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

[/usr/sbin]

You can't be serious. I'd rather spend my money on a video card or nice display than shitty CPU that I put inside the box.

What can't I be serious about? That I upgrade when the performance difference is large enough and time is not a large factor? Give me 50% more performance and I'll upgrade. If it takes you 8 years to do it then you won't see a penny from me until that time. If you do it in 3 years then I'll upgrade then. Personally I weigh out how much more I'm getting for my dollar and for CPUs that is my threshold. Not saying you are wrong, we can all have our own personal criteria of what we think is "worth it".

I treat GPUs the same way. If the best Nvidia could do in 6 years is 25% more performance I'd still use the old GPU. Also, not everyone needs a new GPU or display. I already have a three monitor display setup I'm happy with.

 

[thesmokingman]

Did you guys run anything with real world settings?

It's good to show the IPC differences isolated but that doesn't show what one would experience in real world situations.

 

[WhoBeDaPlaya]

/usr/sbin : This exactly. I waited an eternity for my last upgrade (920 D0 -> 4770K), since it finally got close enough to +50% absolute performance
(comparing a 4GHz 920 to 4.6GHz 4770K ==> ~+43%).

Even then, the fact that it took ~5 years to get there is disappointing. I was expecting +115% (from the 4770K being a 6c/12t part).

Now we are looking at Round 2 of the same sh*t - in fact, it's even worse since the 2500/2600/2700K can clock just about as high as the 7600/7700K,
which means we are only talking about a max of +30% absolute performance assuming a slight clockspeed advantage with Kabylake.

Put another way - what can you do with DD b00bs (Kabylake) that you can't with D b00bs (Sandybridge)

 

[Peter2k]

What is the relevance of gaming benches at 640 x 480 res if as stated:



I realize these ultra low resolutions were pertinent 25 years ago, but they serve absolutely no purpose today.

didn't Kyle write he is going to do more test?
at the end?

real world tests
and VR

I'd think it's just a start to an extensive coverage of the differences
would take time
and many wanted to see something as fast as possible

 

[Highwind]

This is why I tell people looking to make a gaming pc but "don't have the money", they actually do have the money, it is in half-decade old CPUs. Now I can actually cite a source. And keep my 4690K forever.

 

[Ur_Mom]

Chances are you won't see any difference between standard SSDs and newer NVMe drives.

Come on, man. There has to be a reason for me to upgrade. I need a new GPU for sure (I'm on an AMD 7950, which does fine most of the time). But, for everything else, I'm doing just fine. But, I would like to be able to update and see the words I type in Word come up faster... Or something. I don't care.

I want to upgrade, but I don't want to throw my money away.

 

[WhoBeDaPlaya]

This is why I tell people looking to make a gaming pc but "don't have the money", they actually do have the money, it is in half-decade old CPUs. Now I can actually cite a source. And keep my 4690K forever.

You're in the same boat as me since Intel's next IPC increase is ~2018-2019, except not having HT might bite you sooner.
Hopefully that will be a >= 6c/12t part that clocks >= 5GHz, otherwise the upgrade wouldn't be worth it.

Come on, man. There has to be a reason for me to upgrade. I need a new GPU for sure (I'm on an AMD 7950, which does fine most of the time). But, for everything else, I'm doing just fine. But, I would like to be able to update and see the words I type in Word come up faster... Or something. I don't care.

I want to upgrade, but I don't want to throw my money away.

7950 is a getting a little long in the tooth, but you can pick up a used non-reference 290 / 290X for stupid cheap (~$110 on CL) and be fine for quite awhile at 1080p.

Unless you have a workload that demands it, only your wallet will notice the difference between an NVme and SATA SSD :
https://hardforum.com/threads/nvme-...a-iii-ssd-game-load-time-comparisons.1911914/

 

[harddud]

I built two systems last summer ... and used Sandy and Haswell for them. The lack of competition from AMD IMHO has caused Intel to have a brain fart ... OR they just don't care OR Mohr's law has kicked in.

Many discuss why the hobby is "dying". The hobby is not dying ... it is just hobbled by the fact (IMHO) that there is a severe lack of competition in the industry. Why are PC sales down ... look no farther (IMHO) to the lack of competition. Blame Intel for sitting on their laurels ... or AMD for being incompetent ... but YOU, the hobbyist are the one that pays for it.

In My Humble Opinion ...

 

[Iching]

I built two systems last summer ... and used Sandy and Haswell for them. The lack of competition from AMD IMHO has caused Intel to have a brain fart ... OR they just don't care OR Mohr's law has kicked in.

