Intel's 9th Generation Core Family - Coffee Lake (Refresh)

Intel is on the ropes and AMD has a chance to do some real damage with 7nm Zen 2...I think the 8700K is the better upgrade for most right now...

isnt 9700k at similar price but actually offers a bit more performance? why not just get that instead.
 
I am such a dork. 2x performance of an 8350k running at 4.5Ghz vs an 8700k running at 4.3Ghz (NOT 3.7Ghz as shown)

Prime 1024:
9700?: 120s; correct!
8700k: 118s; 119s

H.264:
9700?: 34s; 36s
8700k: 42s; 42s

H.265:
9700?: 79s; 91s
8700k: 99s; 101s

z-zip c:
9700?: 43000 mips; correct!
8700k: 41000 mips; 41000mips

z-zip d:
9700?: 38000 mips; correct!
8700k: 38000 mips; 39000 mips

Java:
9700?: 567; 470
8700k: 431; 433


MySQL:
9700?: 217k TPS; 160k
8700k: 183k TPS; 157k

Blender:
9700?: 292s; 322s
8700k: 302s; 311s

Cinebench:
9700?: 1528; 1572
8700k: 1438; 1432

Euler 3D
9700?: 37s; 54s
8700k: 56s; 56s

So it looks like the 9700k would edge out on most apps and pull away on some like JAVA. Things to consider are Cache size and latency of the 8-core.

https://w.techpowerup.com/reviews/Intel/Core_i3_8350K/

The results for the 9700k were not nearly as good as these estimates, especially with the 4.7ghz "stock" clocks now.

Clock for clock, I am thinking the 8700k would do better. Actual 9700k numbers shown on right, with updated 8700k numbers. For 720p gaming, the 8700k was about 97% of the 9700k and 98% at 1080p.

Finally there is live streaming performance. We have seen a slow 6/12 destroy a fast 6/6 so it is not a stretch that a fast 6/12 would still be better than a fast 8/8.
The first line of the conclusion is that Intel proves that 8/8 is faster than 6/12. That is definately debatable.
 
The first line of the conclusion is that Intel proves that 8/8 is faster than 6/12. That is definately debatable.

Yeah, it should be, for nearly every workload- and shouldn't be appreciably slower when it's not, but could definitely be faster with eight FPUs versus six on the 8700K in FPU-intensive applications that are well-threaded. That would include some games and content creation workloads I'd think, so the 9700K is perhaps more desirable clock-for-clock for most consumer uses.

Stuff that's heavily multi-threaded that's not FPU-intensive, perhaps compiling (SWAG), might be faster on the 8700K clock-for-clock, but no idea how much.

And generally speaking, the 9700K has a larger die and better solder, so it should be a better out-of-box overclocker too.
 
Yeah, it should be, for nearly every workload- and shouldn't be appreciably slower when it's not, but could definitely be faster with eight FPUs versus six on the 8700K in FPU-intensive applications that are well-threaded. That would include some games and content creation workloads I'd think, so the 9700K is perhaps more desirable clock-for-clock for most consumer uses.

Stuff that's heavily multi-threaded that's not FPU-intensive, perhaps compiling (SWAG), might be faster on the 8700K clock-for-clock, but no idea how much.

And generally speaking, the 9700K has a larger die and better solder, so it should be a better out-of-box overclocker too.

The 9700k does seem to clock better, but the 8700k is very close if you delid. Even better if you have an 8086k.

Again, many of the apps did much worse than expected, especially Java and Euler. It is very hard to see where the 9700k is more desirable.
 
The 9700k does seem to clock better, but the 8700k is very close if you delid. Even better if you have an 8086k.

Again, many of the apps did much worse than expected, especially Java and Euler. It is very hard to see where the 9700k is more desirable.

some other synthetic benchmark shows a bit more difference, like around 15-20%
 
Again, many of the apps did much worse than expected, especially Java and Euler. It is very hard to see where the 9700k is more desirable.
some other synthetic benchmark shows a bit more difference, like around 15-20%

But are we talking about actual consumer loads?

I get that synthetics can be representative of actual workloads, but this isn't always the case.
 
But are we talking about actual consumer loads?

