Coffee Lake
#1
Posted 05 October 2017 - 08:44 PM
#2
Posted 05 October 2017 - 10:55 PM
#3
Posted 06 October 2017 - 12:01 AM
#4
Posted 06 October 2017 - 11:48 AM
It looks like without excellent cooling, the 8700K isn't quite as fortunate, but in most tests it either meets or exceeds the Ryzen 8-cores' performance. The clock advantage these chips have over Ryzen make them punch above their weight pretty effectively.
Overall, they really just perform like 6-core Skylake chips. There is a minimal (VERY MINIMAL, 1% AT BEST) IPC improvement. Their real value is the additional 2 cores at similar TDPs, so as games ever so slowly begin to adopt higher core counts, they will finally start to diverge from their Kaby Lake siblings.
The other thing to consider is that it seems like DDR4-3200+ is very easily usable with these chips so any overclocking that you want to do on the bclk/gear ratio should be better (though I'm not sure if that's even required with Coffee Lake).
Unfortunately for me personally, so far it is hard to tell just how much of an improvement an 8700K would be over my 5820K. It's looking like best case scenario, clock for clock it offers about 5% better IPC, and worst-case scenario they're dead even. Since my X99 system has been kind of unstable the entire 3 years I've had it, though, and absolutely will not for any reason touch above 4.3GhZ, the minimum improvement would still be an additional 400MhZ while using something like 40W less.
#5
Posted 06 October 2017 - 11:53 AM
xWiredx, on 06 October 2017 - 11:48 AM, said:
It looks like without excellent cooling, the 8700K isn't quite as fortunate, but in most tests it either meets or exceeds the Ryzen 8-cores' performance. The clock advantage these chips have over Ryzen make them punch above their weight pretty effectively.
Overall, they really just perform like 6-core Skylake chips. There is a minimal (VERY MINIMAL, 1% AT BEST) IPC improvement. Their real value is the additional 2 cores at similar TDPs, so as games ever so slowly begin to adopt higher core counts, they will finally start to diverge from their Kaby Lake siblings.
The other thing to consider is that it seems like DDR4-3200+ is very easily usable with these chips so any overclocking that you want to do on the bclk/gear ratio should be better (though I'm not sure if that's even required with Coffee Lake).
Unfortunately for me personally, so far it is hard to tell just how much of an improvement an 8700K would be over my 5820K. It's looking like best case scenario, clock for clock it offers about 5% better IPC, and worst-case scenario they're dead even. Since my X99 system has been kind of unstable the entire 3 years I've had it, though, and absolutely will not for any reason touch above 4.3GhZ, the minimum improvement would still be an additional 400MhZ while using something like 40W less.
Coffe Lake does have fewer PCI express lanes than X99. I've been reading that 5ghz isn't guaranteed with these new chips either. Still, the 8600K looks pretty great.
#6
Posted 06 October 2017 - 04:59 PM
#8
Posted 06 October 2017 - 05:57 PM
#9
Posted 06 October 2017 - 06:05 PM
part of intels real problem is the lack of long term support for motherboard/cpu's .... its the old obsolete the second you walked out of the shop scenario ....
wouldn't watch that review, youtube wanted me to watch 4 ads in a row .... ahhh no ...
Edited by NARC BAIT, 06 October 2017 - 06:43 PM.
#10
Posted 06 October 2017 - 07:00 PM
LT. HARDCASE, on 06 October 2017 - 05:57 PM, said:
Well, yeah, it's a 7700K with two extra cores and a small clock bump. That should have been obvious from the get-go.
ninjitsu, on 06 October 2017 - 11:53 AM, said:
Coffe Lake does have fewer PCI express lanes than X99. I've been reading that 5ghz isn't guaranteed with these new chips either. Still, the 8600K looks pretty great.
Yeah, but you know how many times I've put those PCIe lanes to use? Only long enough to transition from 660Ti SLI to a new 980. SLI and Crossfire support is on its way out, and I don't plan on putting anything else in those other slots.
A lot of people that went with the 5820K did so because it was the "cheap" way to 6 cores/12 threads and it was also the first way to get quad-channel DDR4 from a consumer point of view. Anybody who actually wanted to use PCIe lanes went with a 5930K instead. I mean, that's the reason I went that route, too. My tower isn't just a gaming machine, it's also a studio audio machine that needs to handle a lot of sound processing without budging. It's been a small disaster with its increasingly weird activity the past 6 months or so, though, and hasn't exactly been a gem from the start, either.
#11
Posted 06 October 2017 - 08:54 PM
#12
Posted 06 October 2017 - 08:57 PM
#13
Posted 06 October 2017 - 09:31 PM
#14
Posted 07 October 2017 - 03:32 PM
Sephrus Shanadar, on 06 October 2017 - 09:31 PM, said:
Pics or it didn't happen man......
#15
Posted 07 October 2017 - 05:57 PM
Bill Lumbar, on 07 October 2017 - 03:32 PM, said:
I tell you what Bill. This isnt the first time you have called me out on my fps differance I've seen between my 1700 and my 6850k. When I get back from my business trip in a few weeks I will link the test vids using both cpus with the same gpu all drivers up to date. I should also be building my 8600k system so I will throw that test in also.
Edited by Sephrus Shanadar, 07 October 2017 - 05:59 PM.
#16
Posted 07 October 2017 - 09:28 PM
GweNTLeR, on 06 October 2017 - 08:54 PM, said:
Sephrus Shanadar, on 07 October 2017 - 05:57 PM, said:
on a side note, it seems to me that the CPU mapping doesn't work out very well on a ryzen 5/7 ... probably mostly due to what was described as the octocore patch .... which ended up being very hit and miss .... and seeing as PGI are not going to actually test on the platform at all, they aren't likely to improve the situation in much of a hurry ...
NARC BAIT, on 01 October 2017 - 11:04 PM, said:
the following is pretty much a quick edit of that TL:DR post ...
sys_job_system_max_worker = 16 sys_main_CPU = 1 sys_physics_CPU = 2 sys_streaming_CPU = 3 e_ParticlesThread = 4 ca_thread0Affinity = 5 ca_thread1Affinity = 6 r_WaterUpdateThread = 7 e_StatObjMergeUseThread = 8 sys_TaskThread0_CPU = 9 sys_TaskThread1_CPU = 10 sys_TaskThread2_CPU = 11 sys_TaskThread3_CPU = 12 sys_TaskThread4_CPU = 13 sys_TaskThread5_CPU = 14
or maybe an alternate stepping map will work out better for some people, depending on overall core load usage blah ...
sys_job_system_max_worker = 16 sys_main_CPU = 1 sys_physics_CPU = 3 sys_streaming_CPU = 5 e_ParticlesThread = 7 ca_thread0Affinity = 9 ca_thread1Affinity = 11 r_WaterUpdateThread = 2 sys_TaskThread0_CPU = 2 sys_TaskThread1_CPU = 4 sys_TaskThread2_CPU = 6 sys_TaskThread3_CPU = 8 sys_TaskThread4_CPU = 10 sys_TaskThread5_CPU = 12
honestly, right now, I'm not sure the 'proper' way to address the CPU topology from within the game, 0 seems to make threads bounce around, it was suggested to me based on an obscure piece someone else wrote that 0 goes to the core with the least load .... which might result in a direct penalty on a ryzen if the chosen new core for task X is on the other CCX ... setting it either way seemed to give me an improvement to *minimal* frame times over the 'auto' settings ... I'm reworking spreadsheets to figure it out one way or the other ... but life likes getting in the way of calc-foo ...
Edited by NARC BAIT, 07 October 2017 - 09:30 PM.
#17
Posted 09 October 2017 - 09:24 PM
#18
Posted 01 January 2018 - 01:05 PM
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users