Intel Kaby Lake Core i7-7700K, i7-7700, i5-7600K, i5-7600 Review

Page 4 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.

So your alternative to a mass market product like the 7700k is to get a specific revision/ES Xeon, which by your own admission:
-You'll be lucky to find on Ebay (and you yourself only got your hands on one because you have connection at HP)
-Certain features may be broken and/or overclocking may or may not work, depending on which chip you get, which you may not be able to tell on your own

Also, I'm guessing you may be talking about older generation Xeons, which means that difference in clock speed + IPC would mean reasonably lower single threaded performance compared to Sky/Kaby Lake, in exchange for more cores/threads than 99% of people have any use for.

Yes, clearly the 7700k/Ryzen are pointless when such a practical and readily available alternative as this exists for the masses /s
 

He doesn't realize regular people who don't ever need 2 and a half dozen threads actually exist. They don't just exist, they outnumber the people who do need that many threads by maybe thousands to one. Calling AMD's next gen of chips and Intel's current gen of consumer chips a "waste of money" when he probably games on that rig 90% of the time (1080 SLI) and he is only using 4 of those 14 cores made me chuckle. Which one is more of a waste of money? Using the ol' car analogy, that is like saying "why should a mom and dad with 4 kids get a brand new mini van when they can get a used school bus for the same price that seats 28 kids and can even be souped up to go just as fast?" Certain vehicles for certain applications, that's why. Certain CPUs for certain applications. Doesn't matter if the prices are similar, it doesn't make sense for the masses which mainly consists of light gamers, web browsers and people who make spreadsheets and powerpoint presentations.
 


I use it for other things not only gaming by doing virtualization, web and application development, testing, video encoding and decoding. I was just talking about enthusiasts. Either way $289 was not a waste of money cause it cost me less than any i7 or upcoming AMD Ryzen.....why not? I just found a way to save money. In my case 7700K, 5960x or AMD Ryzen is a big waste of money. ASRock x99 board i got on sale was $160 + unlocked Xeon 14/28 for $289 and i can use regular quad channel GSkill DDR4 memory on it. Pretty f. good deal to me.

There are masses who don't know any better and pay extra and they get less...and there are people like me who pay less for more.
 
Just because your use is for a Xeon doesn't make it good value, in fact as per Linus one video a $300 pasted a $3000 server setup in just about anything. Unless you are running a 50 workstation rendering firm a Xeon is about as usefull as a bag of nuts.
 


The problem with the Linus Xeon was clock speed. A 4.0ghz Xeon would not have any trouble doing anything.
 


Exactly. I would push more but there would be a problem of heat, better water cooling would do a job. Also i use Google Sketchup. Linus is not really relevant when it comes to tech stuff...i like a guy, he is funny but that's about it beside moving hands so much around.
 
I have a question if i update my H110-ITX BIOS to accept 200 series motherboards could i then buy the 7700 and sell my i5 6500?
swap out my 2x4gb ddr4, for 2x16gb Ram, swap out my itx gigabyte gtx 970 for a gtx 1070 gigabyte itx. Get a Blue snowball M
 
If Intel is hoping to bring a fight to Ryzen, they better speed up the development of CannonIntotheLake. Right now this new intel cpu is not a decent upgrade to my I5 4670K @ 4.2Ghz. Looking at Ryzen for a decent upgrade, but as the specs for ryzen are not very clear yet, guess i will take both Intel and upcoming Ryzen cpu with a truckload of salt....
 

Oh, you didn't specify that you were only referring to your specific needs, we thought you meant they were a waste in general. But, even then, you are talking about a USED CPU that isn't meant to be overclocked. It is pretty much a write off if anything happens to it while, on the other hand, a brand new unlocked i7 has a 3 year warranty and is covered with the optional tuning plan if you OC it. Yeah, your used CPU was a great value but the typical enthusiast is not going to go dumpster diving on ebay or craigslist for PC parts and then pray that they keep working each time before they push the power button. If your CPU was $300 brand new, that is a different story. But, that thing was probably more like $3,000 new so it is pretty much irrelevant to even bring it up when talking about consumer level chips. I'm sure most people "know better" and would still rather buy a new i7 for $300 or so instead of a used 14 core Xeon for the same price. I could do the same thing and sit here and point out that you can get a used i5 2500k for like $100 and call anyone buying a new i3 an uninformed idiot. But, that is just as irrelevant because a lot of the price of a new chip is the warranty and the piece of mind knowing that you are going to have a working computer for at least 3 years after you first buy it.
 

