7800x vs 7700k, hint: the 7800x is slower(even when overclocked)

[Valentyn]

6 minutes ago, tom_w141 said:

Min FPS is a horrible metric as spikes can happen. 0.1%, 1% and avg are the best ways to display fps. But yes I focus on 0.1 and 1 more when comparing because a high avg is a high avg doesn't matter if its 100 or 140 you won't notice. What you will notice is shitty 0.1 and 1%s.

 

That's what I was referring to.

I find it really annoying when most graphs only look at AVG and Max FPS for the rankings; but then 1%'s show a completely different story often.

 

I have a big habit of gaming when my projects and work renders or encodes and if the mins are terrible just gaming, they'll be horrific under pressure often.

[Valentyn]

2 hours ago, mynameisjuan said:

2 less FPS on average at over 100+ FPS and the world loses its shit.

 

Cant we all just be happy that the HEDT are matching consumer i7s for gaming? 

Were you here for the Ryzen launch reviews? Went mental.

Also in many of the tests the 7800X was doing really badly on 1% minimums, even when overclocked compared to Stock 7700K. That really affects gaming performance and consistency of experience.

[ivan134]

Why are they testing at 1080p?

[Enderman]

5 hours ago, tom_w141 said:

You realise the 1.79% difference between 182.96 and 179.74 is well within margin of error right? If Kaby lake had higher IPC then skylake then wouldn't Intel be shouting it from the heavens?

"it's margin of error!!!!11!"

"margin of error this! margin of error that!"

"2% must be margin or error! I can't tell a difference!"

 

Something you might wanna learn: http://www.owlnet.rice.edu/~labgroup/pdf/Error_analysis.pdf

Maybe then you'll understand why two decimal places are included.

[tom_w141]

17 minutes ago, Enderman said:

"it's margin of error!!!!11!"

"margin of error this! margin of error that!"

"2% must be margin or error! I can't tell a difference!"

 

Something you might wanna learn: http://www.owlnet.rice.edu/~labgroup/pdf/Error_analysis.pdf

Maybe then you'll understand why two decimal places are included.

lmao you are still going.

 

You realise by margin of error we mean that 1-2fps variance is likely between tests. e.g. if I bench a 6700k 3 times I'm going to get 3 slightly differing results. I'm not going to get 150 3 times, I might get 147, 150 and 152. A couple frames either side is acceptable error.

 

Something you might wanna learn: https://www.pcper.com/reviews/Processors/Intel-Core-i7-7700K-Review-Kaby-Lake-and-14nm/Clock-Clock-Kaby-Lake-Skylake-Broad

 

Seriously let it go you're making yourself look like an idiot. Quit while you are behind.

 

EDIT:

 

Quote

It's not surprising - but it is a bit shocking to see in this form: Kaby Lake truly does offer zero to the consumer in terms of clock for clock performance. (In fact, a couple of the results show it slower than the Skylake, but these are within the margin of error.)

 

Edited July 12, 2017 by tom_w141

[Enderman]

6 minutes ago, tom_w141 said:

lmao you are still going.

 

You realise by margin of error we mean that 1-2fps variance is likely between tests. e.g. if I bench a 6700k 3 times I'm going to get 3 slightly differing results. I'm not going to get 150 3 times, I might get 147, 150 and 152. A couple frames either side is acceptable error.

 

Something you might wanna learn: https://www.pcper.com/reviews/Processors/Intel-Core-i7-7700K-Review-Kaby-Lake-and-14nm/Clock-Clock-Kaby-Lake-Skylake-Broad

 

Seriously let it go you're making yourself look like an idiot. Quit while you are behind.

So you also this this is margin of error?

And this?

 

Ok sure, whatever you want to believe dude...........

"6700k and 7700k must have the same IPC, that's why these benchmarks show a difference"

 

[tom_w141]

31 minutes ago, Enderman said:

So you also this this is margin of error?

And this?

 

Ok sure, whatever you want to believe dude...........

