It is not as much fun to be in the server part of Advanced Micro Devices these days, with Intel surging in the server racket and expanding out to switching and storage with its Xeon processors and Intel more or less counting the substantial innovations that AMD's engineers crafted for the Opterons a decade ago. The good news if …
AMD has never had a shortage of good engineering
AMD's biggest challenge is execution, not engineering. While they have allowed Intel to take center stage in recent years in the server market, Intel isn't known for delivering the best engineered products but they are good at copying AMD's superior designs. AMD needs to get their act together across all market segments as consumers desire better products than what Intel is offering and at fair prices.
I used to love AMD CPU's
Then Intel pulled a Microsoft and got several game companies to have special optimizations in the game that relied on pieces that the PIII processor had (Battlezone II was one game I know) that seemed to make the games play slower or unreliably on AMD processors.
Re: I used to love AMD CPU's
WTF were the downvotes for?
Of late, Intel has had a reputation for providing chips with better floating-point performance than AMD. Since this isn't one of the big issues for database servers, I don't know if that is going to be addressed, and I would like to get some reassurance from AMD there.
The main problem is simple.
No matter how good chip AMD produces, Intel current chip is always on a better smaller process, which means that its beats any advantages the AMD chip may have. If AMD have 45nm Intel will be on 32nm and so on. Until they are both on the same process node, AMD is screwed!
I think one only needs to look at the performance of 22nm vs 32nm Intel Core i5/i7s available today to get an idea of how small the gains from process shrink really are. Don't get me wrong 2.5% is nothing to scoff at, but it's not revolutionary. Add to that the heat dissipation issues associated with the smaller surface area and I think one could make a compelling case for the argument that we have entered the era of diminishing returns in process shrinks.
One could try to make the argument that smaller chips are cheaper to make (which has historically been the case), but the development costs associated with shrinking die size are going up a an alarming rate. This has been discussed in several articles here on el reg as a matter of fact. While I agree with your conclusion that AMD is in serious trouble, I think it really comes down to $$$ spent on engineering, not die size. Intel is outspending AMD at such a rate that even the thought of AMD bringing something competitive to the table should be seen as a complete embarrassment for Intel.
Re: The main problem is simple.
Actually: NO your conclusions are wrong.
The process node or gate size is only one of many issues that can impact performance. Back when CPU's used 90 nano process dropping the gate size meant increased frequency for the same thermal envelope. Now days APUs on 32nm can offer as good or better performance and lower power consumption than 22nm - as AMD has proven with Trinity laptop vs. both Sandy bridge and Ivy Bridge. So you can forget the false belief that process node is of real significance now that we are at 32nm and dropping because the gains via just the gate size is not significant any longer. There are many more importance design aspects than process node now.
As AMD showed with Trinity laptop, they ain't "screwed" at all because they have better engineering. We can expect the same with the new AMD server and desktop designs. AMD already leads in desktop APUs and even though never intended for severs, APUs have proven to be an excellent package already for certain servers.
Re: Random k
Based on your theory then, Intel should be very embarrassed as Trinity laptop APUs are by far superior to Ivy Bridge in overall performance. IB was only a 5% bump over Sandy Bridge and has thermal issues and doesn't OC well, so anyone who believes Intel hasn't got issues is out of the information loop. AMD certainly has execution issues also but Trinity is a good 25% bump over Llano and Vishera should be a 15+% bump over Bulldozer. Compare those figures to 5% for IB and the technical issues with IB and we see that Intel didn't do so well with their first tri-gate CPU design just as AMD didn't do so well with Bulldozer.
How can you write this without mentioning ARM?
execution, not engineering ????
"AMD's biggest challenge is execution, not engineering" ??????
I couldn't disagree more. Marketing people over-estimate their own ability and talents.
Like HP, there were too many marketing heads at the executive level and innnovation was stifled in favour of "cover-your-backside" short-term gains.
Giving more emphasis back to core engineering is the longer term solution and putting in an experienced engineer at VP is a good start. However it would have been better to clear out the old incumbents and put him at the top.
Re: execution, not engineering ????
Since around 2007, AMD's problem has definitely been in execution.
Their first generation chips (Barcelona, Bulldozer) have had performance affecting bugs or have been delivered late (Magny-Cours, Bobcat) or both. Each time AMD have failed to execute, they have fallen another 1 or 2 quarters behind Intel and lost a few more points of market share.
That's not a marketing issue.
It'd be good to see AMD rise again.
I've been buying Intel chips for a bit now and it still feels dirty.
Last core i remember using was the clawhammer. I think they went into decline since then.
Intel is the 800 pound Gorilla in the room
So if AMD wants to survive, it has to outsmart Intel rather than outdo or compete directly.
That exactly was the case back when AMD introduced its initial Opteron processors and those were the best days of AMD. Intel tried to shift the 64-bit world towards Itanium and AMD made a smart move to go with x86 in the 64-bit game as its resources could not suffice to compete with a different ISA. Intel stuck with the stupid front-side bus in order to sell more chips, AMD introduced HyperTransport and the memory controller in the CPU. Intel was strictly proprietary, AMD chose a more open path with HyperTransport by engaging others to the game. Intel developed its own processes, AMD collaborated with IBM. Intel could pursue the hot magehertz race by being really aggressive on the process front, AMD could not and chose a wiser path of higher IPC - less megahertz that proved correct.
At some point, one would expect Intel to put its act together (as it did) and at that point AMD fell behind expectations. It seemed that its engine lost its steam. One failure after the other. The biggest mistake is that when AMD started to do well, it thought it could be Intel in the place of Intel. Unfortunately that was not the case.
AMD has been "going out of business" according to some folks... for the past 40+ years. the reality is they keep coming up with better products that meet the needs of the masses and thus they continue to be of significance. Intel may be the 800 lb. Gorilla but they make a lot of Biz mistakes due to their arrogance in addition to their criminal ideology.
AMD is regrouping under Rory Read and already pumping out some better products with Trinity and in just a few weeks Vishera. Consumers will be the winners as AMD gets back on track.
Re: AMD regrouping
So say we all.
Piledriver looks good
AMD seems to be back on track and hitting their projected 15% performance bump per iteration. Piledriver and Steamroller will achieve these goals so I expect even better in '14 when they can ship newer designs.