Many discuss why the hobby is "dying". The hobby is not dying ... it is just hobbled by the fact (IMHO) that there is a severe lack of competition in the industry. Why are PC sales down ... look no farther (IMHO) to the lack of competition. Blame Intel for sitting on their laurels ... or AMD for being incompetent ... but YOU, the hobbyist are the one that pays for it.

In My Humble Opinion ...

PC sales have nothing to do with Intel. Millennials and younger people, in general, are computer illiterate apart from the small number who obviously is interested in it. People prefer to spend an exorbitant amount money on some overpriced Pixel or iPhone smartphones since it's a status symbol. It's not like you're going to walk with your expensive new CPU and display around the City showing off.

 

[WhoBeDaPlaya]

It's not like you're going to walk with your expensive new CPU and display around the City showing off.

That's what we used to do, back when men were men...

 

[Xebec_]

What is the relevance of gaming benches at 640 x 480 res if as stated:

I realize these ultra low resolutions were pertinent 25 years ago, but they serve absolutely no purpose today.

The relevance is that they show the CPU in a situation where it's not bottlenecked by GPU.

Some day there will be a powerful enough GPU that same thing will be true at 1920x1080, 2560x1440, etc with that CPU showed today.

An i5-2500K at launch didn't seem to benefit from faster RAM at desktop resolutions. Take a look at some recent reviews (eurogamer) and you'll see upgrading the RAM (i.e. 1333 --> 1866) on 2500K gives noticeable benefit even at 1080p/1440p when you have a modern/fast card such as a GTX 1070 or higher..

 

[Araxie]

The relevance is that they show the CPU in a situation where it's not bottlenecked by GPU.

Some day there will be a powerful enough GPU that same thing will be true at 1920x1080, 2560x1440, etc with that CPU showed today.

An i5-2500K at launch didn't seem to benefit from faster RAM at desktop resolutions. Take a look at some recent reviews (eurogamer) and you'll see upgrading the RAM (i.e. 1333 --> 1866) on 2500K gives noticeable benefit even at 1080p/1440p when you have a modern/fast card such as a GTX 1070 or higher..

the relevance is null at this age as there are settings that actually affect badly the CPU performance even more than a resolution change.

aaand... we are actually in some games in that "day" where CPU bottleneck exist at 1080P with maxed settings..

Hey, and that's is a 4.5ghz i7 6700K..

 

[Iching]

That's what we used to do, back when men were men...

I know but peeps nowadays couldn't be bothered with those things. It might be too much hassle.

 

[Iching]

the relevance is null at this age as there are settings that actually affect badly the CPU performance even more than a resolution change.

aaand... we are actually in some games in that "day" where CPU bottleneck exist at 1080P with maxed settings..

Hey, and that's is a 4.5ghz i7 6700K..

I upped my resolution to 3440sx1440. Problems solved.

 

[harddud]

PC sales have nothing to do with Intel. Millennials and younger people, in general, are computer illiterate apart from the small number who obviously is interested in it. People prefer to spend an exorbitant amount money on some overpriced Pixel or iPhone smartphones since it's a status symbol. It's not like you're going to walk with your expensive new CPU and display around the City showing off.

PC sales have a lot to do with Intel. The lack of performance improvement from Sandy to Kaby is indicative of stagnation ... or Mohr's law has kicked in. Why buy/build a new PC when my 5 year old one works just fine ... and the performance improvement (of the new) would be only 25%? I believe that one of the points that Kyle was trying to make was that.

 

[Hitti2]

I love my wc'd oc'd delidded 3770k.

I see no reason to upgrade unless I'm video editing. Which is not to many times. Handbrake does peg all cores n threads on my [email protected]. But for the normal, it an awesome CPU.

 

[Willseyvilleny]

Kyle,

I just read this post. It's good to see that someone is showing how useful this old hardware is. You really should do a complete comparison between 1366 1155 and 1150 processors as well. I will donate some hardware to the cause if need be. I have a hex-core 1366 system on a rampage III Gene as well as an intel i7 950, as well as a 3570 system that I am not using. Let me know if you could use these. I have put everything from GTX 260's to GTX 1080s and 5870's to r9 390x's in multiple configurations and I just can't seem justify a Skylake build. I know HardOCP is huge on playable game settings. It's the reason why I have been a member for over a decade. In all reality for a majority of us, there just isn't a real world performance gain on some of these newer processors. I'd much rather spend the cash on faster SSD's and GPUs.

