Three AMD Execs Leave The Company Following Management Reorganization

[InvalidError]

AMD need to build some CPU's with Horspower again and continue to push GPU power rather than power efficiency. Let the benchmarks speak for themselves.

AMD desperately needs to improve their IPC and power efficiency: as it is now, the $600, 220W FX-9560 lags behind the $130, 54W i3-4360 in many games. When you step into productivity stuff, the i7-4790k costs nearly half as much, uses less than 90W and beats the FX practically across the board.

There is no market for high-power AMD chips unless AMD at least doubles their power efficiency. With some luck, this might happen with the 16-20nm die shrinks.

 

[Eximo]

They never really lost the top performance slot, just that during that time AMD's chips were very cost-compelling.

I would not really call them "cost-compelling" when AMD was every bit as guilty as Intel for charging ~$1000 for their top-bin chips. They did have generally better performance per buck and performance per Watt than the P4s though.

The way Intel was so successful at merging the best from the P3 with the best from the P4 to create the Core2 and leap-jump AMD was impressive.

quote]Then AMD released the poorly received Phenom series. Twice the cores, but slower single core performance. Phenom II was their last great chip series.[/quote]
What? Phenom (K10) didn't have slower single core performance than previous dual-core Athlons (Brisbane). The fastest Brisbane at the time Phenom launched was the 5200+, which despite it's significant clockspeed advantage was roughly equal to the lower-clocked K10 on single core performance. No, that's not what was wrong with the original Phenom CPUs.

The two biggest faults of the first B2 Phenoms was the TLB bug (which didn't affect typical desktop end-users anyway) and issues with per-core frequency adjustments, which could cause stuttering when threads got shuffled to a down-clocked core until it spooled up. B3 stepping fixed the TLB bug in hardware (which resulted in a very slight performance penalty but not as bad as the software disable fix).

However that did not fix the stuttering. I have heard from several sources the stuttering could be solved by tweaking power management options in the BIOS or in Windows. But alas I do not own a classic Agena Phenom to test this. This was dealt with for the Phenom II/Athlon II.

You could get a contemporary Athlon X2 6000 or 6400 (2006)that ran at 3.0 and 3.2?Ghz respectively while the fastest Phenom (2008) at the time was a quad core at 2.4Ghz. Still struggling with it today, but that was definitely when single core performance was more valuable then multiple cores for the majority of software. Brisbane was a process reduction from 90nm to 65nm, and a power drop from 125W to 89W for the faster chips. When the Phenom II came out it made more sense to upgrade. But then we had AM2/AM2+ issues, so this wasn't an option for a lot of people. Buying a C2Q chip was a better upgrade all around, though I stuck it out and waited for an AMD miracle. Ended up personally going for a i7-950 in early 2010.

 

[synphul]

Hard to speculate if the change will be a positive or negative one, but at least they're actively trying. Leaving the players the same and expecting different results is very difficult. People kept trying to say the desktop market was 'dead' and the wave of the future were mobile devices and consoles but apparently not. The answer isn't to continue blundering along adding more cores since they're already failing with that approach. Hopefully they'll come up with something original and compelling to get investors involved so they'll have the resources they need for r&d to come up with a more efficient architecture. It's going to be really tough overcoming the stigma of being the 'budget' chip. They thought it would be a good marketing plan and I think it's bit them pretty hard. Most everyone equates amd with 'budget', not quality and that's precisely why apple would never consider them. Not until some massive changes have happened. Apple is all about premium, even if the cost isn't completely justified and it's worked for them. Between amd and intel, it's not hard to understand why they picked intel.

 

[childofthekorn]

Seriously the guy that said put them in an apple should think a little bit before responding. Firepro's were in the new mac, and it did nothing. Because no one is gaming on a Apple. They need the benchmarks and the boutique pc places like origin or ibuypower to start featuring equal parts AMD machines. However right now their CPU's can;t compete. Their GPU's did have a great push for a bit, but they are failing to follow up with it and keep things rolling. AMD need to build some CPU's with Horspower again and continue to push GPU power rather than power efficiency. Let the benchmarks speak for themselves.

Firepros aren't meant for gaming, its more for design work, static images and mesh type of deals. Its a workhorse in different avenues but doesn't stand up to the dynamic works of video games. Its more of a workstation graphics (autocad, 3d modeling/animation, etc).

 

[bit_user]

Firepros aren't meant for gaming, its more for design work, static images and mesh type of deals. Its a workhorse in different avenues but doesn't stand up to the dynamic works of video games. Its more of a workstation graphics (autocad, 3d modeling/animation, etc).

It's funny how this reputation sticks to workstation cards. The only differences between Firepro and consumer cards are: higher price, more RAM (with ECC), non-crippled double-precision arithmetic, slightly lower clock speed, and driver optimizations for CAD-type apps.

