The discussion is not about
The discussion is not about 7nm vurses 10nm it’s about first generation Epyc forcing Intel to lower its server SKUs markups/pricing and rhus taking a hit in the quarterly gross margin figures. And that’s already starting to become apparent to Wallstreet in those Market Quants long term figures that also take into their models Intel’s past margin history in the face of, at that time, AMD’s Opteron sales and that 23% server market share that AMD had with Opteron. Intel’s gross margins fell well below that 60%+ gross margin figure with Intel having around 51% gross margins by Dec 2006 the year that AMD’s Opteron sales and market share was at its highest.

So looking at Those Opteron market share figures and Knowing that Epyc is far closer to Xeon than Opteron could ever hope to be is what has got the Wallstreet folks moving their Intel gross margin guidence lower. First generation Epyc and that Platform advantage of 8 memory channels per socket and 128 PCIe lanes for both the Epyc/SP3 MBs with 1(1P) or 2(2P) sockets so that Epyc/SP3 MB/ecosystem puts first generation Epyc at an advantage even before Epyc/Rome is scheduled to arrive. Epyc’s pricing is also lower than Xeon for those Epyc 32 core SKUs and first generation Epyc is not far off in performance from Xeon on some workloads and leads Xeon on many others.

The cost of Epyc alone will force Intel to have to lower its markups in order to compete with Epyc on the price/performance metrics and Intel currently cannot compete on the 8 memory Channels and 128 PCIe 3.0 lanes that all the Epyc platforms come with Standard as AMD has not segemented Epyc by memory channels or total PCIe lanes.

Rome will be fesable as AMD are no fools when it comes to designing for the future but Rome will undergo the same sorts of certification/vetting that all server SKUs from any maker are subjcted to by the customers’ hired engineers and engineer consultants, software and hardware.

Rome with more cores is very feasible due to that Scalable Data Fabric and the Infinity Fabric. Server wrtkloads are not gaming workloads and most Servers are NUMA aware as are the software systems. Look at all the server OEMs that have come onboard with Epyc and that is a great indicator that Intel will be forced to lower its prices and take a gross margin hit in the face of the stiffest competition that it has ever had.

And I’m not even including the Power9 competition or that ARM ThunderX2 competition that will force both AMD and Intel into the lower markup range for low power servers. AMD is much better able to compete on the lower cost end than Intel who can not afford to lower those margins what with all of Intel’s massive fixed costs in having to maintain fabs and costly middle management pay/perks packages etc.

Intel will have to try with lower prices to match Epyc in the price/performance metric and even the TCO(Total cost of ownership) metrics, Intel can not currently match the features like 128 PCIe 3.0 lanes or 8 memory channels per socket that Epyc/Platform offers. And currently the Meltdown fixes are costing many server customers millions in lost performance because the Meltdown mitigations mean the it will take more servers kept online to make up for the lost performance from those mitigations. The one Spectre mitigation has almost no affect on Epyc’s performance is not going to hurt AMD as much! So that’s making things look even better for more Epyc adoption and a faster adoption rate also for AMD’s Epyc based server products based OEMs.

Are you sure Rome will be a
Are you sure Rome will be a feasible SKU?? or it will be a cpu of nice for some rare applications.
TSMC 7nm “early” is suboptimal and only 7nm “performance” will give some real gains but this will happen only in 2020.

Moreover there is any sign of desktop cpus on early 7nm, this is pretty obvious because there are voices that the clock speed is not high enough to beat GF 12nm.

So my prediction Intel will stay in a confortable 2019 and as soon as TSMC will ship 7nm “performance” it will have a 10nm fully functional. We’ll see in 2020 for both AMD and Intel.

they consumer sales have been
they consumer sales have been slow. the server, workstations are where they are making the money. AMD is slowly pecking away at the workstations. apple also has a lot to do with their workstations. since they charge consumers well well over the price of the chip and the cost of the aluminum they use. HOWEVER their employees are NOT as happy as they seem. they took a huge loss on the X series that was put out to outperform the ryzens.

Keeping those margins above
Keeping those margins above 55% for Intel is vital or Intel is going to have to cut its costs in a painful manner. AMD’s Epyc sales biggest hit on Intel will be in the area of higher margins and that’s going to force Xeon markups downward in order for Intel to keep Epyc from getting rapidly back towards the market share that AMD has at the height of its around 23% server market share Opteron days. AMD’s Zen/Zeppelin dies/wafer yields are so high, above 80%, that first generation Epyc’s pricing latitude compared to Intel’s Xeon can not be easily matched. And Intel’s margins were most affected the last time by AMD’s Opteron competition.

Looking at the old Intel Gross Margin History charts and what Opteron was able to do to Intel’s gross margins and that should have any Intel investors worried. Epyc is orders more closer to Xeon than Opteron was in raw performance, and that’s not even considering the Memory channel and PCIe lanes advantage that AMD currently offers with first generation Epyc platform 1P and 2P motherboards.

Epyc looks set to take even more market share faster because of all the pluses the frist generation Epyc platform offers. Those Zen/Zeppelin wafers have been in full production at 80%+ die/wafer yields for a good while and the tweaks/steppings and available quality of the bins is going up. I’d say that the first 10% of Epic’s server market share will be on First generation Epyc sales alone and maybe 10% is being too conservative owing to the interest in AMD’s Epyc from the server/HPC/AI markets.

Achieving 61% margins in the absence of any meaningful competition is not hard and Intel has had even higher gross margin figures than 61%. But Looking at the Epyc benchmarks and the Feature advantage that the Epyc Platform has relative to the 1P and 2P Xeon/MB offerings tells me that Intel is in for some lower margins over the upcoming quarters as Intel can not piviot fast enough to stop Epyc from taking more server market share even before Epyc/Rome sales begin to add to the gains in 2H 2019.

