Who's buying a 6 core Coffee Lake CPU? (Poll Inside)

[VirtualLarry]

I think that the real "pedestrian" battle between Intel and AMD is going to be around the Ryzen 3 1200/1300X level, and the Intel Core i3-8100 (is there a K model SKU too?) 4C/4T CPU as well. Intel will have the clockspeed advantage (for a K model SKU, if they have one in 4C/4T for CFL-S), a slight IPC advantage, and AMD will, in all likelyhood, have a price / value advantage. They could probably drop the price of the Ryzen 3 1200 to $90 or (gasp) possibly even $85 and still make money on it.

But, one other advantage, that Intel would have in such a battle, is an iGPU, for business systems. AMD will have to wait for Raven Ridge for an answer to that, I guess.

I look forward to building some CFL-S i3 4C/4T rigs for people. Not liking the fact that only Z370 chipset boards will be available first, waiting for H310, or B350i.

It's a shame that Intel had to be forced into providing value by AMD, but that's what competition does!

 

[maddie]

I think that the real "pedestrian" battle between Intel and AMD is going to be around the Ryzen 3 1200/1300X level, and the Intel Core i3-8100 (is there a K model SKU too?) 4C/4T CPU as well. Intel will have the clockspeed advantage (for a K model SKU, if they have one in 4C/4T for CFL-S), a slight IPC advantage, and AMD will, in all likelyhood, have a price / value advantage. They could probably drop the price of the Ryzen 3 1200 to $90 or (gasp) possibly even $85 and still make money on it.

But, one other advantage, that Intel would have in such a battle, is an iGPU, for business systems. AMD will have to wait for Raven Ridge for an answer to that, I guess.

I look forward to building some CFL-S i3 4C/4T rigs for people. Not liking the fact that only Z370 chipset boards will be available first, waiting for H310, or B350i.

It's a shame that Intel had to be forced into providing value by AMD, but that's what competition does!

My prediction from months ago is that in the near future we are going to see what a CPU really can drop to, pricewise. So many are brainwashed into thinking that recent pricing history is due to some divine commandment & is sacrosanct.

This CL appears to be strong CPU line and should get a response from the competition.

 

[moonbogg]

I voted no, but if it turns out to be amazing, then who knows what I'll do. I certainly have no idea. Darn it I voted NO though.

 

[whm1974]

I voted no, but if it turns out to be amazing, then who knows what I'll do. I certainly have no idea. Darn it I voted NO though.

Same here. However even if it turns out to be amazing, I can't really justify building a new rig until maybe two or three years from now. Maybe even four.

 

[Excessi0n]

My criteria for my next upgrade are that it must have a better architecture, better clockspeeds, and MOAR CORES. I have a high-clocking 6700K, so it shouldn't be difficult waiting for IceLake or Zen2.

 

[The Stilt]

If CFL can easily (high-end conventional cooling) hit >= 5GHz and provides even a slight overall IPC improvement over SLK/KBL, then I'm sold.

 

[PeterScott]

True, certain workloads (and likely many games, at least current gen ones) will see little benefit from HT, you could put the ~$100 savings towards a better GPU, for example. Well that was my rational when I built my i5 2500K back in the day, but to be perfectly honest this time I'm not sure as I don't tend to upgrade my CPU anywhere near as often as the GPU (my 2500K has gone through 3 GPU upgrades already) and the extra 'future proofing' ability of 6C/12T vs 6C/6T could be worth it in the long run if you plan on keeping the CPU for a few years.

It doesn't really seem like it is will really be any more future proof. It's not like it is the difference between 6 and 12 real cores. Even in best case you might be looking at 20% boost from HT, so the equivalent of about 7 real cores in the case of a few specific loads.

Unless someone has seen something where hyberthreading has more impact than that.

 

[Dufus]

This is not an IPC test, no wonder.

Many do not seem to understand the difference between IPC (Instructions Per Cycle) and Instructions Per Second. Or that it varies with operands for that matter.

 

[frozentundra123456]

It doesn't really seem like it is will really be any more future proof. It's not like it is the difference between 6 and 12 real cores. Even in best case you might be looking at 20% boost from HT, so the equivalent of about 7 real cores in the case of a few specific loads.