"6700k and 7700k must have the same IPC, that's why these benchmarks show a difference"

 

1 reputable source obviously wasn't enough: https://arstechnica.com/gadgets/2017/01/intel-core-i7-7700k-kaby-lake-review/

 

 

 

 

EDIT:

 

@Enderman Just in case you scrape some other unreferenced graph from the corners of the Internet i'll pre-emptively hit you with most of the major reputable sites (I haven't got all day afterall)

 

http://www.anandtech.com/show/10959/intel-launches-7th-generation-kaby-lake-i7-7700k-i5-7600k-i3-7350k/8

http://www.tomshardware.co.uk/intel-kaby-lake-core-i7-7700k-i7-7700-i5-7600k-i5-7600,review-33752-5.html

http://www.pcworld.com/article/3152701/computers/official-intel-7th-gen-kaby-lake-review-one-big-change-makes-up-for-smaller-ones.html?page=3

https://www.eteknix.com/more-i7-7700k-kaby-lake-benchmarks-show-no-ipc-gains/

 

Also wiki for funsies:

 

https://en.wikipedia.org/wiki/Kaby_Lake 

[Drak3]

1 hour ago, ivan134 said:

Why are they testing at 1080p?

Create a bottleneck on the CPU, to test which is better.

[the MBK]

2 hours ago, Valentyn said:

Were you here for the Ryzen launch reviews? Went mental.

Also in many of the tests the 6800X was doing really badly on 1% minimums, even when overclocked compared to Stock 7700K. That really affects gaming performance and consistency of experience.

Good god, there is no 6800X. It does not exist. There is only a 6800k and a 7800X. I read through this whole discussion and this is the only thing that really annoyed me here, despite there being one guy who always claims Skylake and Kabylake IPC are different, which is already quite infuriating.

Well... I should calm down a bit.

[Valentyn]

1 minute ago, the MBK said:

Good god, there is no 6800X. It does not exist. There is only a 6800k and a 7800X. I read through this whole discussion and this is the only thing that really annoyed me here, despite there being one guy who always claims Skylake and Kabylake IPC are different, which is already quite infuriating.

Well... I should calm down a bit.

Was a typo, as meant for 7800X

 

Although looking at performance they might as well have called it the 6800X, looks worse than 6800K.

[Sauron]

Wow, what a surprise... a 6 core draws more power than a quad core and performs slightly lower in games which don't take advantage of the extra cores. I'm shocked.

 

We should stop using games as a metric for cpus, it's a waste of time. If you want the best gaming performance possible get the highest clock speed intel quad core and you'll be just fine, let people who do other things have a performance boost too if they want it, thank you very much. This is the sort of mentality that has allowed intel to keep selling quad cores at the same price for the last 7 years. "Oh here are more cores!" "hurr durr it's 10% slower in games" "ok then".

 

We should be comparing this with the ryzen 8 core chips which come at the same price. We should compare them in multithreaded workloads. Then we'd have some actually useful numbers.

[WkdPaul]

6 hours ago, TheRandomness said:

Please people, calm down.

Another mod already asked that you guys calm down.

 

 

As a reminder, being aggressive, condescending or passive aggressive is against the Community Standards, same for "calling out" another user;

Quote

• Ensure a friendly atmosphere to our visitors and forum members.
• Encourage the freedom of expression and exchange of information in a mature and responsible manner.
• "Don't be a dick" - Wil Wheaton.
• "Be excellent to each other" - Bill and Ted.
• Remember your audience; both present and future.

 

If those simple instructions can't be followed, we'll happily issue warnings to make sure the message hits home.

 

FYI, we're not facts checkers, reporting someone because they have "the wrong facts" or because they misunderstood something isn't an infraction, being a dick to that person is though ...

[MageTank]

3 hours ago, Enderman said:

So you also this this is margin of error?

And this?

 

Ok sure, whatever you want to believe dude...........

"6700k and 7700k must have the same IPC, that's why these benchmarks show a difference"

 

It's not that these two results are within margin of error. It's that you are providing them as evidence to an IPC boost, without understanding how these tests work. I just showed you that photoworxx's score is heavily dependent on memory bandwidth:

 

33k on my 4.5ghz 6700k. The entire reason I score that high is due to my aggressive memory overclock. Because you are not providing the source for us to see their testing methodology (and hardware specs to confirm memory speeds) it makes it impossible for us to validate your claims. I am not above admitting when I am wrong, but I need all of the variables in order to figure that out.