Pretty much this.

I don't necessarily agree that buying a used CPUs is particularly risky; they tend to last a long time. The thing is that making a broad recommendation to buy a used part that you got at a good price assumes that everyone can find it, at that price, with relative ease. And that there's enough of them at that price that a reasonable amount of people could actually make use of your advice.
 
Good review, I'm a little surprised with the voltage and heating issues. Personally I don't see any reason to update my Ivy Bridge it runs everything great. What would be cool and may have already been done is to see bench marks on most of the i5 Generations from the earliest versions of the Sandy bridge to the latest Kaby lake.

I heard a rumor perhaps someone here can confirm it. If you upgrade your system to Kaby CPU and you did the Windows 10 free upgrade you'll need to buy a new copy of Windows 10 if you upgraded the motherboard. I would expect this for OEM system builder versions but many who did the upgrade had retail O/S versions that should be usable on the new motherboard.
 


I had this concern before upgrading from windows 7 to 10 and did a live chat with Microsoft. I was told that when the upgrade was done, Windows 10 inherited the rights of the product that was upgraded. I have 3 machines upgraded, 2 retail and 1 OEM. I fully expect to be able to move the retail license if/when I upgrade.
 
When the time comes (if ever, lol) let us know how the retail versions fare.

 


There is an unofficial way to run WMC on Windows 10 posted on the My Digital Life forum. There is a warning that "W10 Anniversary update broke Cable Card tuners and DRM. No solution has been found as of sept 24th 2016," but it might still be worth a try if you don't need that. You do need to login to the forum in order to see the post.
 


The same could be said for every release since Sandy Bridge.

Coupla notes tho.....

1. Microsoft Word and Excell... really ? The difference in the "scripted workload" in Word is 0.11 seconds (0.14 in Excel) for the best and worst Intel CPU in the test. It would love to see an article examining the relationship between benchmarks and what I like to call "reality". Reality would involve putting a person at the KB who would have to provide input between each of those "scripted tasks".... them measure the time it takes for someone to complete all tasks. If you could even measure a difference, it's npot as if this will impact a business' bottom line.

2. AutoCAD 2D / 3D has historically been always been faster on GeForce cards than Quadro.

3.
There are a lot of observable fluctuations during the extreme stress test. This is because the CPU runs into its thermal limit in spite of our compact water cooling solution, and it throttles accordingly for safety reasons.

While the radiator and fans may be capable of 300 watts, my assumption would be that the cooler, like all CLC's, cooling is limited by the blocks inability to move heat off the die fast enough because of inadequate flow rates.

4.
We learn two things from our experiences with Intel's Core i7-7700K. First, Intel’s Power Thermal Utility is no joke.

I think that depends on what your goals are....

a) The goal is to create an artificial condition by hammering the CPU with single task loads designed to stress the CPU thermally as much as possible... a condition which the system will never see again in its lifetime. If doing cooler comparisons, I see this as a logical usage. If testing for stability or setting my overclocks as high as possible without exceeding thermal of voltage limits, this has no value. In fact it is self-defeating. You may wind up with 4.8 Ghz when without the load that the system will never see again, you might be able to get 4.9 or 5.0

b) Your goal is to create the highest possible loading the CPU might actually see in it's lifetime, and to test for stability in a multitasking environment. I have had 24 hour stable synthetic OCs fail when tested under application based benchmarks like Rog Real Bench.

5.
In light of our experiences, early adopters should be aware that CPU quality can vary widely, and this is especially true for early production runs.

