Intel Reports Record Second-Quarter Revenue of $14.8 Billion

>>61594796
>revenue=net inflow

What's the breakdown?

>>61594796
Your desperation is the best, keep it up!

-7% YoY desktop, clearly you know what happened.

Guess that explains why the AH trading is only 1% up for a 'record quarter'

>>61596337
7% drop in desktop revenues?
Wew.
I wonder what will happen Q1 2018 when full Zen stack is launched.

>>61596337
Well that explains where the extra revenue AMD got this quarter went.

We really need mobile chips though, laptops outsell desktops 3 to 1

>>61596434
AMD's at the mercy of Vega drivers and overloaded schedule for RR though, they're launching something every month.

judgement: fine

>>61594796
>no stock movement
Literally a "whatever" response to the ER

i wanna see the next quarter results

:^)

>>61596786
The Tech sector sellout likely ate up any stock movement for Intel, everyone's down including Amazon and Apple.

>>61596499
The actual reason for the ravenridge late arrival is the uncore power draw, which is really high for whatever reason.
The uncore on the 4 cores actually uses more power than the cores itself, it's baffling.
On the other side, it shows how ultra fucking efficient the zen cores are.

>>61597184
Well you see, Zeppelin has A LOT of uncore.
Fuck, EPYC powerdraw comes from uncore for the most part.
They deliberately made a chip that wastes most of it's power on moving data.
Kudos to Clark & Keller for designing core this efficient and caches this good (considering it's AMD and AMD == bad caches before Zen).

>>61597184
Uncore in BD designs like SR and EXV was pretty efficient, so I think these are time constraints in action.
If they couldn't handle the uncore power, they wouldn't even bother making a mobile design.
Also thankfully this isn't due to IF, AMD confirmed that IF under full load uses some 4 watts on a fully enabled Zeppelin die, they probably haven't given enough attention to the PCH and PCIe power draw, maybe even the memory PHY

File: 1484404940871.png (38KB, 499x338px) Image search: [Google]

38KB, 499x338px

>>61594796
Ahh yes intel brings in a large income thanks to the server market, the majority of all laptops on the market, large partnerships, etc.
>lmao jewtel made more $ fuck ayymd right guys
Shitters in /g/ anymore, its fucking ridiculous.

>>61597210
RR doesn't need all of that shit though. doesn't need the GMI links (unless there's some magical mcm apu coming) and it doesn't need more than 16 io lanes to be compatible with am4, 24 to be fully am4 compliant, might not need all of the built in controllers, I sure you could loose some of the 8sata, 4ethernet, 2 10gbe and whatever else.

Unless there's an extra memory controller for hbm, wishful thinking though.

>>61597495
IF is not fast enough to support bandwith provided by one HBM2 stack.

>>61597518
That means they have to sacrifice some pcie links for that? thank got it has 60+ of those to spare

>>61597518
direct attach it to the gpu section, use a cheap low clocked efficient version etc, not everything has to be perfect full power the whole time. It's not like the ccx will use hbm as an l4 cache with it's shit latency compared to ddr4. even an hbm1 stack would help an igpu greatly and probably works great as a marketing tool.

>>61597540
No, one HBM2 stack requires a pair of specific
PHYs. I mean IF inside the die is not fast enough to support the bandwith provided by one HBM2 stack.
>>61597561
>direct attach it to the gpu section
That may be a possibility.

>>61597518
IF can theoretically scale to GPU bandwidths

>>61598955
>theoretically
That's the point.
You need a physical layer wide and fast enough to handle such bandwith.