• Happy holidays, folks! Thanks to each and every one of you for being part of the Tom's Hardware community!

AMD CEO Lisa Su Interview: Confident In Next Graphics Launch, Zen's Success Is Key

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.


They most certainly did. Before 1995 or so, if you bought a 1 GB HD (I bought one, it was a SCSI 1024 MB and it cost me $1,000) Until around 1995, 1024 was the standard used by the industry. Before that, the 1000 MB GB did not exist in general usage, advertisements, articles, etc. It is a "correct" definition as per the metric, base 10 system but computing was always and still is a base 2 system.

The argument at the time was it was "too confusing" for customers, But you can bet that if it took 976 MB to make a GB, you can be sure this "adoption" would never have taken place.

If wifie hadn't thrown out all my old Byte , Computer Shopper mags .... I could quote the uproar that occurred when this switcheroo was made.

Odd that we have "adopted" this definition for HDs.... well how come 4 GB of RAM still has 4096 MB and the metric definition wasn't adopted there ? How come this isn't too confusing for customers ?

How many liters of water can you fit in a 2 litter bottle ? .... 2
How many gallons of water can you fit in a 10,000 gallon pool ? .... 10,000
How many GB of data* can you store in a 4 GB RAM stick ? .... 4 GB
How many GB of data* can you store on a 500 GB HD ? .... 488 GB

*Ignoring MBR, SPD and other data needed for unit to operate.

 
AMD share of the consumer marketplace is dwindling fast. Intel has 90% share of laptop CPUs, 80% of Desktop CPUS, and NVidia has close to 80% share of the discrete graphics card sales. AMD's share is also getting worse by the day, losing 12% market share of the discrete graphics cards in the past year alone. Finally, in a absurd cost cutting effort, AMD has sliced their R&D expenditures down to their lowest level in 10 years even though it is competing in so many, many markets.
All of these slides are well presented, upbeat, and show new technology. They are also just cheap PR presentations until there are actual products that are for sale. AMD needs a lot more than slides to regain market share and make a profit again.
 

But that "definition" has always been technically incorrect. The CompSci field misappropriated the metric prefixes as good enough approximations for their purposes in most circumstances.

The discrepancy is not odd at all: HDD capacity is a function of arbitrary platter density, arbitrary number of heads and other arbitrary design/addressing decisions so actual HDD capacities hardly ever line up with powers of two and you also have the "lost" capacity to file system structures so people never see the label disk size in their OS anyway. People were already quite confused even before the GB/GiB switch.

With RAM, managing chips with irregular geometry would be a nightmare, so all practical memory chips have whole address lines and consequently, device sizes always grow in exact powers of two. For these, it does not matter which prefix gets used because there is no possible ambiguity even if you tried. If I say 524MB of RAM, you will automatically round that down to 512MiB since there is no possible combination of standard memory that can yield 524MB, MiB or any possible remotely reasonable variant thereof.
 
The point is we had an industry standard for the 1st 14 years (post IBM PC) of the consumer PC industry and then all of a sudden some manufacturers pulled a switcheroo. There was no noble purpose here; the change served one and only one purpose...... to deceive the public. If it was good enough to be the industry standard for 14 years after the IBM PC, why the switch ?

The alleged reasoning (confused consumer) was false. No one was confused from 1981 to 1995 and this was done for the sole purpose of making the consumer think they were getting more than they were. If 1000 MB = 1 Gb than I have to ask, how is it that "everybody had it all wrong" up until 1995 ?

The fact that the HD size discrepancy question still appears on boards even today is proof that consumer confusion has been created not eliminated. I'd have no problem if they started out that way from the getgo.

This change was not initiated by the nerds in the back rooms who designed and built the things, this change came out of the suits in marketing. Disk manufacturers took different stands on the issue as some argued to stay with 1024 and others 1000. In the end the holdouts succumbed as sitting on the shelves, 170 MB/GB looked better than 160 MB/GB.

By the logic presented, gallons should be 60 ounces, nice round base 10 number, instead of 64. That's probably not far off as, what do we have on supermarket shelves today ?