Soooo true tho this is the 1st time I have seen it so bluntly stated in a published review. Those rushing out to be the 1st one on the block to have the new tech, seem immune to recognizing the risks involved. Living on the bleeding edge means:

a) CPU in later production runs will, on average of course, subject to normal lottery variations, be faster, more power efficient and run at lower voltage / temps than later productions runs.
b) You MoBo will have to go thru numerous BIOS upgrades before all issue get addressed, resulting in redoing all your overclocks each time.
c) You will get stuck with various issues that are corrected in later steppings. I waited about 3 months for Asus to fix an issue with external devices not waking up from sleep ... it was not a BIOS change that fixed it, it required a board revision (C1 stepping).
d) You don't get to see which MoBo designs stand out and which ones are clunkers because you didn't wait the the bleeding edgers to finish beta testing and discover which is which.

6.
The Corsair H100i v2 in our U.S. lab gave us sporadic issues and generally didn’t cool as well.

I understand the logic (easy install) of using a AIO for testing purposes, but as the H100i doesn't do as well as even a NH-D15 or Cryorig Ultimate air cooler, I think you need to reach a bit higher. An EK Predator or better yet Swiftech H240- X2 would move the heat off the block much more quickly. When we used our test rig to vary pump flow (Swiftech MPC 35x2 dual pump), we saw rapid temperature improvements with increases in pump flow which started tapering just below 1.0 gpm and were nearing flat at 1.3 gpm.... tho never quite got horizontal. The 0.11 gpm in the H100i is just too far below the accepted flow range recommended for optimal cooling
 
Cooling:
We also tried the Noctua and it was even worse. I also checked the i7 7700K with a good custom open loop (brandnew Alphacool), that was able to cool down my 5960X @5GHz and the 6950X@4.4 GHz without any problems. Same issues with the KL. Ok, it is cooler, but not optimal. The Corsair AiO is fun, sure, but in each case better than all this air coolers. The reason is the high density (small hot spot) and the very slow heat transport over the heatpipes. I monitored this with my Hi-Res IR cam in time lapse. Intel must solder the heatspreader to get better results, that's all.

Mainboard:
You're wrong. I'm in a very close contact to the BIOS guys and the experience from Z170 and Skylake was used for this (very similar) CPU and Chipset. The memory speed was this time in the focus of the developers, nothing else. The problems are the CPUs and the very different quality. A new (and cheaper) production process also means, that it makes no sense to buy such silicon as early adaptor. Maybe you're lucky, but mostly not. I talked with guys from a shop and they proofed 15 more KL and found only one good sample. All others were junk. Mainboard stability is one thing but in this case a new BIOS will not fix production issues and a bad chip quality. And to be honest - I have a good reason to not use any Asus mainboards after the 3rd or 4th BIOS revision.

BTW:
I'm testing stability over 3 hours (and more) with Creo or Solidworks and a virtual machine in background. This causes more issues than any games or synthetics. Prime is a joke and nothing to believe. AutoCAD is good for GeForce since they moved to DirectX. But I know a lot of AddOns and scripts (simulations), where the similar Quadro is faster :)


 
i7-7700K, ASUS Strix Z270 mobo, Cooler Master Hyper 212 EVO air cooler, and 32GB of Corsair 3000 C15 RAM. With the EZ overclock utility from the BIOS, I have 4.84GHz overclock and 2880 memory at a maximum temp of 80C. 4.84GHz on air with a $40 cooler. I am ecstatic over Kaby Lake's great performance.
 
So you are lucky with your chip :)

One of my own 7600K runs up to 5.1 GHz with air (be quiet! Dark Rock Pro III), the official samples only 4.9 and 5 GHz under water. But mostly you can find only junk in the stores, the best CPUs of the first batches were binned for special events and promo. We tried again to select in the store from what the guys got officially and it was not the bomb. I would suggest everybody to wait a few weeks for newer batches.

BTW: The Asus OC utility gives every time too high voltages. Try to reduce the Vcore a little bit to get it cooler and use only all four cores with a max Turbo Core frequency of 4.8 GHz and forget the stupid multiplicator.
 
Some mainboard manufactures are providing now modified onboard drivers, I saw this also.
But this isn't official distributed as final drivers from Intel. The iGP in Skylake is the same as as in KL.
 
Status
Not open for further replies.