 

The same can be said of cinebench. You can manipulate it's results with memory speed (to some degree) and depending on how it's configured, speedshift 2.0 can dramatically impact the results. It's not evidence of an IPC boost until we can get a real apples to apples comparison. If you'd like I can run photoworxx again on my 7700k at 4.5ghz, with identical memory clock speeds and show you that it's the same as my 6700k. If that will put this entire argument to bed, I'll gladly do so.

 

 

[VagabondWraith]

Did he overclock the cache to 3Ghz? That really speeds things up. Stock 2.4 is limiting performance. Overclocking the cache is a must much like faster ram on Ryzen.

[Enderman]

10 minutes ago, MageTank said:

It's not that these two results are within margin of error. It's that you are providing them as evidence to an IPC boost, without understanding how these tests work. I just showed you that photoworxx's score is heavily dependent on memory bandwidth:

 

33k on my 4.5ghz 6700k. The entire reason I score that high is due to my aggressive memory overclock. Because you are not providing the source for us to see their testing methodology (and hardware specs to confirm memory speeds) it makes it impossible for us to validate your claims. I am not above admitting when I am wrong, but I need all of the variables in order to figure that out.

 

The same can be said of cinebench. You can manipulate it's results with memory speed (to some degree) and depending on how it's configured, speedshift 2.0 can dramatically impact the results. It's not evidence of an IPC boost until we can get a real apples to apples comparison. If you'd like I can run photoworxx again on my 7700k at 4.5ghz, with identical memory clock speeds and show you that it's the same as my 6700k. If that will put this entire argument to bed, I'll gladly do so.

 

 

Reverse image search is not hard, or just google "7700k vs 6700k same clock speed benchmarks".

The image URL is the review site it is from too.

 

If you want to run some tests with same memory clock then sure, go ahead, I'll take a look.

But test something that stresses the CPU to 100%, not like games, because otherwise the difference is very little, "within margin of error" as you claim

 

[ApolloFury]

10 hours ago, Enderman said:

It was a pretty significant improvement when talking about CPUs.

 

Past 8:12 of that video they compare all CPUs at the same clock speed of 4.5GHz. Kabylake and Skylake are identical clock for clock, Kabylake may change state slightly quicker than Skylake thus the moment they started the benchmarks the 7700K will have a brief FPS 1-2FPS advantage for the first couple of seconds, but after that they are producing the same FPS.

[sazrocks]

9 minutes ago, Enderman said:

Reverse image search is not hard, or just google "7700k vs 6700k same clock speed benchmarks".

The image URL is the review site it is from too.

 

If you want to run some tests with same memory clock then sure, go ahead, I'll take a look.

But test something that stresses the CPU to 100%, not like games, because otherwise the difference is very little, "within margin of error" as you claim

 

Here is the conclusion of your article.

Quote

There is no actual performance increase on the IPC and aside from a 200MHz base clock bump and 300MHz boost clock bump, there are absolutely no gains to be had.

https://play3r.net/reviews/cpus/intel-core-i7-7700k-kaby-lake-cpu-review-i7-7700k-vs-i7-6700k-z270/

 

There also seems to be something funky about their testing methodology, as I don't see how two systems using identical RAM could possibly get a difference in RAM latencies unless there is a significant difference in the IMCs in KL vs SL, which I am unaware of existing.

[ApolloFury]

10 hours ago, Enderman said:

2 fps huh?

It's almost like you have a 6700k and are trying to convince yourself that the 7700k is not better...

You only just proved that the 6700K is better since it has a higher average of 127.1FPS over Kabylake's 124.1. It only just happens at that exact moment that the Kabylake was producing 138.0FPS at that split second but we all know that if we benchmark the same scene again the numbers will change but the averages will be similar. Has it ever crossed your mind that doing the same benchmarks over and over again will never product the exact same result? It is certainly possible that the 6700K and 7700K current FPS (122, 138) and average FPS (127, 124) could completely change when we rerun the scene.

[MageTank]

30 minutes ago, Enderman said:

Reverse image search is not hard, or just google "7700k vs 6700k same clock speed benchmarks".

The image URL is the review site it is from too.

 

If you want to run some tests with same memory clock then sure, go ahead, I'll take a look.

