8700k/8600k Simulated Results - How close to actual??

TahoeDust

Senior member
Nov 29, 2011
557
404
136
I used my 7820x to mimic what looks to be the type of performance we will see out of the 8700k and 8600k when overclocked. From what I have seen 5.0GHz should be pretty easily obtained...even by using auto overclock on some motherboards. To simulate the 8700k I disabled two cores, left on hyperthreading, and overclocked to 5.0GHz effectively giving me 6c/12t. To simulate the 8600k I disabled two cores, disabled hyperthreading, and overclocked to 5.0GHz effectively giving me 6c/6t.

Supposedly, Coffee Lake has some IPC increase so theoretically the real 8700k and 8600k numbers should be even better.

How close do you think these results will be to actual? Better or worse and by how much?

Here were my results...

8700k - 7820x 6c/12t 5.0GHz

wsaOfgR.jpg




8600k - 7820x 6c/6t 5.0GHz

wX2Gz7T.jpg
 
Last edited:

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
Is the ROG CPU-Z reading those volts correctly? Impressive chip, nonetheless. Those numbers are about what I expect. The 8600K without HT is quite a strong performance compared to 4c/8t. I wonder what the 7700k scores at 5GHz with HT enabled?
 

TahoeDust

Senior member
Nov 29, 2011
557
404
136
Is the ROG CPU-Z reading those volts correctly? Impressive chip, nonetheless. Those numbers are about what I expect. The 8600K without HT is quite a strong performance compared to 4c/8t. I wonder what the 7700k scores at 5GHz with HT enabled?
No. CPU-Z is displaying VRIN as VCore. It is a bug in the motherboard bios. I have not gotten around to updating the bios...too lazy to want to re-enter my OC profiles.
 

Zucker2k

Golden Member
Feb 15, 2006
1,810
1,159
136
No. CPU-Z is displaying VRIN as VCore. It is a bug in the motherboard bios. I have not gotten around to updating the bios...too lazy to want to re-enter my OC profiles.
Woah! All these while? You're probably missing a few performance tweaks. Doesn't your board allow saving your oc profiles?
 

TahoeDust

Senior member
Nov 29, 2011
557
404
136
Woah! All these while? You're probably missing a few performance tweaks. Doesn't your board allow saving your oc profiles?
I can save the profiles, but they are not accepted if saved in a previous bios version. I updated it once about a month ago and it did not fix the CPU-z reporting. Maybe I will try the newest later.
 

beginner99

Diamond Member
Jun 2, 2009
5,320
1,768
136
Since Skylake-X has completely different cache architecture I call this comparison problematic. Yes it will be in the same ballpark but in some benches the increased L2 will help and in some the lack of L3 will be an issue. Also uncore speed of 7820x will hence influence the results as well.

It will behave like a 7700k with more cores and preliminary leaked benches confirm this.
 

TahoeDust

Senior member
Nov 29, 2011
557
404
136
Since Skylake-X has completely different cache architecture I call this comparison problematic. Yes it will be in the same ballpark but in some benches the increased L2 will help and in some the lack of L3 will be an issue. Also uncore speed of 7820x will hence influence the results as well.

It will behave like a 7700k with more cores and preliminary leaked benches confirm this.
That was kind of the point of the thread. I was just curious if others thought the actual performance would be higher or lower, and by how much.
 

Pookums

Member
Mar 6, 2017
32
13
36

That is Clearly cherry picking. Its taking the best coffeelake scores and the shittiest 7700k scores. most 7700k with decent ram at 5ghz reach 1100. The highest I've ever seen is around 5.2ghz, with OC'ed uncore and samsung b die was around 1200. My 7700k at 4.5 enchanced all core and 3000mhz ram scores just a few points under THAT shown 7700k above. My ram at 2400cl15 before xmp at 4.5 is around 995. That is problematic. 930-950 multi for 7700k is what occurs when no turbo function is working running 4.2 all core.

Its possible the above picture is 3333mhz ram, but might also be the worst latency times imaginable for that speed(for example 16-17), with crap cooling and crap board to cause some minor throttling. Thats the only scenario possible.

Some of the early Coffee leaks when averaged (done to reduce cherry picking) suggested that coffeelake should be possibly 47% faster. Therefore if 7700k is usually closer to 1100 at 5ghz(which can be seen around the web), Then coffeelake should score maybe 1620 (though core/uncore might have a few more diminishing returns than usual at such a high clock)...and so 1600 seems a fair guess for likely 5ghz paired with fast ram for coffeelake. That 1663 with 219 seems like its probably the best respective run for coffeelake on cinebench. This also suggests that coffeelake might have SLIGHTLY worse ST per clock than kabylake, as the highest 5 ghz 7700k I have seen is 224. So its possible it loses 1-2% in order to vastly improvement MT efficiency through cores.