 

[Max9R]

I've been talking friends and family members with anything Sandybridge and newer out of CPU upgrades for years (in some cases even Bloomfield). SSD's, RAM, and graphics cards is where their money has been spent instead. My 3770K @ 4.8 was still serving me fine up until I sold my 780ti's early last year. I still have it, it's just resting in my Z68 board until I put a GPU back in it this year (hoping for some decent GPU competition in 2017). Frankly, I've had zero interest or desire to spend money on a CPU since I bought it because the reality is there just hasn't been anything to get excited over in the CPU world in a long time. Hopefully Ryzen will change that... time will tell.

 

[SvenBent]

What is the relevance of gaming benches at 640 x 480 res if as stated:

I realize these ultra low resolutions were pertinent 25 years ago, but they serve absolutely no purpose today.

You basically quoted the text that explain you why it is run this way. we are looking at CPU performance. how much is it worth looking at GPU benchmarks then?

 

[raghu78]

With rumours of Intel 10nm Cannonlake being delayed to mid 2018 and that too only for ultra low power and Coffee Lake 14 nm being confirmed for 2018 for desktops and notebooks we are going to see 10nm for desktops in 2019 . We are probably going to see the next big architectural improvement (tock) from Intel by late 2019. AMD should use this golden opportunity to catch up and overtake Intel. Mark Papermaster has stated that we will see tock tock tock after Zen. So there should be good IPC improvements in 2018 and 2019 along with higher frequencies due to process maturity. Intel is going to have a tough time for being complacent and downright indifferent towards the enthusiast market.

 

[amenx]

The relevance is that they show the CPU in a situation where it's not bottlenecked by GPU..

Nobody, I repeat NOBODY, games at 640 x 480 TODAY. Therefore 640x 480 is a useless bench for those wondering how the CPU practically affects their gaming performance when on a mainstream res of 1920x1080. May as well just state to readers to forget about CPU gaming performance if on 1080p or above (even if not true).

 

[thesmokingman]

Nobody, I repeat NOBODY, games at 640 x 480 TODAY. Therefore 640x 480 is a useless bench for those wondering how the CPU practically affects their gaming performance when on a mainstream res of 1920x1080. May as well just state to readers to forget about CPU gaming performance if on 1080p or above (even if not true).

You're missing the point and just repeating it shows that you didn't understand the point in the first place. This review isn't a reflection of real world gaming. It is a test to isolate the cpu by moving the load from the gpu to the cpu as much as possible. This allows [H] to shine a light on the IPC difference between the two cpus.

 

[amenx]

You're missing the point and just repeating it shows that you didn't understand the point in the first place. This review isn't a reflection of real world gaming. It is a test to isolate the cpu by moving the load from the gpu to the cpu as much as possible. This allows [H] to shine a light on the IPC difference between the two cpus.

Except that all [H] CPU reviews show gaming benches with 640x480 as standard practice, incl the proper 7700k and 6700k reviews. This can mislead many uninformed folks who may not be privy to resolution impact on CPU gaming performance. This I believe is just a silly carry-over from the days when 640x480 was indeed more relevant. Most sites now do CPU benches @ 1080p as standard practice, some even include 1440p (guru3d, TPU) and you may be surprised at results (even though minimal) in some games.

 

[Shintai]

With rumours of Intel 10nm Cannonlake being delayed to mid 2018 and that too only for ultra low power and Coffee Lake 14 nm being confirmed for 2018 for desktops and notebooks we are going to see 10nm for desktops in 2019 . We are probably going to see the next big architectural improvement (tock) from Intel by late 2019. AMD should use this golden opportunity to catch up and overtake Intel. Mark Papermaster has stated that we will see tock tock tock after Zen. So there should be good IPC improvements in 2018 and 2019 along with higher frequencies due to process maturity. Intel is going to have a tough time for being complacent and downright indifferent towards the enthusiast market.

Coffee Lake is 14nm++, yet another node revision just as others do. 10nm isn't mid 2018 either. The Icelake desktop will be using 10nm+ on a second generation 10nm.

And seems fairy tales and unicorns have struck again for the rest.

 

[Shintai]

You're missing the point and just repeating it shows that you didn't understand the point in the first place. This review isn't a reflection of real world gaming. It is a test to isolate the cpu by moving the load from the gpu to the cpu as much as possible. This allows [H] to shine a light on the IPC difference between the two cpus.

The problem tho is there is another bottleneck present for at least one of the CPUs. Memory bandwidth.

