TaintedSquirrel
[H]F Junkie
- Joined
- Aug 5, 2013
- Messages
- 12,691
Little ridges help, apparently. I saw it on my D15 and thought it was a defect so I looked it up.
Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
Was the problem Koduri? Or was the problem Lisa Su eliminating resources from the GPU group to reassign them to the CPU group? Koduri moved to Intel because Intel is giving him the resources he needs.
Check also
https://www.hardocp.com/article/2016/05/27/from_ati_to_amd_back_journey_in_futility
You would read your own links, whereas paying attention to the difference between actual sales vs projected sales.
Even if you have no time to read the articles that you bring to us. At least check the table in that article. There is only one table in the article and it states that AMD Desktop Unit Share was 9.9% in 4Q16. So even accepting those "12.9% in 4Q17" as actual sales, RyZen gave only a ridiculous 3% share gain during the whole 2017.
And as stated before those numbers don't consider the recent impact of CoffeeLake sales. AMD won 3% marketshare the past year and then lost it. Now AMD marketshare is back ot pre-Zen time.
One has to know how to use Passmark marketshare data correctly. Clickbait articles took the early July marketshare value to generate false headlines. I discussed those false headlines here in the forums. I prepared the next graph
showing how the clickbait articles mentioned only the original fluctuation (31%) in the database to spread false information. Once the fluctuations went away (it took several days) the graph reported the final stable value, which of course the clickbait articles ignored.
The passmark graph also suffered fluctuations the first days of January of this year, but I expected the fluctuations to vanish and I started mentioning the graph only when the marketshare for 1Q18 was stable.
Of course, passmark share isn't the only indicator that shows the same trend: Intel is back and stole marketshare from AMD, thanks to CoffeeLake. Check the Amazon and Mindfactory stuff given in #4183
It just looks like CNC tooling marks, and they didnt spend the man time finishing/polishing it, but Noctua should know what they are doing and I dont think they would cut corners so maybe the new tech is reverting from mirror smooth on the HS base and needs more surface area for the tim?Little ridges help, apparently. I saw it on my D15 and thought it was a defect so I looked it up.
no that is their theory and not only theirs. it the "more surface area" theory....
I would be incredibly surprised if there were any measurable difference at all. These are very minor effects in the overall scheme of things.
45W Coffee Lake-H Series
Core i5-8300H: 4C/8T, 8M L3 Default 2.3 GHz, All Core 3.9 GHz, Single Core 4.0 GHz
Core i5-8400H: 4C/8T, 8MB L3 Default 2.5 GHz, All Core 4.1 GHz, Single Core 4.2 GHz
Core i7-8750H: 6C/12T, 9MB L3 Default 2.2 GHz, All Core 3.9 GHz, single Core 4.1 GHz
Core i7-8850H: 6C/12T, 9MB L3 Default 2.6 GHz, All Core 4.0 GHz, Single Core 4.3 GHz
Core i9-8950HK 6C/12T, 12MB L3 Default 4.3 GHz, Single Core 4.8 GHz
Xeon 2176M 6C/12T 12MB L3 Default 2.7 GHz, 4.1 All Core, Single Core 4.4 GHz
Xeon E-2186M 6C/12T, 12MB Default 2.9 GHz, All Core 4.3 GHz, Single Core 4.8 GHz
Intel UHD Graphics 630
Source: www.chiphell.com/thread-1817327-1-2.html
And now we get mobile chips boosting higher than desktop chips(8700k) Holy SHIET
So the 4/4 i3-8300 is launching at $170 USD.
Meanwhile, the 4/8 2400g is launching for about the same price, meaning the slower and Vega-less R3 1400 will have to drop down to at least $125... on a motherboard that costs half as much.
So the 4/4 i3-8300 is launching at $170 USD.
Meanwhile, the 4/8 2400g is launching for about the same price, meaning the slower and Vega-less R3 1400 will have to drop down to at least $125... on a motherboard that costs half as much.
This is the cut-off point where I'd switch my recommendation to Ryzen- the lower clockspeeds will bite a little for gaming, but not nearly as much as not having enough cores in the first place.
Contingent on AMD getting performance up with more pedestrian RAM, though.
High speed ram is so meaningless tho. To get the higher speeds they have to increase latency and there comes a tipping point that the latency hurts more then what you gain by higher speed. Only on a apu does the higher speeds tend to make much difference and with HBM now there is a even better fix for the memory bottleneck on a apu. After most of the bios updates tho memory support improved quite a bit and 3200 speeds are pretty easy to hit now, beyond that it can be much more difficult.
You're right that the speed itself is meaningless- except in the case of Ryzen, where RAM speed increases decrease internal latency. That needed fixing, because the CPU itself didn't need the extra memory bandwidth.
While you are right that the fabric speed and latency is affected by the higher clocked memory, it's only more noticeable from 2400 to 3200 speeds. People managed to get to 4000 speeds and it made little difference in benchmarks compared to going from 2400 to 3200.
That is the normal behavior. First because 2400-->3200 is a higher overclock than 3200-->4000. Second because the cache/memory subsystem in Zen bottlenecks passed the 3200MHz.
Your not very good at math, 2400 to 3200 is a 800 difference or the same as 3200 to 4000.
The CPU portion of the 2400g will perform best at 3200 cas 14, which will be just as good as 3600 cas 16 or even faster speeds with cas 17/18.
However, the igpu will want the fastest ram, with latentcy having a smaller impact. There, it would be nice to see 3600 mhz or higher supported.
I believe AMD was running a 2400g at 3600 mhz ram on one of their slides.
Yeah with a igpu higher speed is better then low latency and why you see AMD using 3600 speed ram. But HBM being used will eventually make that no longer necessary. It will be interesting to see how the igpu evolves over the next few generations.
Do we have any information on an AMD APU with HBM hooked up for the iGPU yet?
Intel's solution is a full daughterboard...
Does it also note on which "fixes" they are using from Intel/Windows or if they are not patched at all?
Whiskey Lake is based on 14nm and is coming in summer~ apparently - no idea what it entails - just that it is 4 cores. I can't imagine Intel pushing back 10nm products any further, if anything they are trying to make up for lost time.
10nm just wont hit mainstream until they can get EUVL tech installed. Right now the only way to make them is using lots of masks and it creates yield issues with bigger chips and it's takes quite a bit longer to make that way as well. I expect they will stay delayed until the end of the year where I think they will finally enter 10nm production for server and desktops chips. Intel tried to make it work without the EUVL tech and it just didnt work out and why everyone is catching up to them now.
Intel has said in their recent Earnings Call (yesterday) that it is currently ramping(shipping low volume skus) and full ramp is in H2'18, and expect a Gross Margin hit in H2'18. I doubt Intel will use EUVL tech until 10nm++ or 7nm. While it is ramping - I doubt the products you or I want will arrive this year.
what is EUVL? so if silicon fix will do the trick then hope theres no delay.. so 10nm+ still scheduled for end of this year then?
SemiAccurate...