A nice CPU with regards to performance, but compared to the 1950X it's ~25-30% more performance at a 100% higher cost and 20 fewer PCI-E lanes. It's absolute shit value.
Does anyone else have the problem that when you stress test the 7900x with OC that it doesn't even bother to try and clock the cores at target speeds and just let's them down at 900-1300mhz? Even with power limit and current limit etc set so that they aren't an issue anymore? (3000watts limit and 140% current limit) Temps aren't the problem either (both CPU and VRM)...
CPU intensive benchmarks and stuff works just fine tho (cinebench 2620 pts / 209 pts)
I wish, it took out my CPU (Core i7-990X) and RAM (Corsair Dominator GT 12GB (6x2GB) 2000MHz CL8) to it's grave with it.
Replacements in this case... not going to work.
- - - Updated - - -
There's something weird about the test bench according to Videocardz that very likely shouldn't be happening as others of the same series don't do it either.
The CPU auto-boosts to 4,2GHz on all cores to establish this speed, all other chips tested only boost 2 cores to between 4,2 and 4,4GHz and drop the rest down considerably... so there's a little bit of shenanigans going on there either by the "leaker" or by Intel themselves.
I'm guessing the leaker has been dicking about somehow.
Yeah that's a function available to low core count CPUs, but not on the HCC ones.
Since you'd have a voltage range and HUUUUUUUGE TDP difference.
- - - Updated - - -
Update!
http://wccftech.com/intel-delays-10n...cpus-end-2018/
Worth a read... Intel's 10nm woes continue.
As afaik the cores are still onlocked it could be a manual OC (would require a very hefty cooling setup (Atleast a custom loop Water solution if not something better) though seeing that would probably go well north of 400W usage i reckon (CPU only).
And from what i could see while playing with a Asus Rampage X299 mobo the core sync option still in the bios, and it doesn't look restricted with the higher core parts, and also the manual also doesnt list any restriction of this not being available on higher core parts. TDP does goes out of the window though if you use this option, so very, very good cooling is well advised.
I also dont really get the whole TDP fetish esp not with K/X cpu's or Overclocked AMD CPU's as (potentionally) TDP goes out of the window anyway when you enable any OC setting in the bios, even something like sync all cores that alot if not most motherboard manufacturers enable by default makes it so that you cannot look at factory TDP anymore. Only if you run stock settings on both the CPU and motherboard and disable any OC features that the mobo manufacurer has enabled by default you can expect do see operation to be always with or very close to factory specced TDP else it just depends on load and temperature management, but the CPU and mobo wont try to stay within factory TDP anymore.
Last edited by chronia; 2017-09-20 at 10:34 PM.
If the 7900X is anything to go by... it certainly will be.
But we'll see won't we? I'm absolutely curious to see the power draw for that.
Hmm ... from the EXTREMELY limited time I had in the same mobo the option was gone entirely from the AI Tweaker page and CPU management page.
Unless it was moved to a different page or perhaps the one I saw didn't have that option enabled yet.
That said though I'm curious to see the overall picture, the one recurring thing is that the current leaks keep stating the 7980X is incredibly low temperature compared to previous iterations ... but I've seen those statements on the 7900X as well with an example, look how that turned out.
That said... the 10nm barrier seems to be a complete bitch for Intel, they delayed it again.
Can't seem to copy from the PDF, but its in here, Page 3-15 Under CPU Core ratio http://dlcdnet.asus.com/pub/ASUS/mb/...875.1503942854
At this point i'm even wondering if we will see Canonlake anyway, i was already going to be used for Mobile Only, and now its coming so close to Ice Lake if these rumours are true (it still is WCCFtech, they are very hit and miss since they kinda publish every rumour they get theirs hands on, not the most reliable source at times) that it might not even be worthwhile to release canonlake, unless they make Canonlake Mobile only (which they already want to do) and Ice Lake desktop only for the first period.
Last edited by chronia; 2017-09-20 at 10:50 PM.
Hmm... perhaps I was just blind then.
I certainly didn't see that with the 7900X I had extremely limited time with.
Regardless you use that to OC because stock voltage certainly isn't going to supply enough power for 18 cores on sub-1V for 4,2GHz.
That'd be both insane and pretty freaking awesome if it could though.
Regarding 10nm... I think it's entirely possible as Intel has had MAJOR difficulties with both 14nm and 10nm where Samsung and GloFo had far less issues with their LPE/LPP 14nm and even the 7nm that was developed in conjunction with IBM from Samsung/GloFo's part...
Because according to the latest reports the 7nm IBM/Samsung/GloFo will enter mass production capability in "early 2H 2018" ... which if it's true on both that front and Intel's delay will NOT be a good time for them.
That roadmap for Globalfoundries is wrong because they announced today that they were starting with 12LP (12nm FinFET) next year.
https://www.centralcharts.com/en/new...e-applications
Last edited by Gray_Matter; 2017-09-21 at 12:14 AM.
I was referring to the table at the bottom where they had GF with 14LPP and 7nm DUV for 2018. That should be 12LP and 7nm DUV or possibly 14LPP, 12LP and 7nm DUV.
There are big advantages for a company like Intel to have their own Fabs. The problem is that sometimes the wheel turns and they end up in the position where they are effectively competing against 3 companies and it's easy to get in trouble.
Ah right... I was wondering why it sounded off and without you explaining the link would've been difficult to make :P
I believe I read somewhere a few weeks back that experts believe that AMD will hold the lithography advantage for a year or 2 - 3.
Most people assumed that Intel's 10nm would be smaller when the actual space difference is in the 7nm advantage so the 7nm is actually "more advanced".
(I really REALLY don't like that term when speaking about uArch Lithography)
Its really hard to compare nodes by name though, as Intel is quite conservative, while the other parties let marketing do their thing alot more and give their nodes names "better than they actually are". ASML has a formula to rank nodes to a standard node value instead of the factories naming scheme. For example Intels current 14nm is actually more close to 12nm. And there you can see that Intels 10nm while it will loose the advantage is actually not that far behind as the naming scheme will tell you.
While the dates might be wrong (it doesnt state if its projected release dates, or a first production date, as Intel is already running test runs on 10nm with other products), but Intels 10nm should actually be quite close to what others call 7nm. Intels 10nm process is actually more around 8.3nm, while the others their "7nm"processes are between 7.8 and 8.2nm. And Intels upcoming "7nm" node (which will come late, 2020 are the rumours)" is actually rumoured to be 5.9nm while using this formula from ASML.
Last edited by chronia; 2017-09-21 at 06:33 AM.