-A box of pasta now contains 13.5 ounces instead of a pound (16 oz).
-That "pint sized" container of Tropicana now contains 14.5 ounces.
-That bag of sugar now contains 4 lbs instead of 5
-That big bottle of juice that used to contain 2 quarts now hold 1 quart, 28 ounces
-That can of tuna that used to contain 6 ounces now has a weight of 5 ounces, net 4 ounces after draining the liquid.

Yes, they all contain the government mandated correct weights *in fine print* but few would argue that the intent here is not the hope that the consumer isn't noticing that the price is the same but they are getting less. When this first started, I remember seeing the * and fine print explaining that in this context a MB = 1000 not 1024 MB.
 
AMD right now is in a relatively power position. Sure they are not in a position that enthusiasts like, they are in another position. This being the low end market. There current 28nm offerings are simply amazing. Decent CPU/GPU performance at 15w for $50 with a $35 mobo. Its good enough to do anything except for modern 3D Games and heavy computation workloads.
 


AMD should focus in my opinion on the cheaper gaming market. I think fighting high-end battles is not worth it because more people buy cheaper graphics cards than expensive. We need to remember that we here are enthusiasts but in the real world many people go with things like the 270.
 


Are you just making up percentages or do you have sources?
 
Wish she had more to say about Fiji, it looks like it will not launch with the rest of the re-brands on the 16th. When it does it had better be outstanding or AMD will be in for a rough year in the graphics market.
 
http://store.steampowered.com/hwsurvey is one source

This is as of end of 2014.... no doubt it's grown
bNqJYgA.png




 
I have issues with articles and people talking about all this technology, Sure HBM sounds great, yes it is stacked or whatever. But at the end it is just a technology, what user cares about is the end performance. If i have a GTX960 and i am running into issues because it is not powerful enough can i buy a $200 AMD card that will be powerful enough to run the games that i want. That's really all we want to know, we don't care about the technology you use.
Click to expand...

Then you've subscribed to the wrong site...
 
I am looking out for AMD G-card with there next GPU. If the rumor are correct.
The Fury and Fury X. With HBM
As for AMD next CPU architecture. Well that to far away around a year if the smoothly execute there plans.
14nm is near a full step down compare to 28nm.
So they can put a lot more on there DIE's then ever before. So even if they IPC get that much better. You can compete with a equal dieshrink much better. If iNtel goes to 10nm then this equalness would be short term. But it will be a half node. If IPC are on par. Then it still difficult to beat iNtel. But keep the possibility to compete and have better ASP.

As for Fury ( X ) I wonder if it wil support the Direct3D_FearureLevel_12_1 and more.
 
AMD should focus in my opinion on the cheaper gaming market. I think fighting high-end battles is not worth it because more people buy cheaper graphics cards than expensive. ...
Click to expand...

Wow do you ever have that the wrong way round. 😀 Low-end cards have much smaller margins. Many store owners survive on the profits made on high-end parts, because that's where the real money is. One store owner in CA told me 40% of his profit comes from high-end CPUs/GPUs, whereas he has to absorb a loss on things like consumer HDDs in order to keep customers coming in who eventually buy other stuff.

Plus, the GPU biz is a trickle-down affair. They need to attack at all scales, but they'd never survive financially if they stuck to the low-end, etc.

Apart from the rebranding (which sucks, though both sides are guilty of that), my worry for Fury is that AMD will again make the mistake of sacrificing noise, heat & power characteristics in order to have something which competes with NVIDIA on speed, ie. what they did with the 290(X); they can get away with this to some extent if they have a price advantage, but otherwise it'll put many people off. Users care a lot more these days about noise, power/heat issues (at least for enthusiasts) have knock-on effects on oc'ing, and many people are concerned about power consumption now, especially as high-end builds are approaching the limits of a typical US power socket (thankfully not a problem yet with 240V/13A areas). NVIDIA had a big win with it's stock cooler for recent product lines, whereas AMD's stock coolers have been terrible, something AMD really has to get right this time.

