 
Very nice article and it was nice to see the 2600K pulled back into service for a while. My only desire for Kaby Lake to perfectly honest would be the HEVC QuickSync hardware accelerated encoding aspects because HEVC takes so damned long to encode on anything but Kaby Lake at this point. QuickSync should make encodes damned fast indeed but the quality still won't match what a proper h.265 software based encode will be capable of but it'll be interesting to see how things work out. I haven't built a desktop class machine in a very long time, I use business class laptops nowadays (not a gamer so a big desktop just isn't a need) so I'll have to keep an eye out on Dell's offerings in the Latitude line for Kaby Lake support.

Would be nice to get a monster piece of hardware again after years of just using whatever crosses my path.
 
QuickSync should make encodes damned fast indeed but the quality still won't match what a proper h.265 software based encode will be capable of but it'll be interesting to see how things work out.
We covered QuickSync in our 7700K IPC review to compare it to the 6700K, and yes, it is a lot faster on the same workload compared to H.264 on CPU.
 
Well, i am actually impressed because now it looks like Intel has made even less progress than it seems since Sandy Bridge. Basically the largest gains are in usage of new instructions, little else.
 
Looking at the test results:
* The one test done to show single thread performance end in a tied result. (A 2% difference that can be related to different RAM.)
* All other tests are multithreading, with HT activated.

Question that comes to mind:
How much of the multithread performance difference comes from more efficient HT and how much comes from more efficient spread between cores?
(I suspect the difference would be decidedly less if comparing 2500K to 7600K in multithread.)
 
Excellent review and mirrors my research findings when contemplating a move from my 4.2GHz 2500K to "something" back in October 2016. Bottom line: unless you need multithreaded support (e.g., Handbrake and the like) there isn't a compelling reason to upgrade. The biggest bang for the buck is to use and SSD for the OS...I use one for the OS and one for games and transient video files for editing. *Those* two items made the biggest difference.

If you use multithreaded apps, the 6700K at 4.6GHz (on air, stock voltages) rocks it over 2x faster than my 2500K. So Kaby is really a fistful of 'meh'.
 
wow, I was expecting closer to 40%, not 25 maybe 30%...


AMD PLEASE DON'T BLOW IT LIKE YOU ALWAYS DO.

I had precisely the same expectations. Given the modest 3% to 5% average increase in IPC I expected just that when all lumped together.
 
Nice to see someone actually do the comparison. I'm still running a 3570K @ 4.2GHz that I got sometime back in 2012 with an R9-290X, 32GB RAM and a Samsung EVO 850 SSD on a Z77 motherboard. Every once in awhile, I get the bug to build a new rig, but never really find an incentive to. I've been starting to think maybe I was missing something, but I guess not.
 
Nice to see someone actually do the comparison. I'm still running a 3570K @ 4.2GHz that I got sometime back in 2012 with an R9-290X, 32GB RAM and a Samsung EVO 850 SSD on a Z77 motherboard. Every once in awhile, I get the bug to build a new rig, but never really find an incentive to. I've been starting to think maybe I was missing something, but I guess not.

The biggest thing you are missing out on is M.2 / NVMe based SSD's. That's pretty much it. I've got an EVO 850 in my laptop and I've used them on desktops before. They are not bad drives for the price but they don't hold a candle to the good NVMe drives. If gaming performance is your primary focus then the only benefit there are slightly faster level load times in some games. Disk I/O performance is a big deal for some usage scenarios as well but I think people who need that probably already made the jump or would if they could.

There are a lot of improvements in the newer motherboards. The UEFI BIOS is better, there are more redundancies in some BIOS ROMs, (BIOS Flashback, resume flash on power failure, etc.) additional tools and features such as secure SSD erase software being built in. Onboard fan control is vastly improved over those LGA 1155 based motherboards. You get reduced power consumption as well although this is a minor issue as the cost of powering a motherboard is almost nothing regardless of brand, model or generation. Sure PCIe, USB 3.0, 3.1 etc. are all great things if you are doing something to leverage those technologies. Most people really aren't going to outside of the occasional flash drive transfers. If you like built in WiFi that's significantly improved as well.

None of these upgrades are a big deal outside of storage I/O options and even the benefits of those are situational based on usage scenarios.
 