I get that synthetics can be representative of actual workloads, but this isn't always the case.

most consumer load would be within 1% because of clock speed, in that case its IPC theres no improvement. a lot of consumer stuff still only use 1 core btw so this issue always existed, from 6700k to 7700k then 8700k etc
 
most consumer load would be within 1% because of clock speed, in that case its IPC theres no improvement. a lot of consumer stuff still only use 1 core btw so this issue always existed, from 6700k to 7700k then 8700k etc

Gaming is thread-limited; rarely would the extra threads from HT with eight cores make a difference, unlike how it does at four cores and to some extent six. But other stuff like photography and videography, which with gaming is what I mean when speaking to 'consumer' loads, also probably benefit very little from a benchmark perspective.
 
Gaming is thread-limited; rarely would the extra threads from HT with eight cores make a difference, unlike how it does at four cores and to some extent six. But other stuff like photography and videography, which with gaming is what I mean when speaking to 'consumer' loads, also probably benefit very little from a benchmark perspective.

gaming would also be dependent on thread and clock. if a game uses 8 threads. then 8700k would be 4c/8t at most and it'll loes to 8c/8t big time. if game design to use full cpu resources then 9700k wouldnt be that much better than 8700k. then if a game uses 12 threads, we got full 8700k being used vs 6/8 cores of 9700k.

but quite honestly i dont care much about games, like 3-5 fps out of 100 is nothing really. but if i were to do say 7zip, winrar, video encoding, browser loading, vmware and i need those to be noticeably faster, which seems 9700k does come out ontop vs 8700k by about 10-15%, thats good enough.
 
if a game uses 8 threads. then 8700k would be 4c/8t at most and it'll loes to 8c/8t big time.

This really depends on the load split between ALU usage, which is easy to farm off to virtual cores provided by hyperthreading, and FPU usage, of which is not so easy. Assuming good thread usage, FPU heavy stuff is likely to favor the 9700K, while ALU heavy stuff is likely to favor the 8700K. Further, the difference can be not just application dependent, but workload dependent, and that's where you have to be really careful about the synthetics and line up the specific use cases.

but quite honestly i dont care much about games, like 3-5 fps out of 100 is nothing really. but if i were to do say 7zip, winrar, video encoding, browser loading, vmware and i need those to be noticeably faster, which seems 9700k does come out ontop vs 8700k by about 10-15%, thats good enough.

And this is why I ask about 'consumer workloads'- you have to do a lot of zipping/unzipping for compression performance to make a difference; same for browser loading, which a dual-core CPU is already fast at. Do you reload fifty tabs every few minutes?

Video encoding is one that I get in addition to gaming, but that also has to be a primary purpose of the system or being done for profit to really be a differentiator one way or the other. VMs I get too- but is a consumer-socketed system being used for dev work or for server work? My homelab server has a 7600k in it and it runs VMs, but the load is still negligible. At what point is two extra FPUs vs. four extra virtual HT threads going to be a decision point there?

This is why I err toward gaming and video/photo editing: this stuff is more likely to use the extra FPUs on the 9700K, which is what I'd recommend in nearly all consumer cases.
 
gaming would also be dependent on thread and clock. if a game uses 8 threads. then 8700k would be 4c/8t at most and it'll loes to 8c/8t big time. if game design to use full cpu resources then 9700k wouldnt be that much better than 8700k. then if a game uses 12 threads, we got full 8700k being used vs 6/8 cores of 9700k.

but quite honestly i dont care much about games, like 3-5 fps out of 100 is nothing really. but if i were to do say 7zip, winrar, video encoding, browser loading, vmware and i need those to be noticeably faster, which seems 9700k does come out ontop vs 8700k by about 10-15%, thats good enough.

7zip and rendering were similar despite the 8700k clock deficiency. Video encoding, VMware and "browser loading" seems pretty proportional to the clock speed advantage.
It is only 10% faster due to clock differences.

I think Intel sold all of the reviewers a bill of goods with the 9700k. By having MCE-like clocks of 4.7 ghz, it looks better than the 8700k.

Priced the same, the 9700k is probably the safer route if you don't delid. 5.0 ghz+ seems like a sure thing with moderate cooling. For the delidders, the 8700k and 8086k seem more attractive.
 
There was alot of critisism for the 8th gen that some used MCE settings. Board makers were being shady for having MCE on by default in some cases and Intel had no control of this.

The time around, MB makers seem to be a bit more sneaky. They are giving turbo boost for an infinite time which negates the 95w tdp:



I guess you could say that this is the new MCE. Do you guys think that this is completely the board makers fault, or does Intel share some blame here? They seem to be encouraging this while still looking good with a lower TDP.
 
If running fully within stock specs is important for you, you almost certainly won't be buying enthusiast aftermarket motherboards.