If anything I have more confidence in what they can do with Zen. I don't necessarily think they'll come out with something which stomps on Intel, but if they manage to be on a par with comparable pricing (certainly with any price advantage) then that could pull back a fair bit of custom. Doubly so if they release lower-level unlocked parts which Intel has ignored for years.

Anyway, we shall see. Atm all of this is just a pitched battle of fear, hope & bias. 😀

Ian.

 
All this said about AMD, they seem to have nowhere to go but UP! The stock's at an all-time low, but they show no signs of going out of business. I think that's a good sign to throw some money at their stock, no matter what you think about their GPUs or CPUs.
 


God this sounds very familiar. It is almost the same as what was said back when AMDs stock hit less than $2/share a few years ago. Then it started getting better. Then worse again.
 





Very interesting comment. I just scanned Dell and HP websites (i merely glanced, maybe i missed it) but i did not see any AMD APU desktops there. At least i do not see these APUs on desktops in mainstream PCs either, maybe they appear in laptops. So it's like enthusiasts don't buy AMD APUs and at least i don't see them in mainstream desktops either.

AMD reminds me of Sprint, no matter what they do they just screw up.
 


The APUs are more designed for lower end desktops/mobile devices. The more current one is more of a UMD based CPU.

AMD doesn't seem to have anything in Dell/HP right now short of their laptops. Maybe they are not able to supply them as much as they want. I know that when the Athlon 64 was the big thing to have, AMD was having supply issues as they only had one FAB in Dresden, Germany pumping out CPUs and Dell/HP require a lot of CPUs since they sell a lot of their PCs.
 


The month of November 2012 was the only time in the last few years it ever fell below $2 (and not by much). If you would have bought at that time and then sold at practically any time from May 2013 to September 2014, you would have doubled your investment.

But if you were good, bought in November of 2008, and sold in April of 2010, you could have turned $5,000 into $25,000. Other than the 2012 and 2008 lows, this is the lowest, which means they've always bounced back and made people money.

https://www.google.com/finance?q=NASDAQ:AMD
 
The claims of 40% higher IPC are always to be taken with a pinch of salt. For example:

Is this against Excavator, the performance of which is generally unknown right now?
Is L3 cache factored into the results?
Is this per core or per thread?
What workload(s) is this applicable to? (I imagine dumping shared FPUs will yield a significant performance boost, well above 40%)
Are clocks being reduced to more sane levels in order to keep power usage in check?
How much of this increase is from improved cache, esp. with regards to write speeds?

(etc. etc. blah blah)
 


Carrizo will probably tell us more about Excavator soon.

I would assume a lot of this is based on the entire CPU as a whole performance wise and not just the core. Most references to IPC are including the L3 cache.

And the only boost removing shared FPUs will give is in multithreaded performance as the threads will no longer have to share them. That is of course only in say 8 core Zen vs a FX8000 series. Single threaded could benefit if they are dropping back to the K10 as a base design, I do remember K12 was supposed to come after K10.5 but they scrapped it for BD and now it is Zen, and doing improvements to the uArch and cache.
 


When your last marketing effort caused massive hype for a product that massively underperformed, then yea it is a good idea to be humble. They were not humble with the FX series at all.

And you get the wrong view of cars salesmen. I used to sell cars and most are decent people trying to make a living. It is the small used car lot jerks that give them a bad name.
 


The FX Series came out under the previous CEO. They booted Rory Read and promoted Lisa Su during Q4 of last year (October, I think), so I don't think her humility is tied to Read's tenure. It's probably just how she is.

http://en.wikipedia.org/wiki/Advanced_Micro_Devices
 


I know that. I was talking about the company as a whole. The CEO is the head but is not the only one talking about it. AMDs marketing has been much quieter this time around. In fact since the failed launch of BD they have been much quieter. Steamroller and the other APUs have not gotten as much hype as BD did and that was still under Rory Read.

Even with a great CEO, marketing can still cause hype to build up.
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom