AMD Ryzen: Good For Enthusiasts, Bad For Investors

[mitch074]

With the manufacturing process being new, and the core design too, AMD didn't have the cash to validate more than one design - the 8-core one was the only way to really make a difference and gain mindshare among enthusiats and workstation users - the most vocal groups in tech.

After a few runs, it is more than likely that defect rates has fallen down enough for them to consider a "true" 4-core design that would save on wafer sizes, and allow them to get the lowest R5 and R3 out - and that's where we still don't know if they'll go CPU-only (save on wafer size thus better margin) or APU (same wafer size as Ryzen, but with one CCU and one GPU on die).

The second is probably what they're working on for mobile chips, as we'd get "real" quad cores (with SMP) with a solid GPU; throw in a dual channel controller and some reasonably fast DDR4, you'd be able to run most games (with reduced details of course) on a single-chip laptop - the very thing that we could almost-but-not-quite-because-cpu-sucked enjoy with Godavari.

 

[InvalidError]

They could have sold all 3 models for hundreds more, but they needed the minds and market share of high end desktop users because they were the laughing stock of the industry.

I think current Ryzen prices are right where they needed to be, maybe a bit on the high side for the 1700X and 1800X since they provide next to no benefit over the 1700 on top of not having a stock cooler included. Intel's current prices have nothing to do with silicon and engineering costs, they are the product of a competition vacuum which may take a while to break due to Ryzen not systematically wiping the floor with Intel's CPUs across the board in benchmarks.

At the same time, AMD desperately needs to improve margins so it can turn a profit and pay off some soon-to-be-overdue debts, which means it can't sacrifice more of its margin than absolutely necessary for market share.

But even the 1800X is cheap relative to the transistor count

On a per-transistor basis, the 1700 is still extremely expensive compared to other chips. If you want cheap, DRAM comes in at ~$0.70 per billion transistors and MLC NAND undercuts that at ~$0.14 per billion. Transistors are cheap. It is adding all the extra metal layers on top that drives costs up for irregular ICs like CPUs. For a fairer comparison, mobile SoCs ring in at ~$25 per billion, which is still much cheaper than a 1700.

 

[Burstaholic]

Not convinced. Is the margin difference really that large? I think the difference in the physical bill of materials between 8 and 4 cores should be tiny - surely the silicon itself is not nearly the most expensive thing here.

What costs a lot of money is engineering expertise and factory tooling - it might actually be much cheaper to build and maintain six assembly lines that make 8-core chips than two that make 8-cores, two for 6-cores, and two for 4-cores.

It's usually cheaper at large scale to maintain many identical things than the same number of things that are all a little different. Without being an expert in semiconductor manufacturing, your premise seems pretty shaky.

 

[InvalidError]

What costs a lot of money is engineering expertise and factory tooling

These are one-time costs that get amortized over production volume. If you have high enough production (and sales) volume, those costs become an increasingly smaller proportion of total costs, especially for modular designs that are intended to scale with minimal re-engineering effort. A new set of masks cost a few million dollars to make from engineering changes and validation to first wafer, which dwindles to under a dollar per chip when you produce tens of millions over the design's lifespan.

BTW, AMD made a new die for the R-540/550 which is 101sqmm vs 123sqmm for the RX-460/560. If AMD could afford the re-engineering and tooling costs to save 18% in die size on extremely low margin products, I think it can easily afford doing the same to reduce 4C Ryzen's size by 40+% where its margins are several times better.

Either AMD is so happy with Ryzen's margins at the moment that it cannot be bothered with optimizing costs (that could turn into a costly decision long-term) or it had no confidence that sales volumes would be good enough to justify the costs of having two die designs and picked the one-size-fits-all approach for first-gen Ryzen. Likely both.

 

[iam2thecrowe]

They could have sold all 3 models for hundreds more, but they needed the minds and market share of high end desktop users because they were the laughing stock of the industry.

I think current Ryzen prices are right where they needed to be, maybe a bit on the high side for the 1700X and 1800X since they provide next to no benefit over the 1700 on top of not having a stock cooler included. Intel's current prices have nothing to do with silicon and engineering costs, they are the product of a competition vacuum which may take a while to break due to Ryzen not systematically wiping the floor with Intel's CPUs across the board in benchmarks.

As a product, if you removed the branding from the equivalent priced Intel and AMD cpu's to a consumer, and called them brand X and brand Y, then yes, maybe they are priced not too badly. Unfortunately, the average consumer sees Intel as the premium and trusted product. To claw back market share they are going to need to reduce the price. In Australia the Ryzen 5 is approx the same price as an Intel i5, with no integrated graphics and at a lower clock speed. For my needs I would buy the i5 every time. Its also more than enough power for 99% of PC users. And that leave a bunch of 8 core models 99% of the world doesn't need....

The other issue that has always been Intel's competitors undoing, is Advertising and consumer awareness. If AMD dumped a whole lot more resources into advertising and consumer awareness they would reap the rewards. They have not done enough advertising with any cpu they have had to date, even when it outperformed Intel's offerings. I work in a tech industry, and most of my colleagues are relatively tech savvy, but are totally unaware that Ryzen exists, or what it is about. Now I don't think that AMD can rely on only word of mouth to sell their product.....

 

[rhysiam]

... the Ryzen 5 1600 IMHO makes it very difficult to recommend an i5 build at all. ...

How do you come to that conclusion?
I ordered a Ryzen 5 1600 for myself just the other day. I think it's a fantastic multi-purpose CPU at just the right price point. Compared to a Core i5 at the same price I get now roughly the same (more than sufficient) gaming performance and decidedly better productivity. Comparing these very same CPUs again three years from now I expect the Ryzen to be better in just about every aspect.

That's exactly what I'm saying! "Difficult to recommend an i5", meaning that a Ryzen 5 1600 makes much more sense right now. On the forums I'm rarely recommending i5 builds to anyone anymore. For almost any use-case the Ryzen 5 1600 is straight up a better buy.

 

[InvalidError]

To claw back market share they are going to need to reduce the price. In Australia the Ryzen 5 is approx the same price as an Intel i5, with no integrated graphics and at a lower clock speed.

Which Ryzen 5 and i5 would that be? Based on US pricing, if you compare the R5-1400 against similarly priced i5, you get four extra threads, unlocked multipliers and 2MB extra L3. Intel's unlocked i5 are closer in price with the R5-1600 which has two more cores, four more threads and 8MB more L3 than Intel's much more expensive i7-7700k. For about the same price as the i7-7700k, the R7-1700 gives you performance closer to Intel's $1000 i7-6900k in applications where the extra threads matter.

If those aren't sufficient benefits to convince you to get Ryzen for your next build, then you aren't in AMD's target market. At least not yet.

Marketing-wise, AMD doesn't have much money to spend on that and the bulk of PC sales are through PC OEMs. As long as AMD can convince Dell and friends to put Ryzen in some of their PCs, AMD doesn't need to waste much of its own money on marketing but with most PCs shipping without discrete graphics, that won't be happening at full scale until Raven Ridge comes out, which will hopefully also mean much cheaper 4C4T/4C8T CPUs.

 

[synphul]

Some folks may be waiting for ryzen 3. Many were hoping for i5 or i7 equivalents for the price of fx cpu's around the $90-130 range. Currently the cheapest r5 is $170+ which for what it is isn't exactly poor value but it means for amd fans the cost of admission went straight to i5 pricing levels. Prices on older fx may have dropped some and no doubt fx's ipc is nowhere close to ryzen. Currently folks comparing to fx are seeing 6c fx 6350's for <$100 and fx 8350 8c's for <$125.

Only those who haven't been paying attention to performance benchmarks or don't understand that core speed/count isn't the only metric to compare cpu's by would likely be put off by this. Fx and ryzen are clearly apples and oranges. For those who see fx as 4 cores with the 4350 for $75 or those who dispute it and consider each module to be less than 2 cores, the fx 8350 is still significantly less than the 4c/8t r5 1400.

In short amd fans have grown accustomed to cheap cpu's and amd isn't giving them much at this point. Starting price at $180 is well above what budget fans have experienced for the last 5 years. I would expect it to be similarly difficult if it were intel releasing a new batch of cpu's and no sub $200 pentiums or i3's, just i5+ or nothing.

 

[Olle P]

That's exactly what I'm saying! "Difficult to recommend an i5"...

My bad! I totally misread it!