Intel of course is allowing you to do what you want. Run "safe" or run "fast" - you can do what you want. They publish what they certify is "safe".
 
So do you think the infinite turbo should be considered a motherboard overclock and if not, should TDP still be 95w?
 
So do you think the infinite turbo should be considered a motherboard overclock and if not, should TDP still be 95w?

I don't like the term "overclock", as basically all computational chips have variable/dynamic clocking now. If you're asking if running with infinite turbo exceeds the specced TDP - yes, absolutely. The TDP spec has limits on time spent at boost frequencies. Blowing past that will mean you're using more power in the steady state, and thus have a new TDP.

But I have to admit I don't understand why people are upset about the motherboard makers making generally very safe ways of getting higher performance from high end chips. MCE is incredibly safe, as is infinite turbo with even a moderate cooling system.
 
I don't like the term "overclock", as basically all computational chips have variable/dynamic clocking now. If you're asking if running with infinite turbo exceeds the specced TDP - yes, absolutely. The TDP spec has limits on time spent at boost frequencies. Blowing past that will mean you're using more power in the steady state, and thus have a new TDP.

But I have to admit I don't understand why people are upset about the motherboard makers making generally very safe ways of getting higher performance from high end chips. MCE is incredibly safe, as is infinite turbo with even a moderate cooling system.

It just seems that Intel is getting their cake and eating it too with this scenario ie, high performance while advertising low tdp.

What would you consider moderate cooling and do you know anyone or any reviewer that used moderate cooling during this type of MCE?
 
It just seems that Intel is getting their cake and eating it too with this scenario ie, high performance while advertising low tdp.

What would you consider moderate cooling and do you know anyone or any reviewer that used moderate cooling during this type of MCE?

They are advertising a given TDP for the speeds which correlate with that TDP, along with the information that it is unlocked and you can do what you want - including running at higher speeds / power levels.

MCE will increase the TDP to the ~120W range (reasonably typical). That level of cooling is pretty easy to come by. We're not really shooting up into a specialized bracket at this point, which was what I was getting at.
 
There was alot of critisism for the 8th gen that some used MCE settings. Board makers were being shady for having MCE on by default in some cases and Intel had no control of this.

The time around, MB makers seem to be a bit more sneaky. They are giving turbo boost for an infinite time which negates the 95w tdp:



I guess you could say that this is the new MCE. Do you guys think that this is completely the board makers fault, or does Intel share some blame here? They seem to be encouraging this while still looking good with a lower TDP.


who really cares though the people getting 9900k to run it at stock and stock turbo? honestly i think thats just stupid.

if we getting 9900k it'll obviously all core OC at 4.8-5ghz when we can, for those who want to run it at stock might as well wait for other SKU thats locked no point paying for unlocked chips.

and when we OC, TDP goes out the window, not a problem.
 
who really cares though the people getting 9900k to run it at stock and stock turbo? honestly i think thats just stupid.

if we getting 9900k it'll obviously all core OC at 4.8-5ghz when we can, for those who want to run it at stock might as well wait for other SKU thats locked no point paying for unlocked chips.

and when we OC, TDP goes out the window, not a problem.

Yeah I get that, it just seems that there needs to be some guidance by Intel for the MB manufactures. If Intel advertises TDP at 95w, which is around 4.3 ghz, that should be the stock settings.

Most importantly, not everyone is buying $100+ cooling solution for these. Most likely those using a standard fan rated a TDP will throttle down.
 
Yeah I get that, it just seems that there needs to be some guidance by Intel for the MB manufactures. If Intel advertises TDP at 95w, which is around 4.3 ghz, that should be the stock settings.

Most importantly, not everyone is buying $100+ cooling solution for these. Most likely those using a standard fan rated a TDP will throttle down.

If it is stable, then there isn't an issue on an enthusiast board. Hell, there's really not an issue at all- if the CPU is pushing into unstable territory, it should throttle down. That's good. Those that are picky about actual power usage have those settings available to them in the BIOS.
 
Sweating intensifies...

upload_2018-11-9_11-52-16.png
 
Was looking at my B+H order, and it looks like my 9900K may finally be coming soon. (9900K page says in stock) May want to check it if your interested, but it probably won't last long if so.
 
Looks like they bracketed in the performance pretty nicely!

For the most part, yes. I think I would fork out a little more for the 8700k. It is a munch more versatile chip. The 9600k looks to be a dissapointment for overclocking.