Unless someone has seen something where hyberthreading has more impact than that.

Cant link the source, but I know in a few games hyperthreading gives more boost than that. I think in crysis 3 some tests show 35 to 40 percent. Now that is from a 4 core cpu. How much boost it will give to six cores remains to be seen. As I said earlier in the thread though, my choice would most likely be the i5.

 

[VirtualLarry]

Even in best case you might be looking at 20% boost from HT, so the equivalent of about 7 real cores in the case of a few specific loads.

CPU-Z's benchmark includes a "Core Scaling" number. (Edit: My bad, it calls it "Multi Thread Ratio". A Ryzen 5 1600 at stock, hits 8.06. So, 6 cores + SMT == 8.06 cores, for benchmarking purposes?)

 

[Dufus]

So does that mean it equals 6.0 if HTT is disabled? Try it

 

[CHADBOGA]

You say IPC then you apply that logic to games, then backtrack & say that test's not IPC & now again 20% more in games?

You do realize that it's you shifting the goalposts right, how about you add a caveat to every statement of yours so that you don't have to resort to Trump like tactics? As for GPU bottleneck, where did you see an 8k bench in there or do you think 1080p is irrelevant to your claim?

One does not use GPU limited results to determine CPU capability, unless one is profoundly ignorant and/or has a clear agenda.

 

[R0H1T]

One does not use GPU limited results to determine CPU capability, unless one is profoundly ignorant and/or has a clear agenda.

Right so why would one apply the logic of IPC (20% above Zen according to him) to games, since you're clearly not ignorant &/or agenda driven you would know that CPU is generally a secondary concern in games. Also I asked him how would you determine gaming IPC (by forcing games to run on single core?), maybe you can answer that on his behalf.

P.S. I don't pretend to be unbiased, never have, but it's always fun exposing those who like to wear that cloak.

 

[R0H1T]

http://www.legitreviews.com/cpu-bot...ed-on-amd-ryzen-versus-intel-kaby-lake_192585

Here you go, please look at the 1080P results for thr 1080Ti.

Keep in mind that the 1800X has twice the amount of cores/threads compared to the 7700K and still is easily 20%+ slower in the CPU limited 1080P scores. Coffee Lake with its slightly higher IPC and extra 50% core/thread count will only extend this lead further.

I'm not sure why are you supporting his flawed argument that since CFL might have an advantage of 20% in some applications, wrt Zen, that games will also be 20% faster?

You know very well, like the other poster, that gaming results largely depend on GPU, resolution, AA, AF et al. The CPU, same class, from both sides will be inside 5~15% of each others performance barring outliers, like heavy OC or MP option. Just to give you an idea, we have same TDP in the first case & same clock speeds for the latter one ~

http://www.anandtech.com/bench/product/1851?vs=1832
http://www.anandtech.com/bench/product/1950?vs=1832

The IPC debate is all fine till you restrict it to certain applications, (over)extending it to games however is highly disingenuous, that's what I'm against.

 

[R0H1T]

Did you even look at the link I posted?

In terms of running game code, in a CPU limited scenario Kaby Lake (and by extension Coffee Lake) is easily 20% faster clock for clock, if not more. Yes I'm well aware that you can turn all settings to ultra and max out the AA etc and put more burden on the GPU side but that is not the point of this discussion.

There is actually a practical benefit for a faster CPU in gaming, and that is to maintain 144fps (or higher) for Freesync/Gsync gaming. Gaming isn't just about maxing out settings and being happy with 60fps. Some people may be happy with a 60fps average with maxed settings, but I'm sure competitive FPS players for example would be more than happy to lower settings in order to maintain 144fps.

Did you, the 7700k is not running at fixed clocks? How can you claim this, when you aren't even paying attention to your own statements?

We also gave the AMD Ryzen 7 1700 a head start by overclocking it from 3.0GHz (3.7GHz Turbo) stock clock speeds all the way up to 4GHz on all 8-cores. The Intel Core i7-7700K 'Kaby Lake' processor was left at stock settings, which are 4.2GHz base and 4.5GHz boost

Seriously if any of you are lecturing me on IPC, or clock for clock, then go read a book or find reviews that're actually doing tests at fixed clock, not stock vs stock or whatever!