However, I also have my doubts that 8700k will be able to turbo above 4.7 or 4.8 all core average as a result of heat. Golden chips might occur, but on average it should turbo less than a 7700k, as even though its more effecient per core, its still 50% more cores and thus heat. Early benchmarks have also shown 8700k is hotter than 7700k. While kabylake can reach 5ghz, it usually requires strong cooling or a delid to accomplish if you want to pair it with good ram. So 1550 maximum would probably be a good estimate for a multithreaded non delidded 8700k.
 

coercitiv

Diamond Member
Jan 24, 2014
7,448
17,753
136
That is Clearly cherry picking. Its taking the best coffeelake scores and the shittiest 7700k scores.
They are using the same 3333Mhz memory for both systems, how in the world are they picking the best CFL results and the worst KBL ones?! Sure, the best KFL setups will reach higher scores, but so will the best CFL systems.
 

coercitiv

Diamond Member
Jan 24, 2014
7,448
17,753
136
Damn. Pretty freaking close.
Yup, although it may be close due to different factors canceling each other out (considering the differences in cache, memory channels and memory speeds - CB does not scale much with RAM, but it does scale nevertheless)
 

Pookums

Member
Mar 6, 2017
32
13
36
They are using the same 3333Mhz memory for both systems, how in the world are they picking the best CFL results and the worst KBL ones?! Sure, the best KFL setups will reach higher scores, but so will the best CFL systems.

CFL does have larger cache, so may be less effected by poor timings. Regardless, Those aren't even AVERAGE 7700k scores. That is one of the worst 5 ghz 7700k scores I have ever seen. I explained above, and even mentioned my own scores. I'm reaching that score at 4.5ghz and 3000mhz ram. 1100 CB score Is an average 5ghz. That is also the highest non-LN or active cooled bench produced for coffee lake presented of all the leaks out in the wild.

Its cherry picked. Coffee lake may very well end up as a great product. However, the amount of BS I've seen posted about the 8700k lately is ridiculous. Every time there is a leak, and it compares to another architecture (whether previous intel or amd), it always uses poor benchmarks for the competitors, and stretches specific games or programs where 4 cores starts to falter, and 8 cores is unnecessary in order to trump itself up. Look at the previous releases that said 8700k would be in one instance 18% or 27% faster in ST than 7700k. It's the same architecture with more cores. If anything it will be +- 1 to 3%. The fact that at stock it can gain several fps in low thread programs is from a ST turbo set to 4.7, compared to 7700k 4.5. However, It will likely be SLOWER in Low core averages once OCed than 7700k due to heat management. For OC, 8600k or 8350k are the coffeelake processors with a chance at faster ST.

Assuming coffeelake reaches a higher maximum OC (lets for example say 5.3 or 5.4ghz) to take the ST title, it will require a good chip, lots of power, a delid, and Great cooling in order to achieve. At that point a $350+ 8700k is now a $600+ processor.

I have a 7700k. I understand the heat issues associated with an increase in 50% more cores. The marketing being passed around on coffee lake is getting out of hand. I'm sure everyone would love an amazing new all around CPU for a cheap price, however theres nothing in coffee lake technical specs that suggests coffeelake is going any more of a savior than broadwell, skylake-x, or ryzen. It will do a little better than the previous 3 for an All-around chip architecture.

The Cores to look at (assuming leaked pricing is true) are the 8100, 8350k, and 8400. They should offer performance at a MASSIVE price drop. 8700k will likely be the best all-around chip(good ST and good MT). However, depending on how 8600k and 8700 non-k score, intel will possibly cannibalize one of its own core positions. Which is odd, as its managed to do that twice now, and this time it will possibly occur at the higher price end of its release.
 
  • Like
Reactions: coercitiv

coercitiv

Diamond Member
Jan 24, 2014
7,448
17,753
136
Regardless, Those aren't even AVERAGE 7700k scores. That is one of the worst 5 ghz 7700k scores I have ever seen. I explained above, and even mentioned my own scores. I'm reaching that score at 4.5ghz and 3000mhz ram. 1100 CB score Is an average 5ghz.
You could have saved yourself a lot of time by just posting a score from a reliable source. Credit where it's due though, I was obviously on the wrong path here.

85151.png


Considering the MT 4.8Ghz score with 2400MHz RAM is higher than the 5Ghz score from the other review, I would say it's not a cherry picked result, but rather a throttled overclock. The difference in ST score makes that even more probable.

That having been said I guess we'll have to ditch the 8700K result as well, since the reviewer may have botched the other overclock in the same manner.