Quick update on this: going back to Auto was buggy, the system simply kept the values I had already set manually. I did however set the TDP to 65W, also disabled the Short Term boost past 65W to simulate continuous load.
With this setting Package power never went above 65W and CB 15 score was 1226. The cores kept jumping between 3200 and 4300Mhz.
Quick update on this: going back to Auto was buggy, the system simply kept the values I had already set manually. I did however set the TDP to 65W, also disabled the Short Term boost past 65W to simulate continuous load.
With this setting Package power never went above 65W and CB 15 score was 1226. The cores kept jumping between 3200 and 4300Mhz.
For some loads like current games the cache may help right now as MT distribution is still optimized for 4c loads, but as they get better threaded we might see them become more reliant on memory performance. Also, as already pointed out, a test meant to look for 3-5% differences between memory configs should also include the top GPU on the market (the 1080 is a good data point though).Some of their results seem strange. It also doesn't jibe with the Skylake memory scaling chart Anandtech put together years ago. Only x-factor I can think of here, is that the 8700k has more l3 . . .
Yes, in my view it should be a must. The CB score suggests the 8700 operated with roughly 15% lower clocks when capped to 65W. Depending on load this difference will vary though, from 0% to 30%+.So it looks like you'll need a board that supports cTDP to maximize performance even if you don't plan to OC. This will be relevant if looking into upcoming H and possibly B series boards.
Yes, in my view it should be a must. The CB score suggests the 8700 operated with roughly 15% lower clocks when capped to 65W. Depending on load this difference will vary though, from 0% to 30%+.
This is MSI specific bug I've seen on my previous MSI Z board as well, some power related features tend to remain tied to previous manual values once they get set to Auto.My main interest would be in the standard default behavior, which you claim doesn't actually work, but follows you last settings.
It would be quite a big problem if it wasn't capped by default, remember the tiny stock heatsink that comes bundled with 8700? It doesn't stand a chance at more than 65W.Though, from what you have shown so far, the evidence isn't in, if it is actually capped by default. You had to do it manually.
So I did a Load Optimized Defaults and modified no setting other than some needed adjustments to fan behavior.Load Defaults option and report back.
I peaked at 85C on Noctua NH D14 with 3 consecutive runs of CB15 on "stock" config. Don't make me mount the stock cooler (aluminum core)The older Intel stock 115X coolers were rated at 82W for the aluminum core E97379-001 and 95W for the copper core E97378-001.
It turns out this board does not enforce the 8700 TDP with stock settings. Package Power in CB 15 climbed to 110W+, score was 1410. Checked power limits via software and they aren't even set. The only limit is Max Current at 135 Amps. If this CPU had been mounted with the stock heatsink the only throttling it would do would be ... thermal.
Some of their results seem strange. It also doesn't jibe with the Skylake memory scaling chart Anandtech put together years ago. Only x-factor I can think of here, is that the 8700k has more l3 . . .
Please dont say that, this is why I dropped Ryzen and went with Intel. I am totally not in the mood to go through any of those troubles again.IMC is apparently slightly different from Kaby Lake, given that G.Skill's new "Coffee Lake-compatible" lineup of TridentZ RGB has different subtimings despite using the same IC (Samsung B-die), particularly for tRAS. Either that, or the Z370 platform has some teething issues.
Source:
https://www.anandtech.com/show/11904/gskill-launches-lineup-of-trident-z-dimms-for-coffee-lake-cpus
IMC is apparently slightly different from Kaby Lake, given that G.Skill's new "Coffee Lake-compatible" lineup of TridentZ RGB has different subtimings despite using the same IC (Samsung B-die), particularly for tRAS. Either that, or the Z370 platform has some teething issues.
Source:
https://www.anandtech.com/show/11904/gskill-launches-lineup-of-trident-z-dimms-for-coffee-lake-cpus
Andrew: Because this is a refresh platform, Intel has somehow made it even more mature, so it will be even more stable than the previous generation. From what I can see, this refresh doesn’t mean a bad thing for users. Usually, the kind of customers that already have a Z270 board will probably not upgrade for this generation. For the majority of Z370 users, maybe they have been using their PC for three or five years, so these are still good products for them to upgrade to.
So I did a Load Optimized Defaults and modified no setting other than some needed adjustments to fan behavior.
It turns out this board does not enforce the 8700 TDP with stock settings. Package Power in CB 15 climbed to 110W+, score was 1410. Checked power limits via software and they aren't even set. The only limit is Max Current at 135 Amps. If this CPU had been mounted with the stock heatsink the only throttling it would do would be ... thermal.
I peaked at 85C on Noctua NH D14 with 3 consecutive runs of CB15 on "stock" config. Don't make me mount the stock cooler (aluminum core)
Please dont say that, this is why I dropped Ryzen and went with Intel. I am totally not in the mood to go through any of those troubles again.
It's the same exact memory controller. The platform is virtually identical aside from the changed pin-out.
See: https://www.bit-tech.net/features/tech/motherboards/asus-interview-andrew-wu-rog-motherboard-pm/1/
Went back into BIOS and changed 1 setting only: CPU ratio from Auto to 46. Rebooted and ran CB15: score remained the same ~1410, but Package Power dropped to 90W+. Temps dropped accordingly by 3-5C. At this point it's becoming clear I'm evaluating the MSI BIOS rather than the 8700 CPU.
And my review rating is... BETA!
PS: XMP is working again.
This is kind of what I would expect from a high end desktop CPU.
Thanks for checking, though it may be different on different MBs.
I just got my Fedex# from Newegg for the 8700k. So excited
25-26C ambient temp. Case is Fractal Define Mini C, with 3x140mm + 1x120mm fans. Top lid is off and the heatsink is oriented to make the most of vertical airflow. Idle CPU temps are around 35C with main fans off. No dGPU installed yet.What were your ambient temps? Is this in a small case? I'm using a Noctua U14S in a Fractal Define S.
25-26C ambient temp. Case is Fractal Define Mini C, with 3x140mm + 1x120mm fans. Top lid is off and the heatsink is oriented to make the most of vertical airflow. Idle CPU temps are around 35C with main fans off. No dGPU installed yet.
As said before, my setup is silence oriented, all fans are running at low RPM, so it should not be taken as a point of reference for performance/cooling oriented builds.
<sigh> When I saw this I clicked to add to cart and in that moment it went out of stock and told me it had been removed from my cart.
I have an amazon backorder still from October 5th.
Three Arctic AC F14 fans and one of the 120 mm fans that come with the case. All of them are voltage controlled, motherboard supports both PWM and DC, has 1 CPU fan header, 1 Pump header, 3 System Fan headers. One System Fan header is situated near the center of the MB, the rest are near the margins, thus easy to access via the case cable routing holes.What case fans do you have, and how are you controlling them PWM/Voltage? Motherboard or fan controller?