 

[Dufus]

@VirtualLarry Here you go, example of CPUZ with HSW 14 core. Sorry I don't have CFL.

Running a flat 30x multi across all cores with HTT disabled.

CBR gets close to the expected 14.0, some background tasks running maybe, but CPUZ quite a bit off. Seems reminiscent of it's buggy beta days in that if 12 threads or less are selected then the result is very close to the number of cores but if we try all cores then there's a problem. Perhaps the thread creator is hogging the resources of one core.

Now with HTT enabled

Personally I'd have more faith in the CBR ratio and for HTT gain it may be simpler to take the ratio of CPU scores.


Of course it depends on the opcodes used as well.

No HTT, 14 cores and 14 threads

570GFLOPS, could also be had by enabling HTT but setting only one thread per physical core.

HTT, 14 cores and 28 threads

456GFLOPS. That's a 20% loss from using HTT.

So a little care needs to be taken when quoting performance and of course people need to quote the correct reference. For instance approximating IPC from the data in this link http://www.cpu-monkey.com/en/compare_cpu-amd_ryzen_5_1600x-708-vs-intel_core_i7_8700k-763 shows one result at 12% and all others at under 10%. IPS however is quite a bit higher.

 

[Dribble]

i2500K owner, probably will get an i8700K - this is the first chip that feels like a worthwhile upgrade giving me both better single core performance and more cores. As I held onto the i2500K for ever will probably spend the extra on the 12 thread version this time.

 

[Excessi0n]

It doesn't really seem like it is will really be any more future proof. It's not like it is the difference between 6 and 12 real cores. Even in best case you might be looking at 20% boost from HT, so the equivalent of about 7 real cores in the case of a few specific loads.

Unless someone has seen something where hyberthreading has more impact than that.

Multitasking? Review sites don't do benchmarks with a bunch of stuff running in the background. They run a single thing at a time on clean installs of Windows.

 

[Tup3x]

http://www.guru3d.com/news-story/intel-core-i7-8700k-benchmarks.html

Cinebench 15 scores. 1230 doesn't seem quite right in multithreaded test.

 

[LTC8K6]

http://www.guru3d.com/news-story/intel-core-i7-8700k-benchmarks.html

Cinebench 15 scores. 1230 doesn't seem quite right in multithreaded test.

They reference Turbo 3.0, which the 8700K does not have.

 

[Dave2150]

Nope. Had my 6700k for two years, would feel rather moronic to upgrade to an identical architecture with 2 more cores that I won't utilize in gaming.

I'll wait for the 9700k, which I presume will be Icelake, the first new architecture since Skylake.

 

[Insert_Nickname]

A certain amount of speculation regarding CFL's expected performance vs its competition is inevitable, and not entirely off-topic.

Expected performance? I don't think we need much speculation. This is basically a 7700K with 2 additional cores, and 4MB extra L3 cache. I don't expect anything more or less.

CPU performance will likely land @ stock 7700K-level, but a fair bit faster in MT. IGP is more-or-less identical to Kaby Lake, not much surprise there either.

 

[crashtech]

Expected performance? I don't think we need much speculation. This is basically a 7700K with 2 additional cores, and 4MB extra L3 cache. I don't expect anything more or less.

CPU performance will likely land @ stock 7700K-level, but a fair bit faster in MT. IGP is more-or-less identical to Kaby Lake, not much surprise there either.

The unknown is mainly sustained max clocks with those 2 extra cores. A 5GHz hexacore sounds pretty nice, this level of performance may actually put it ahead of an 1800X in MT even with two fewer cores.

 

[slashy16]

I will definitely be purchasing a 8700k if it offers better single thread performance than the 7700k. I wish to replace my work computer I7 990x@4.4 with something better and unfortunately, ryzen isn't up to the task.

 

[Insert_Nickname]

The unknown is mainly sustained max clocks with those 2 extra cores. A 5GHz hexacore sounds pretty nice, this level of performance may actually put it ahead of an 1800X in MT even with two fewer cores.

Yes, that is the question. It is pretty binary actually*, if it has similar frequency to the 7700K, it will use more power. Or if it uses the same power, it will have lower frequency.

I'm actually more interested in how the non-HT i5s fare power-wise. Especially vs. 4 core i7s.

*pun was intended...