Once Epyc/Rome supplants Epyc/Naples for top spot on AMD’s lineup then AMD will be in a position to fruther lower the pricing of its forst generation Epyc/Naples SKUs and have them compete more fiercely with Intel’s lower tiered Xeon Offerings! So that AMD and Intel inter-generational server SKU market competition will begin once Epyc/Rome assumes the Flagship position for AMD and that just makes matters worse for Intel in the lower tiered server market come the latter half of 2019 when AMD will be pricing first generation Epyc SKUs even more agressively.

“I’m not entirely sure how
“I’m not entirely sure how analyzing pretty black and white numbers for 2018 so far has made me biased towards Intel.”

‘Reality has a distinct Intel bias’.

It certainly does cost a lot

It certainly does cost a lot to own a fab, but so far Intel has made it work for them. Achieving 61% margins while running those monstrosities is still impressive. But I think that a company the size of Intel with that income is better off running their own for at least the next 5 to 10 years. Yes, the cost of keeping one will continue to rise at a pretty drastic rate as the technology advances. For now Intel has more direct control over their product stack and that has been good for them. Consider that being fabless, AMD had a strong quarter but their profits were only $116M from $1.76B in revenue (6.5%). Now compare that to Intel's $5B net from $17B revenue (29.4%).

So while some of what you state about Fabs is certainly correct, I think that Intel's position is such that your analysis on how it affects the company is off. Currently it is not an anchor weighing them down, but it certainly could be if they do not get their manufacturing in order for 2019.

Intel’s got margins to look
Intel’s got margins to look out for and a high business overhead operating its own fabs and other expenses that AMD does not have. AMD’s turning profits on just 33-37 percent gross margins and Intel has big revenues and big operating expenses that will require maintaining those 60% gross margins. AMD being fabless has none of those worries about the billions necessary to maintain expensive chip fabs and it’s the third party fabs that are responsible for keeping their fabs operating at full capacity. AMD’s got TSMC and GlobalFoundries and can even use Samsung if necessary with that modified contract with GF that allows AMD to have more fab capacity options. GF has stated that they alone can not supply AMD with enough leading edge fab capacity.

Those first generation Zen/Zeppelin dies will remain in production for a good long while as for Epyc and servers there is that 5 years of guaranteed CPU SKU availbility that the server/HPC/Cloud markets demand. AMD’s Ryzen Pro SKUs will also have that extended product availability so GF will still be turing out some 14nm production while it transistions to 7nm along with TSMC. GF is doing a good business on its FDSOI line of fab services and togather with TSMC/Samsung that’s plenty of fab capacity for AMD’s future needs.

Intel just lost a billion dollar fab customer as a result of that 10nm delay and Intel’s got some very large fab expenses that will bring down any earnings if AMD’s server market share begins to get larger faster than expected. AMD’s first generation Epyc has passed the vetting/certification process with flying colors and has now begun its first few quarters of production usage. Intel will have to begin to lower its markups/margins in its cash cow server/HPC market Xeon lines jsut to compete of the price/performance metrics with AMD’s Epyc abd there are some things like PCIe lanes amd memory channels offered on 1P and 2P server MBs that Intel still can not match and 1P and 2P servers make up a good majority of the server market so that’s to AMD’s advantage.

If Intel beins showing any unusual gross margin declines then Wallstreet will devinitely take notice, and Intel’s 10nm fab process node delays will put them even fruther behind in getting to 7nm-5nm while AMD is already beginning to sample 7nm parts in sdvance of full production in 2019.

Record revenue numbers can be shown by just keeping up with inflation if those profit figures are not looked at more closely and Intel is a very large company so those revenue streams are going to be larger. Stock buy backs can also modify any per share figures so that has to be factored out of any company’s quarterly figures.

The Wallstreet quants are measuring the fundamentals of Intel’s performance in the face of renewed competition from AMD’s Epyc competition and that’s measuring by the quants includes many different model scenarios. And the quants will run many scenarios and look at the consensus and maby that’s why Intel’s share value too a hit. You can Google Intel’s Gross Margin history charts and just about Pinpoint on those charts where and when AMD’s Opteron server competition had its high water mark relative to Intel’s lowerst gross margin figures on that gross margin history chart.

Well, let me ask you a few

Well, let me ask you a few questions…

Did Intel have a bad quarter? Does it look like Intel is going to have a bad year? Ryzen is competitive and has lead AMD to have some historically good quarters, but not great quarters. Why has it shaken out that way? Even with 10nm issues, most analysts still see 2019 as a $70B year for the Intel, and they should end 2018 at $69.5B. So even with the uncertainty of 10nm, why is Intel still valued so highly and expected to have another strong year?

I want to see AMD thrive in the years to come and have meaningful competition for Intel. The results from the past year have shown Ryzen to be successful, but not nearly at the levels that Athlon 64 was after its release. It is a very different market in 2003/2004 than this past year. If 10nm is even more delayed and Intel's response is slow coming, then AMD should have a banner year in 2019.

I'm not entirely sure how analyzing pretty black and white numbers for 2018 so far has made me biased towards Intel. The financial results from the past year give a pretty accurate accounting to how both companies have fared.

I really have hoped for PCPer to become more unbiased, but you still have business interests with them, so it is understandable (although unethical) that you would not have them look bad.

EVERY chip manufacturer plays
EVERY chip manufacturer plays pretty fast and loose when it comes to the nomenclature of their particular “process node”. Accuracy and consistency don’t apply here, just marketing. Don’t fall for their game.