AMD CEO vows 'maniacal' chip-baking fix AMD's newish president and CEO fairly flaggelated himself for his company's failure in execution during its last quarter, a failure that caused Intel's only real competition to do the unthinkable: leave money on the table. "We saw both 32 and 45 nanometer supply challenges during the third quarter," Rory Read told analysts and …
AMD release a chip aimed fair and square at heavily threaded server loads.
A load of enthusiast sites grab one, run a load of lightly threaded desktop/gaming benchmarks over it and find that, er, it's not very good at that sort of thing. Guess what? Bears shit in the woods too.
When the promised desktop variant ships with its clocks upped, more cache and a ceepie-geepie "Fusion" setup to push FP performance, then we can start to judge.....
"A load of enthusiast sites grab one, run a load of lightly threaded desktop/gaming benchmarks over it and find that, er, it's not very good at that sort of thing."
Actually, they ran a variety of heavy multi-threaded applications as well, to judge benefits of the "8 cores" vs the "4 core / 8 thread" of the i7 series chips. AnandTech is just one example. The sad part is that Bulldozer pulls close to the i7-2600K, but still falls behind the i7-980X (6 core, 12 thread). The even worse part is the i7-2600K can be OCed substantially with little effort, which then easily beats Bulldozer, which is already clocked so high as to only have another 200MHz of OCing headroom on air.
Therefore, it is not "execution" issues that have failed, it's their technology that isn't keeping up. Here's hoping their new tick-tock policy will keep them closer to Intel at a decent price-point. Sure, they're not grabbing the high-margin top-end, but volume and mid-market is where the revenue is at, which is where their supply issue comes into play. Unfortunately for AMD, where they're a K-series chip within range of an AMD chip, the K-series will win out easily for anyone willing to overclock, even by just 400MHz. A few of the locked-down chips still perform on-par or better than their AMD counterpart for everyday tasks, and meet or beat threaded performance as well.
Bulldozer was supposed to save AMD, but it hasn't. It's still clock-for-clock the same as a Phenom II, hidden by their high GHz settings, but has a better ability to multi-core and power gate. If they can compete price-wise by bringing costs and retail down, Bulldozer is strong enough to make AMD sit well in the bottom half of the market, which may be where they wanted to be all along. Performance crowns are expensive to keep, but if you can do nearly as good, but at a lot less cost, why not?
Is NOT Bulldozer but Brazos, which is why the CEO wisely slowed desktop production to fill the orders on mobile. I myself have recently managed to snatch up one of the EEE Brazos netbooks (A 1215B if anyone is interested) and it is sweeeeeeet! Dual core, radeon 6310, 720p screen with 1080p HDMI and VGA outputs, I was able to watch THREE HD movies on the 6 cell before needing to recharge, that's nearly 6 hours, and the thing barely got warm. oh and with an upgrade to 8Gb of RAM (Yeah I know its overkill, it was on sale and I couldn't resist!) and a nice briefcase to carry it in I got the whole smash for just $350 USD! heck there are even videos of guys playing L4D and other shooters on it with decent framerates.
So the key is NOT Bulldozer, which i'm sure they'll fix in the next rev like they did with the Phenom (Sold a lot of Deneb Phenom II desktops at my shop, they are quite nice and VERY affordable) but IMHO they need to be concentrating on Brazos and Fusion, both for the crucial mobile market AND for the low end desktop line. HP and a few others are now using Brazos for "all in one" PCs built into nice flatscreens and that is making those chips just that much harder to get. better to focus on Fusion and Brazos than worry about trying to get the high end gamer market which has traditionally been Intel's anyway.
...that maybe everyone at GloFlo and AMD have been working diligently to try and resolve production problems A.S.A.P. as everyone knows what's at stake. The fact that you simply can't wave a magic wand and fix these types of issues quickly should be obvious to anyone with a clue. Perhaps now that demand for AMD products is ramping rapidly, there will be better pre-production validation and scheduling to reduce or eliminate lost revenues in the future. Quite frankly with the demand for Llano APUs, the new Opterons and even the FX CPUs, AMD has never been in such a position before, so this is trial by fire for them.
Every Semi that I have worked for or seen in action have been the same. They are all driven by senior management marketing men to deliver the bottom line, with little real interaction with the engineers tasked with getting new products to market.
Middle mangers act like arse licks. Then the engineers have to figure a way to meet the numbers. Corners are cut, reliability goes down, while yield goes up. Great in the short term, but it comes back to haunt you.
One place that I have personal experience of, cut the leverage - the number of gross die per wafer - to increase the % yield. They still shipped the same number of passing die per wafer and achieved the same revenue per wafer, but management were pleased, because yield was up. Twats.
Give AMD a break. It will take time for them to properly figure out the root cause yield issues and that is no small task. I hope the industry recognises that and it was a brave and correct move by them to leave the money on the table rather than ship and be damned.
And I quote:
"We will continue an aggressive effort with out foundry partner to improve manufacturing performance at this important 32-nanometer technology,"
There you go. Get yourself a foundry partner AMD. It'll help.
For those of you reading without sarcasm enabled... Tut tut el reg proofread fail!
Where's the tombstone?
"Being a customer in my past job," he explained – Read came from Lenovo, where he had been president and COO – "it's about building trust:
This guys still gets work?
Let me break the news, as an actual customer. I couldn't give a flying rat about who the COO of XYZ pc corp trusts, doesn't trust, goes on day spa breaks with etc. When I buy PCs I'm interested in one consideration of my own and I'm not distracted by half eaten fruit, dancing pink radiation suits, or statistics on how many units shipped to businesses with clueless IT depts.
(£ / Ghz + £ / capacity) - your competitors cost for same or nearest quivalent
I *may*, in the event of a tie, pick the one that isn't trying to insult me further with a jingle...
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2024