Just as you could use a 5775C and then claim high IPC increase, yet its due to something else.

 

[Stimpy88]

The problem tho is there is another bottleneck present for at least one of the CPUs. Memory bandwidth.

Just as you could use a 5775C and then claim high IPC increase, yet its due to something else.

I would love to see memory bandwidth equalized between the CPU's/Platforms that Kyle tested.

It could be fun to rule out memory bandwidth advantages to the newer CPU. Maybe using Sandra or another synthetic memory benchmark to get same or similar memory scored between them would be quite enlightening...

 

[zone74]

Nobody, I repeat NOBODY, games at 640 x 480 TODAY. Therefore 640x 480 is a useless bench for those wondering how the CPU practically affects their gaming performance when on a mainstream res of 1920x1080. May as well just state to readers to forget about CPU gaming performance if on 1080p or above (even if not true).

Resolution has little-to-no bearing on CPU performance.
The only thing testing at higher resolutions will do is cause the test to be GPU-bound rather than CPU-bound, reducing the framerate and invalidating the test.

Let's say that you have three CPUs.
When tested at 640x480, GPU load is below 50% at all times.
CPU1 scores 200 FPS, CPU2 scores 150 FPS, and CPU3 scores 100 FPS.
When tested at 4K with a GTX 1050 GPU load hits 100% and all three tests give a 50 FPS result.

Does the 4K test tell us anything about the CPUs, or does it only tell us that a GTX 1050 is not enough for 4K gaming?
Is that second test a measure of CPU performance at all?


Now let's say that we did a set of GPU benchmarks at 4K using the fastest CPU we have available (CPU1) in all tests.
A Titan XP scores 200 FPS
GTX 1080 scores 175 FPS
GTX 1070 scores 150 FPS
GTX 1060 scores 100 FPS
GTX 1050 scores 50 FPS

So if we did our CPU testing again at 4K with these cards, the results would be:
Titan X P : 200 / 150 / 100 FPS
GTX1080: 175 / 150 / 100 FPS
GTX1070: 150 / 150 / 100 FPS
GTX1060: 100 / 100 / 100 FPS
GTX1050: 050 / 050 / 050 FPS
Good Result / Indeterminate Result / Bad Result

At 4K resolution, only the Titan XP shows us the real performance of these CPUs - and that's assuming that GPU load in the Titan XP test remains below 100% at all times.
Testing with a 1080 instead of a Titan XP tells us that CPU1 is faster than CPU2, but does not tell us how much faster it really is.
Testing with a 1070 tells us that CPU1 and CPU2 are faster than CPU3, but shows no difference between CPU 1 & 2.
Testing with a 1060 or 1050 shows us no difference between the CPUs at all - though the performance level is different with both GPUs.

If you tested at 640x480 instead of 4K to eliminate the GPU load, the result should be 200/150/100 FPS, regardless of which GPU is being used to run the test.
The difference is that the 1050 might be at 50% GPU load, while the Titan XP would only be at 12.5% GPU load during testing - but as long as it never approaches 100% GPU load, the results should be the same across all GPUs.

 

[amenx]

Fair enough. I guess [H] will just have to stand proudly as the only site left in 2017 with the 640x480 benches. And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

 

[zone74]

Fair enough. I guess [H] will just have to stand proudly as the only site left in 2017 with the 640x480 benches. And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

Did you not understand my post?
Increasing the resolution tells you nothing about CPU performance.
The only thing it does is make the test more likely to be bottlenecked by the GPU.
If your test is affected by the GPU at all* it's not a CPU test.

I would probably select newer games to test though, where the difference is going to be that one CPU runs the game below 60 FPS, while another is above it, than older games where the results are 300 FPS vs 360 FPS.



*The exception being across GPU vendors, since there are different driver overheads.
AMD will perform worse than NVIDIA for any given CPU in DX11 titles.

 

[ChoGGi]

And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

higher resolutions = GPU factoring into the difference
If you want to see the differences between CPUs then you use low resolutions (baseline numbers, not gaming numbers).
This is trying to show the IPC differences between Sandy and Kaby, but you already knew that.

 

[WhoBeDaPlaya]

Fair enough. I guess [H] will just have to stand proudly as the only site left in 2017 with the 640x480 benches. And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

The point was not to show the CPU impact in a realistic scenario, but the maximum "theoretical" impact

 

[amenx]

The point was not to show the CPU impact in a realistic scenario, but the maximum "theoretical" impact

Thank you for jumping straight to the heart of the matter. I (and many others, review sites) are more interested in 'realistic scenarios' rather than the theoretical. Just find it humorous that others have not been able to grasp that and basically insisting that the theoretical is more valid because it shows bigger differences. Differences that are of no use to anyone for practical purposes whatsoever.