But test something that stresses the CPU to 100%, not like games, because otherwise the difference is very little, "within margin of error" as you claim

 

Photoworxx doesn't stress CPU's to 100%, not even close. It doesn't even care about clock speeds. It simply cares about thread count and memory bandwidth. 

 

If you want 100% stress, then Linpack is the way to go. It will also tell us the raw flops that the CPU can achieve (and since it's AVX, it's also hot and tied to memory bandwidth again). I can run my CPU's at a static 4.5ghz, and use a more achievable 3200 C14 XMP profile (or less, if you'd rather I run a speed more people can actively test and dispute).

 

18 minutes ago, sazrocks said:

Here is the conclusion of your article.

https://play3r.net/reviews/cpus/intel-core-i7-7700k-kaby-lake-cpu-review-i7-7700k-vs-i7-6700k-z270/

 

There also seems to be something funky about their testing methodology, as I don't see how two systems using identical RAM could possibly get a difference in RAM latencies unless there is a significant difference in the IMCs in KL vs SL, which I am unaware of existing.

There is a little known secret to achieving better latency scores, something even I kept under my sleeve, telling only @Lays and @done12many2, but i'll share it here to further dispute the results. If you have a higher CPU cache clock (uncore), it will reduce your memory latency. Assuming he kept the 6700k's cache at stock (only raising the CPU multiplier), the 7700k has a 300mhz cache clock advantage (or 100mhz, if you suffer that -200 cache offset bug). I can say that my 7700k's IMC is worlds stronger than my 6700k's, so that could also be the case here. My 7700k was able to push my ram to 3866 C15 (or 3733 C14) while my 6700k absolutely could not push beyond 3600 C14. I've even booted 4000 C16 on this 7700k, on a board that isn't exactly the best for memory overclocking. 

 

Either way, go test your cache speed vs latency in Aida64, and you will see how dramatic of a difference it actually makes. 

[MageTank]

To further illustrate my point on cache speed, and it's impact on the IMC (and latency): 

 

Stock 3200 C14 XMP, 4.8ghz cache (north bridge clock):

 

Stock 3200 C14 XMP, 4.0ghz cache (north bridge clock):

Now, I am not even going to sit here and pretend that I know why it's that dramatic a difference in latency. Someone more educated on caches relationship to the IMC will have to clarify that point, but I know that I, along with some of my enthusiast memory overclocking friends, have kept this a secret for quite some time now. That 800mhz difference was a 12ns difference in latency for some reason. Now, I will also say this: the higher you push your memory overclocks, the less dramatic an impact cache seems to have on latency. So diminishing returns do exist. For my 3600 C14 profile, I go from 39ns at 4.2ghz cache, down to 36ns at 4.8ghz cache. Basically 1ns per 200mhz cache. For this slower XMP speed, the difference seems to be far more dramatic.

 

If it helps, I can redo the tests with my Aida64 SPD page opened in the background, along with ASRock timing configurator open so that you can see I did not adjust tRFC, tREFI, or any tertiary timings to deflate this result. I take memory overclocking very seriously, and I do not benefit from deceiving with false information. If it helps, @done12many2can chime in with his results after tinkering with cache speeds. He actually spent more time testing cache than I did, lol. 

[Sauron]

5 hours ago, Enderman said:

So you also this this is margin of error?

And this?

 

Ok sure, whatever you want to believe dude...........

"6700k and 7700k must have the same IPC, that's why these benchmarks show a difference"

 

If you actually check the numbers the difference is much lower than that graph would suggest visually - the difference in cinebench is about 6%, in the other one it's like 0.3%.....

[Sauron]

11 minutes ago, MageTank said:

Now, I am not even going to sit here and pretend that I know why it's that dramatic a difference in latency. Someone more educated on caches relationship to the IMC will have to clarify that point, but I know that I, along with some of my enthusiast memory overclocking friends, have kept this a secret for quite some time now. That 800mhz difference was a 12ns difference in latency for some reason. Now, I will also say this: the higher you push your memory overclocks, the less dramatic an impact cache seems to have on latency. So diminishing returns do exist. For my 3600 C14 profile, I go from 39ns at 4.2ghz cache, down to 36ns at 4.8ghz cache. Basically 1ns per 200mhz cache. For this slower XMP speed, the difference seems to be far more dramatic.

