I didnt bother reading any of this thread after I saw the typical avatars. QQ team red and QQ team blue going on I take it. LOL.
I didnt bother reading any of this thread after I saw the typical avatars. QQ team red and QQ team blue going on I take it. LOL.
Just as an FYI I do not post here because of TEAM RED or TEAM BLUE, I post here to correct the incorrect information given out.
My post history should've actually shown that to anyone who's known me for a while objectively.
I bash equally on both brands in terms of CPU quality but I also correct the other when absolutely atrocious stuff is called out.
Hell in example of current discussion I even stated Intel's 8700K to be underperforming and me believing it will likely be higher and I'm still called an AMD fanboy for that.
Go figure with the logic side of things...
Regardless I will attempt to reduce my "walls of texts" if it allows to be done, I will however not ignore incorrect information being posted because when that happens that information propagates to people whom actually either want to learn or know for whatever reason.
All I want is a decent 6/12 from Intel with good single thread performance. I don't see any reasonable need for 8/16 at the moment.
I just hope this silly cores race won't hurt single thread performance.
The whole problem with the world is that fools and fanatics are always so certain of themselves, but wiser people so full of doubts.
And it will be replaced in less than a year by Z4XX and so on.
Honestly, why do you care? When it comes to CPUs - high end ones really age well, simply because they are usually way overspecced for what you actually need. People buying in insane number of cores and shit and proceed to boast how they have 30% utilization when they crank everything up, which is total foolishness as I see it.
if you buy high-end coffee lake or skylake-x CPU - it will be easily enough for next 4 years.
http://tieba.baidu.com/p/5328908012?...d=112180312773
https://www.anandtech.com/bench/CPU/1603
6900K - 1437
7800X - 1333
Hard to believe it has 95W TDP. I expect some pretty hefty spikes in temp, like 7700K had on 'stock' tim and voltage.
Mm, that 1400 score seems more like what it should be pulling. will be interesting to see where it will OC though.. And the wattage it will require to get there.
MMO-Champion Rules and Guidelines
Like I stated before, which unfortunately is lost in argument, it's unlikely it's 95W TDP...
Currently the estimates fall exactly within my estimates because of the simple fact that it likely has 0 IPC improvement over Kaby Lake.
Which means that TDP is closer to ~130 - 135W range and 95W being some shenanigans by Intel like basing it on the non-turbo base clocks.
Even though technically the "base clocks" on motherboards will be the All-Core Turbo of 4,3GHz... it's a little stupid.
- - - Updated - - -
Expect it to be around 4,8GHz with a small +/- 100MHz variance.. and consider the power draw to be about 160 - 190W
TBH my question lies more with previously mentioned paper launch or not and what can be done if it's TIM between the IHS/die.
What many people tend to forget is that the temperature spikes caused by those chips it's largely due to the lack of thermal conductivity from the die to the IHS and not just because it's the uArch design so the whole "shitposting lack of solder" attitude and claiming Intel doesn't use solder because of thermal spikes and cracked solder is largely invalid.
(Not saying you said it, just making a general statement)
Depends if they improved the manufactoring proces again. Even with Kaby Lake all samples i tested had no issues with temps on stock settings (so just using the normal boost clocks, not base clock only). OCing its gets different though, while the average is stil around 4.7 / 4.8 Ghz without delid, but temps are alot more over the place between the samples. Some do really good, some are quite a bit worse when OC'd. But from what i can tell its not actually mostly the TIM (As Intels TIM looks to be quite good, just not liquid metal good) causing this, but variance in the space between die and IHS due to the glue layer. If Intel finally manages to get this right (and remove / make that variance smaller) the temps between samples should also be alot more consistant.
From what i can tell atleast removing the glue layer gets the temps down by quite a margin already. Adding a slightly better TIM is just icing on the cake.
Like I said .. the actually efficiency improvement has to be ~60% to be able to remain in 95W TDP over Kaby Lake.
That is generally only possible with a process shrink, being on the same process and uArch that's almost impossible.
But regardless the TIM actually causes the temperature spike because the silicon cannot actually push the heat away fast enough through the TIM.
That causes heat to build up and poor through more slowly than it could be.
The TIM itself is actually pretty garbage in terms of throughput ... but Intel's TIM is based upon lasting forever rather than best throughput.
Almost every aftermarket TIM has a higher throughput but has to be renewed every couple of years, same for the Liquid Metal AFAIK.
So the choice is either such a TIM or solder it and if you compare the difference between them it's rather gigantic but ignoring the max temps of either if we just focus on the "spikes" you'll see that they are almost non-existant because the heat can be transferred fast in comparison to the TIM setup.
Note again: Ignore the max temps difference, simply look at the temp spikes when normal TIM is applied and Liquid Metal.
Hell ... you can even apply higher quality normal TIM and see the spikes get severely reduced.
And those spikes are what Intel (and others) call the reason for solder cracking, which is in part true no doubt but the chance of it and length of time required is a LONG ass time.
Why do you think the Xeon chips are soldered (or were prior to Skylake-X)?
Of course the space that is higher due to the glue is also a factor but that doesn't change the spikes, that simply alters the actual max temp.
But the 60% improvement required is why I don't believe Intel's 95W TDP for Coffee Lake is true on advertised specs as the jump is generally a rather large lithography jump... and doing that on both the same uArch and lithography is generally not a thing.
(and yes uArch design is also extremely important in power efficiency improvement.. whatever others may state on this forum)
Also with heavy heart I have to inform you my Intel Core i7-990X system died this morning.
A VRM burnout occurred after it didn't want to power on and post (1 sec power on and turning off, repeating indefinitely) and with it going it it took my RAM and CPU along with it to it's death.
(It looked like a Roman Candle firework when the VRM failed, burned a needle sized hole clean through the motherboard along with of course the blackened ICs)
I have yet to be able to test whether or not it took my graphics card along with it and other peripherals but hopefully those should be fine.
Fuck ... this is really at a bad time for me
Itll likely perform just about the same. Its not really a new architecture. I wouldnt expect to hit higher than 5ghz.
- - - Updated - - -
I know, right? If they priced that thing at like.. ~140/150, tops, it might be worthwhile. At the rumored 185$, i dunno wth Intel is thinking.