 

[SpeedyVV]

We sure are a queer bunch!

Anyone one else that buys something that does not get obsolete quickly is very happy.

[H]ardOCPers, want quick obsolescence! I know I do.

 

[thesmokingman]

Except that all [H] CPU reviews show gaming benches with 640x480 as standard practice, incl the proper 7700k and 6700k reviews. This can mislead many uninformed folks who may not be privy to resolution impact on CPU gaming performance. This I believe is just a silly carry-over from the days when 640x480 was indeed more relevant. Most sites now do CPU benches @ 1080p as standard practice, some even include 1440p (guru3d, TPU) and you may be surprised at results (even though minimal) in some games.

Misinformed, like yourself?

 

[lolfail9001]

And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

Everyone else interested in CPU impact on gaming is not actually interested in IPC comparisons.

Folks complaining about resolution choice are basically complaining that they have found apples in bag of apples... because they were expecting oranges there!

 

[WhoBeDaPlaya]

Thank you for jumping straight to the heart of the matter. I (and many others, review sites) are more interested in 'realistic scenarios' rather than the theoretical. Just find it humorous that others have not been able to grasp that and basically insisting that the theoretical is more valid because it shows bigger differences. Differences that are of no use to anyone for practical purposes whatsoever.

Fair enough, but the same can be said for a lot of synthetic benches.
Nothing wrong with having a complete picture since I do believe Kyle mentioned that more realistic resolution benches are forthcoming?

 

[ncknck]

Hello, a great review.

IPC is an incredibly important stat in apps that are single-threaded. There are tons of singlethreaded apps around (AMD's financial status is a good show of that). So here is an idea, how about benching these apps? That will give provide both a real world and IPC performance gains.

There are 3 major kinds of apps that require IPC.
-Bad console ports:
Final Fantasy F13
Final Fantasy F13-2

-Console Emulators
Dolphin Benchmark

-Games that are actually strictly single-threaded
Dwarf Fortress
Stellaris


i personally was looking for a cpu that would give me best performance in all those games.

 

[pendragon1]

Fair enough. I guess [H] will just have to stand proudly as the only site left in 2017 with the 640x480 benches. And everyone else interested in CPU impact on gaming with resolutions relevant to them (1080p) will have to look to the myriad of other sites that provide their answer.

you haven't really read any [H] reviews have you? they test at appropriate res for the hardware(1080, 1440, 4k) and when trying to show cpu difference and take gpu out of the equation, they use 640x480.

Misinformed, like yourself?

perhaps try and be a bit more useful with you post next time, maybe why hes misinformed...

 

[jbltecnicspro]

I see this as a double-edge sword. For those of us who like to be on the bleeding edge, Kaby Lake isn't anything to get excited over. However, for those of us who can't afford to be on the bleeding edge (yours truly), I'm okay with shitty IPC increases. Still rocking a Core i5-3570 here and I'm still GPU limited.

 

[Nimisys]

Thank you for jumping straight to the heart of the matter. I (and many others, review sites) are more interested in 'realistic scenarios' rather than the theoretical. Just find it humorous that others have not been able to grasp that and basically insisting that the theoretical is more valid because it shows bigger differences. Differences that are of no use to anyone for practical purposes whatsoever.

Go read the first paragraph of the Conclusions Page again:

First and foremost, of all the benchmarks we have run through here today, none are GPU limited, which is by design. So we want to follow up with some real gaming comparisons. In the next week I am going to be getting back into our VR GPU coverage. I have been quite busy with Kaby Lake and the Z270 chipset for the last month and a half. Since then we have seen a lot of solid VR titles come out that we surely need to cover. I am going to change up our platform a bit and move to the 7700K as our go-to CPU for VR gaming reviews. Nimisys has given me the OK on keeping his Z68 motherboard and 2600K CPU for a bit longer, so we will do some 7700K vs 2600K in VR gaming and see what performance differences there are, if any

real world, real resolution benches are coming. the point of this article was to show how much of a difference there was between Sandy Bridge and Kaby Lake, thus the reason it was labeled an IPC review. it was not intended, nor written, as a review of the platforms as a whole. a lot of other sites do an overview of the platform and call it good, Kyle and Co, take a deeper look at various areas of the platform. this was focused on the strength of the chip alone.