I could be completely wrong but might that not have something to do with the fact that faster cache finishes looking for data faster and allows the cpu to move on to ram sooner?

[ravenshrike]

5 hours ago, Enderman said:

"it's margin of error!!!!11!"

"margin of error this! margin of error that!"

"2% must be margin or error! I can't tell a difference!"

 

Something you might wanna learn: http://www.owlnet.rice.edu/~labgroup/pdf/Error_analysis.pdf

Maybe then you'll understand why two decimal places are included.

It's within the margin of error in two respects. The first being within the test itself. Unless they're running their series of benches at least 10 times on both processors while completely shutting down and restarting between series and averaging that 1-2% is well within the margin. There's also the fact that between the same CPUs you can easily get up to a 1% difference in IPC between golden chips and crap chips. As such even if that particular 7700k is consistently 1-2% faster than that particular 6700k that does not mean the series of chips have a different IPC.

[MageTank]

6 minutes ago, Sauron said:

I could be completely wrong but might that not have something to do with the fact that faster cache finishes looking for data faster and allows the cpu to move on to ram sooner?

I have no idea. I just know that uncore is tied to your memory controller,QPI controllers (L3, IMC, etc).

 

https://en.wikipedia.org/wiki/Uncore

Quote

The uncore or system agent is a term used by Intel to describe the functions of a microprocessor that are not in the core, but which must be closely connected to the core to achieve high performance.[1] It has been called "system agent" since the release of Sandy Bridge Intel microarchitecture.[2] The core contains the components of the processor involved in executing instructions, including the ALU, FPU, L1 and L2 cache. Uncore functions include QPI controllers, L3 cache, snoop agent pipeline, on-die memory controller, and Thunderbolt controller.[3] Other bus controllers such as SPI and LPC are part of the chipset.

So I am assuming, if your memory controller is bottlenecked by your uncore (cache) speed in any way, that increasing it will remove said bottleneck. I don't really understand the science behind it, I am just a simple overclocker, but someone might be able to clarify for us if they understand how it works. Either way, that's how a lot of the higher end memory overclockers set their records. Using tricks most people deemed "worthless". 

[done12many2]

21 minutes ago, MageTank said:

Now, I am not even going to sit here and pretend that I know why it's that dramatic a difference in latency. Someone more educated on caches relationship to the IMC will have to clarify that point, but I know that I, along with some of my enthusiast memory overclocking friends, have kept this a secret for quite some time now. That 800mhz difference was a 12ns difference in cache for some reason. Now, I will also say this: the higher you push your memory overclocks, the less dramatic an impact cache seems to have on latency. So diminishing returns do exist. For my 3600 C14 profile, I go from 39ns at 4.2ghz cache, down to 36ns at 4.8ghz cache. Basically 1ns per 200mhz cache. For this slower XMP speed, the difference seems to be far more dramatic.

 

Just as it is with core clocks, the goal in cache overclocking should be to minimize wasted cycles.  Increasing its speed increases the opportunity for faster cycles from time of requests to time received (lower latency).  

 

I think the fact that higher memory overclocks net less return from cache overclocking can be explained in the fact that the goal of minimizing wasted cycles is a product of both ends.  If it's a miss, the request is then sent to the next in line.  If the next batter up has his game face on, he's helping to keep the overall goal of minimized wasted cycles to a minimum.  In the end, the faster memory is already doing a great job of sending stuff back so the cache doesn't benefit as much from running faster.  That's not to say that it doesn't benefit at all because the faster cache always has better opportunity.  

 

Just a hunch.  

 

Quote

If it helps, I can redo the tests with my Aida64 SPD page opened in the background, along with ASRock timing configurator open o that you can see I did not adjust tRFC, tREFI, or any tertiary timings to deflate this result. I take memory overclocking very seriously, and I do not benefit from deceiving with false information. If it helps, @done12many2can chime in with his results after tinkering with cache speeds. He actually spent more time testing cache than I did, lol. 

 

Cache (mesh) speed on Skylake-X is pretty dramatic.  Increasing cache from stock 2400 MHz to just 3000 MHz can bring about a 10ns + drop in request to received latency, which is what AIDA64 measures.