In addition, measuring only fps seems archaic nowadays.
 
As an Amazon Associate, HardForum may earn from qualifying purchases.
Couldn't help myself, so I bit on the 9900K. Was going to go for the 9700K, but I had a $500 gift card that the sale created opportunity for and it now has $0.01 left on it :ROFLMAO:.

in two years from now you will realize the right and wise decision you have made to go with a 8 core 16 thread cpu, enjoy it.
 
I've been very happy with my upgrade to the 9900k. With a Kraken X62 I am fully stable at 5.1 using up to 2 cores and 5G using all cores at 1.3 volts. I have not tried all cores at 5.1G yet. Temps go up to 60 deg C . I had to manually set the voltage as the Asus mobo was giving it above 1.6volts in auto mode.
 
I had to go up to 1.34V to get 5 GHz all core stable on my system. It hit 100C in Prime95 after an hour, but during my gaming tests it didn't go above 70C. I'm going to do some additional tuning to see if I can get the voltage down. Would the memory I'm running cause the CPU to need more juice? All 4 sticks are running well with the XMP at 4000 MT/s.
 
I had to go up to 1.34V to get 5 GHz all core stable on my system. It hit 100C in Prime95 after an hour, but during my gaming tests it didn't go above 70C. I'm going to do some additional tuning to see if I can get the voltage down. Would the memory I'm running cause the CPU to need more juice? All 4 sticks are running well with the XMP at 4000 MT/s.

Why do you need a 6% clock speed boost? Do you really feel anything different?

That's an awful lot of trouble to go through for a chip that is already pre-overclocked for you. Unless you're adding phase change or LN2, then touching the stock clocks is really kind of pointless.
 
Why do you need a 6% clock speed boost? Do you really feel anything different?

That's an awful lot of trouble to go through for a chip that is already pre-overclocked for you. Unless you're adding phase change or LN2, then touching the stock clocks is really kind of pointless.
Why not? I'm not trying to hit 6 GHz or anything. The thing was hitting 1.43V under load with the default settings on this motherboard, anyway.
 
Why not? I'm not trying to hit 6 GHz or anything. The thing was hitting 1.43V under load with the default settings on this motherboard, anyway.

Gotcha on the overzealous factory turbo bin vvoltages:D

When overclocking both CPU and memory controller, you most likely will need more voltage to do it.

The more portions of the chip are suddenly in your critical path, the more current you need to supply it with.

Unfortunately no one has done a memory performance scaling article for 9900k yet, so you're probably best-off with memory as fast as possible.
 
I tried tweaking various things over the weekend but could not get stable on anything less than 1.34V. I got a X72 on the way to replace the X62. Maybe 4100 mm² more surface area on the radiator will make an appreciable difference in temperature. I swear I'll do a custom loop one of these days :cautious:.
 
Some official documents showing CFL-R expansion:
https://www.fanlesstech.com/2019/03/intels-coffee-lake-refresh-lineup.html?m=1

Lots of 'F' models as expected.
The 9350KF sounds fun but it will no doubt be WAY overpriced and not very practical.

Really curious to see what the 2288g brings to the table. It will most likely be clocked lower than the 9900k, but possibly with tighter voltage tolerances meaning the ability to achieve 4.5 ghz+ with 'minimal' cooling such as in a SFF. ECC support is also a plus.
 
Hmm, at true 95w TDP, which they will most likely want the Xeon lineup to run at, it looks as if the 9900k runs closer to 4.0 ghz. Perhaps the Xeon will do a bit better than that.

Curious to see what frequency is needed to keep the 9900T at TDP spec with only 35w. Hopefully it gets a nice boost freq!
 
so what is the consensus on the lack of hyperthreading on the 9700K?...big deal?...little deal?...will Intel continue doing this for their upcoming 10nm Ice Lake series?
 
so what is the consensus on the lack of hyperthreading on the 9700K?...big deal?...little deal?...will Intel continue doing this for their upcoming 10nm Ice Lake series?


If all you do is play games, little deal. If you do a lot of workstation tasks, or you stream on the same machine, then big deal. The latest multiplayer console ports are not using more than six threads (that won't change for another 5 years), and Intel definitely has the advantage there. 8 real cores would be overkill for some time.

Intel is segmenting the market by turning off HT on most processors because they can dominate in games and compete in everything else with their current lineup. If AMD whips out a winner with Zen 2.0, Intel will be forced to compete with whatever pricing AMD offers (HT enabled except at $100).
 
Back
Top