QOTD: How Would You Change AMD?
AMD, arguably the CPU company that's gone through the most significant changes in the last several years, has yet to scratch the surface of what it can really do.
More recently, AMD decided to establish a separate company to handle semiconductor manufacturing, and is now focusing much on designing and engineering, as well as emerging technologies. It's purchase of ATI was definitely a big announcement for the industry several years ago. AMD hasn't always been this bold however. In its early days, it mainly followed in Intel's shadows. This is in stark contrast from the AMD we all know today.
Granted, not all is rosy. AMD is in the middle of a heated disagreement with Intel over the use of Intel's x86 technologies. Financially, it's still up and down.
The question of the day is: How would you run or change AMD?
Would you have purchased ATI? Perhaps another company, or none? Would you attempt to even create a whole new CPU architecture? Would you also have formed Globalfoundries? Let us know.
Buying ATI was a great move and now nvidia is shitting due to the competiton from ATI. I used to be a nvidia fan now im all ati!!!
plese keep up the good work and do something about the i7. i donnt wanna spend $1000 for a rig!
on the high end they should turn their gpu into 64-bit cpu with hundreds of computing cells. can you imagine what kind of games could be made with a cpu that had 256 cores.
re: next five years
they certainly could partner with china now and get money that will keep them going and building a brighter future..
ATI: Great move
New PRocessor Architecture? Absolutely not.
AMD needs to build a stronger portfolio in the x86 space. they need to build a crown-winner. They also need to get on board with Intel in the sense that they need to invest in die shrinks. They're lagging behind in architecture, engineering, and performance. All of AMD's innovation is gone. They need to step up the game.
Buying ATI was a great move and now nvidia is shitting due to the competiton from ATI. I used to be a nvidia fan now im all ati!!!
plese keep up the good work and do something about the i7. i donnt wanna spend $1000 for a rig!
2. Opteron server chipsets need to be released YESTERDAY. No one wants to buy Opterons based on nVidia chips now that nVidia & AMD are enemies.
3. Southbridge RAID for enterprise -- Intel's RAID SUCKS. Intel's chipsets have been terribly unstable, and their RAID has serious problems like the cache settings won't survive a reboot, etc. AMD could do better and gain a following here.
idiot... you think a monopoly would result in lower prices?
I think AMD should have gone skipped phenom I and skipped straight to a 45nm Athlon 6400X2... but then with HT3.0 ofc and the other phenom features... I bet that if they had skipped phenom I, then Phenom II would JUST start coming out right about now- or maybe a few months from now... but we would be sitting on Athlon 6800x2's or something running at 3.4-3.6 ghz at stock speeds... and overclockable to higher speeds than the E8400/E5200.
But as for the future, they need to look into merging th compare and jump instructions like intel did, its just too big of a performance gain. Im seriously surprised they haven't done it yet.
They also need to see how easy they could add multiple threads per core to the chip. This is essential in the server world, not so much for your average user.
Beyond that, work on cache latency, needs to be brought down.
--------
On the manufacturing side, their 45nm process seems strong. However, they need to execute 32nm flawlessly and get it out the door ASAP.
I think they should focus on new and improved products instead of rebranding and also creating more low/mid range stuff. Take a break from releasing “mid-range” processors and spend ALL that cash on surpassing Intel / Nvidia.
ATI already had the next two generations in the works, maybe more. Their price wouldn't have gotten much lower...
That said.
I'd like to see AMD make their own architecture. A true 64 bit (maybe even step up to 128 bit or something else) architecture that they could design to be backwards compatible with x64 (which is theirs anyway) and design a new platform for it but keep the standard open. That way Intel could go hither with their x86 licensing, AMD could make an open platform of the future for itself and maybe even license out to Intel or other companies so they could make products for their platform.
I know you could say that they don't have the support to go forward with a unique platform, but if its made well and holds great advantage over existing platforms, it could work.
2) Show your employees how valuable they are to you(much more then the useless management)
C) BREAK AWAY FROM THE BULLSH!! ADVERTISING MODE. SHOW THE BENEFITS OF 64-bit processing, because Intel's crack at it SUCKS! IF I could get a 16 core 64-bit processor running 2.0 ghz for under 200 w/ motherboard (not ram) I would sell my teeth instantly.
GREEN) Most of the time, management doesnt have a clue what the market wants. They've spent too long managing employees and getting over paid and underworked. Start listening to your engineers. Start listening to the Customer Service agents, start listening to everyone but stock holders and corporate management....most of the time, they are honestly... useless
APPLE) ATi is known for taking what works, throwing more at it and selling it at current NVDA prices. Break that mold and get more specialized. Get in talks with Newtek or whomever to help your 3D Rendering packages. The common user isn't the guy spending $20,000 its the one spending 100-200 dollars...
For CPUs:
K8: The last K8s would be 90nm, on Socket 939 with a DDR controller and two types: Single and Dual Core.
K9: It would be 65nm, the introduction of Socket AM2, a DDR2 controller, and a CPU with two cores or four cores designed kind of like the K8 but different: all cores would have two or three treads (two for Athlon X2, three for Athlon FX) 1MB or 2MB of L2 cache per two cores and 4MB L3 cache and make the quad cores just like the Core 2 Quad, put two Dual-Core K9s on one die. This would have beaten the Conroe.
K10: Use the same architecture but build a true quad core (just like today) but introduce Socket AM3 (which is backwards compatible with AM2) and a DDR3/DDR2 hybrid memory controller. Oh and 45nm. It would be built in the same way as the Phenom II but have many other additions that AMD didn't have time to incorporate like Multi-Treading (three treads or more per core) and I would have a Phenom FX which could mop the floor with Core i7.
For chipsets:
I would have made more 6xx series chipsets, not just the 690G
And today I would have a similar but very different AMD, one that wouldn't have gone the way of Global Foundaries because it would have a lot more money to play with and at this time I would have completely dominated the server market. And K11 is coming with 32nm or even jump over 32nm and go all the way to 22nm beating Intel completely.
Ahhh, one can dream, hehe!