AMD CES 2019 Keynote Live Coverage

Page 5 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.



My thoughts ...

The power consumption they showed was system power - not CPU package power.
A 9900K that is allowed to pull as much watts as it wants scores just over 2000 in Cinebench.
A 9900K that is actually restricted to use "95w" or so scores just under 1800.
The demo 9900K scored just over 2000 points.

So we know that the 9900k processor itself was unrestricted and likely using a reasonable amount more than 95w. 130-140w is commonly seen, and some stress testing will push this CPU well over 200w consumption - this is known.


The Ryzen system used ~130 W and the Intel system ~180w during the test, according to the graphs.
We can reasonable guess that the i9 was using maybe 135w (very likely) to get the score that it did.
That leaves 55w for the rest of the system.

130w (Ryzen demo CPU) minus 55w (system power) = 65w.

It actually seems more than plausible - it actually seems likely. I think Lisa has actually left us just enough breadcrumbs ... :)

Lisa's style sure seems to be to not let anyone see all the cards in her deck until the time is right ... considering the launch is several months out, I'd say that is a wise strategy; and one that shows that AMD won't be throwing its consumers under a bus just to impress investors, like some other companies ...

 


AMD has and will. They are a company. If they don't please investors and share holders Lisa could be easily ousted. Do you not remember the QuadFX? Their answer to Core 2 Quad. Cost more than the best Core 2 Quad with performance at less than the lowest end Q6600 and used way more power, probably 2x as much.

I to this day don't understand why people think AMD is some consumer-centric company. They are now because its the only way to regain trust beyond the enthusiast market and to regain market share. Once they can, and probably will, regain a bigger share they will price accordingly. Yes Ryzen is a really good CPU. It really has thrown Intel for a loop but they have a lot more to do. There is marketing to get the brand out, to this day the majority of consumers have never heard of AMD but most everyone knows who Intel is (this is similar to people with Creative Labs who makes some of the best audio products on the market but are unknown by the masses) that alone is a massive advantage as the normal consumer doesn't look as in depth as we do.

And you know what? I wouldn't fault AMD for pricing accordingly. If they had an even 50% of the market from Intel and had performance that was even or traded blows then they should demand what the product is worth. Their biggest problem in the past few years was not having enough money for R&D. Its part of why they went FABless as there is no way they could compete in that space, look at Intels FAB42 cost. It was $5 billion to build and tool and is $7 billion to retool to 7nm.
 


I disagree with some of that ...

Don't mistake "not throwing consumers under the bus for the sake of impressing investors" with "not caring about investors". The two are not the same.

My reference is to current AMD under Lisa Su (I thought I was clear on that) ... not some ancient bad product decision by AMD from some 12 years ago as cited in your example. (but I'm with you on that - that was a stupid idea) But please don't conflate these two either - this is well outside the point I was making.

Jump to recent history (to the timeframe in which I am referring) and I am reminded of the 5ghz 28 core promise; the chip they finally launched, that I guess was supposed to be this one they referenced during that display, cost $4000 ... consumers where expecting a Threadripper competitor, not just another re-badged Xeon ... Investors responded well though, as that was probably goal.

The greatest way to gain traction over a giant, extremely well established competitor is to consistently offer more for less. AMD is executing this strategy quite well and only consumers win. I have noticed that many investors are complaining that AMD's revenue is too low and that they are low-balling Intel on pricing, so that, in their eyes, devalues the stock. A perfect example of giving consumers what they want despite what investors may think.

At the end of the day low prices = good for consumer, and gets a "meh" from investors - how is this not catering to the consumer before the investor? Your point about AMD and low prices feeds fuel directly into my point.

So while you wouldn't be critical if they raised their pricing a little, the fact that they have not (and in fact keep lowering prices it seems - zen vs zen+), is exactly what I am talking about here.

I do agree with you it in a sense that it will take years of offering more for less (no doubt this is a marketing strategy) before they are on equal footing with Intel, (assuming they can keep that up for years), and at that time it would make little sense to continue low-balling their product prices, as the "offer more for less" strategy would gain them a lesser advantage than it does now ... but we aren't there yet at all.


Note, I am referring to Ryzen ... Vega is a different scenario altogether and what I mentioned above doesn't apply as its just to costly to build very high performance variants - AMD doesn't have much choice there until they can get onto Navi (hoping the Navi will be more gamer friendly or at least have dedicated tuned variants for gaming, unlike Vega).


Edited several times to make additions and fix stuff ...



 

Investors have little to legitimately complain about here: a company that has been on the verge of bankruptcy for the better part of a decade is finally turning a profit. That's a pretty substantial value increase from trash stock right there regardless of gross income being nowhere near Intel's.

For the investors who wish AMD matched Intel dollar for dollar, performance for performance in the retail market, I bet I'm far from being the only one who wouldn't bother considering AMD if AMD was nothing more than Intel-II. If the competition doesn't add meaningful value to a product category, I stick with the original. I wonder how many sales AMD would lose if it lowered its performance-per-buck to Intel's standard. Probably enough to overkill gains in per-unit profit and cause a net loss.
 


That and they have brand name issues. As siad everyone knows Intel very few outside of enthusiast circles know AMD.

As well they still have a ton of work to do in the server and HPC market where Intel still heavily dominates and where more money is to be made.
 


Seems like Epyc is AMD's solution to gain market share in servers from Intel. So far, it looks to be working, despite slow server adoption, due to the nature of such markets.

 


I am not sure. My boss is content with sticking with Intel as its been the go to. The other issue is that most servers use VMWare now and you have to be careful when switching those out. You can't change one host out at a time you would have to do, most likely, a cluster at a time and they are not cheap individually.

Then there is the fact that Epyc is only 2S solutions for now, maybe with the third iteration they will move to 4S+, which leaves that market for Intel.

Only time will tell really.
 

Pretty sure Lisa said EPYC was only going to be 2S because 4S is a tiny piece of the global server market and unnecessary with EPYC bringing more IO and nearly as much performance in 2S as Intel had on 4S at the time, a situation which is going to be refreshed by EPYC2 doubling core count and bumping connectivity to PCIe 4.0.
 


They still have a long way to go and depending on a lot of factors it may or may not become a permanent thing. AMD has a lot of trust to rebuild and if Intel makes a decent play in the next few years it can make it harder for them to continue to take market share.
 
Moore's law relates to "the number of transistors on a chip", which I take to mean includes all the cores. Doesn't the much higher core count in today's chips compensate for smaller gains in shrinking the process node when calculating Moores law?

Perhaps not...it's not like the total chip or die sizes have gotten exponentially larger. Curious what someone who knows about such things thinks.