Now, I get that you want low video settings in games to see the big difference. But, you should throw in high res video settings too. Maybe limited to 1080p. It's important, Because, that's what someone running that old proc wants to know. They can immediately relate to those settings for their old setup. Will their 1080p setup see any improvement? It's nice to see if it does. Be it only a few FPS, they are looking for every reason possible to upgrade. Give them something to really compare to right away. :)

Great stuff tho, as usual Kyle.

Special thank you to Nimisys
 
Loved reading the comparison - GREAT ARTICLE! Clearly gives prospective for many here.

Also somewhat shocking the lack of CPU performance progress. Intel's time to up the core count probably should have been with Skylake from 4 core to 6 core. This is not compelling at all for most folks who have an Intel machine built in the last 5 years at the same type of cost structure. I can only imagine what Skylake E will cost and plateform - probably will be the best performers but at a rather steep price.

As for AMD, just waiting for the hard earned results of their endeavors - if they are good I will probably build a Zen setup replacing the FX.
 
Now, I get that you want low video settings in games to see the big difference. But, you should throw in high res video settings too. Maybe limited to 1080p. It's important, Because, that's what someone running that old proc wants to know. They can immediately relate to those settings for their old setup. Will their 1080p setup see any improvement? It's nice to see if it does. Be it only a few FPS, they are looking for every reason possible to upgrade. Give them something to really compare to right away. :)

Great stuff tho, as usual Kyle.

Special thank you to Nimisys
As noted, I am going to put that system into some VR tests this coming weekend. :) If it can make it there, it can make it anywhere.
 
My son has been toying with an upgrade and this is exactly the article I wanted to see. I've been telling him he wouldn't see much improvement over what he's already got (4770K).... But the upgrade bug is hitting him. I told him he'd be better off just investing in a newer video card (he's currently on a 970) and more memory (he's got 8) if he just really feels like he needs to spend some money right now.

Personally, my 5820 will probably still be in use 4 years from now unless something seriously changes.
 
I'm currently running a 2600k @4.0ghz and really don't want to upgrade but I want to move to matx and can't find a z77 motherboard at a decent price. So that might force me to make the jump to Kaby.
 
Well FFS!

...anybody got a link to a thorough Gigabyte Z68 bios guide?
 
My son has been toying with an upgrade and this is exactly the article I wanted to see. I've been telling him he wouldn't see much improvement over what he's already got (4770K).... But the upgrade bug is hitting him. I told him he'd be better off just investing in a newer video card (he's currently on a 970) and more memory (he's got 8) if he just really feels like he needs to spend some money right now.

Personally, my 5820 will probably still be in use 4 years from now unless something seriously changes.
Put all your money into GPU then display...then GPU....then display...
 
Honestly, the only reason anyone should be considering an upgrade at this point is if you need new motherboard chipset features like USB 3.0, USB type-C, NVMe... and (depending how old your Sandy Bridge motherboard is) PCI-E 3.0, and so on... Sandy Bridge motherboard chipsets aren't exactly top of the line anymore, you know? I think the motherboards have aged worse than the CPUs. Newer motherboards can run RAM at higher frequencies, and there are a ton of subtle improvements that don't seem like much at first, but then you really feel it when you go back to your hot, noisy Sandy Bridge system with slower gear.

I have yet to see Intel release anything that shoots down my theory that that, in the post-Sandy Bridge era, we'll be reluctantly pushed to upgrade for the sake of chipset features or hardware failures rather than being able to wait around for a better CPU. It's going to be more like replacing an appliance than like something where you're just constantly getting better and better stuff.

The only thing I feel the need to do on my system at this point is add another 8GB of RAM and install a GTX 1070. After that, it will be fine for gaming again. Graphics cards are still "alive," somehow. Although now I'm really just upgrading because I'm annoyed that newer games don't run any better on my PC than they do on the Xbox One, rather than because they're actually unplayable.
 
Last edited:
Well FFS!

...anybody got a link to a thorough Gigabyte Z68 bios guide?
I actually doubt that GBT even has that. You might go back through some of our GBT Z68 mobo reviews and see what you find there though.
 
Well then.

I'll continue to sit on my 2600K until it explodes (it seems to have degraded as 4.4 is no longer stable but 4.2 is). Sounds like I'll be hitting guides and reading to see if there are settings I should optimize to squeeze more speed out of it.
 
Back
Top