I doubt the driver optimizations actually hurt game performance much, if any. And while NVidia used a slightly different core for workstations/servers (that was more optimized towards GPU workloads), AMD hasn't been doing that.

See: http://www.tomshardware.com/reviews/firepro-w8100-workstation-graphics-card,3868.html (AMD FirePro W8100 Review: The Professional Radeon R9 290) and check the specs for yourself. If you look at the gaming benchmarks and simply scale the workstation cards' results by the clock speed ratio between the workstation and corresponding consumer cards, you get almost exactly the same results.

This proves that the impact on gaming is simply the 13% and 7% lower clock speeds than their consumer counterparts.

 

[eatmeimadanish]

What I can't understand is why they did not enter the tablet/phone arena. The future in computing is in that direction. PC's on the floor are being replaced by laptops and tablets. It's not a mystery why AMD is struggling.

 

[antilycus]

Intel did exactly what they planned. They pushed AMD out of the market by using Intels' endless supply of money and their bad business practices of monopolization and threatening of vendors. I will ALWAYS pick AMD, ESPECIALLY IN VIRTUAL SERVERS because I've had much better results on VM's that run on AMD architecture as opposed to the 4x's more expensive (1100 bucks just for a CPU, come on Intel?!). AMD has the GHz working their favor but their thread optimization isn't as good as Intel, which is why Intel gets better results (plus their endless supply of $ for R and D). No matter what, Intel will not AMD compete again, but what people don't know is that Intel still pays AMD for the 64bit rights. IA64 is specific to Intel but the market prefers AMD's version of 64bit. Right now, AMD needs to stick to the A10 APU's, FX CPU's and bring back the powerhouse Opteron Server CPU's at a MASSIVE discount. They got greedy with their prices (because they outperformed Intel)and that's exactly when they started to Tank. However, now NVIDIA is going to get involved as ARM starts to take over the world.

 

[InvalidError]

PC's on the floor are being replaced by laptops and tablets. It's not a mystery why AMD is struggling.

Intel is hurting too from the longer PC and server replacement cycles, along with people for opting for lower-end computers now that even the low-end is adequate for most people's everyday home and office uses.

As far as tablets go as a growth market, tablet sales growth dropped from around 70% for 2013 to about 12% for 2014, so it seems very likely sales will start dropping in 2015. Tablets are becoming a mature market.

 

[iam2thecrowe]

Seriously the guy that said put them in an apple should think a little bit before responding. Firepro's were in the new mac, and it did nothing. Because no one is gaming on a Apple.

Do you really think that gamers hold such a market share in pc's, that it's the best area for AMD to invest in?

 

[Eximo]

Intel did exactly what they planned. They pushed AMD out of the market by using Intels' endless supply of money and their bad business practices of monopolization and threatening of vendors. I will ALWAYS pick AMD, ESPECIALLY IN VIRTUAL SERVERS because I've had much better results on VM's that run on AMD architecture as opposed to the 4x's more expensive (1100 bucks just for a CPU, come on Intel?!). AMD has the GHz working their favor but their thread optimization isn't as good as Intel, which is why Intel gets better results (plus their endless supply of $ for R and D). No matter what, Intel will not AMD compete again, but what people don't know is that Intel still pays AMD for the 64bit rights. IA64 is specific to Intel but the market prefers AMD's version of 64bit. Right now, AMD needs to stick to the A10 APU's, FX CPU's and bring back the powerhouse Opteron Server CPU's at a MASSIVE discount. They got greedy with their prices (because they outperformed Intel)and that's exactly when they started to Tank. However, now NVIDIA is going to get involved as ARM starts to take over the world.

And AMD still pays Intel for the rights to use x86 technology. AMD64 / x86-64 / x64. IA64 made it to the market last and was only adopted in the server environment. Last I checked only HP was still supporting Itanium based systems, but no new production exists and it is de-supported by Intel.

Not sure I would go so far as to say they have a monopoly, even in the Windows PC world you do have AMD as an alternative, and long ago there was Cyrix, and before that many other choices. VIA is still around, but they make mostly industrial grade x86 chips. But they all pay licensing fees to Intel, and Intel's innovation drove the market to select them since they could deliver the bulk contracts. Pentium name alone was worth a lot during that era. Apple using Intel is a more recent development as well and according to rumors, might change.

If you are talking the computing world in general, there have always been alternatives. Everything that isn't a PC but is still a computer has either ARM or other RISC style chips. FCPGA chips on PLCs and all kinds of niche computing markets.

Then we have smartphones, tablets, and things like Chromebooks (What is the generic term for a laptop shaped object that runs an Android OS?) as alternatives.

It was mentioned earlier, but I agree with the sentiment that Apple should buy out AMD. They could provide the financial muscle for development for something that could compete with Intel. (Only bad thing would be that they would probably make it an Mac exclusive CPU). Not sure how they would try to make the transition to ARM without letting all their software vendors know a long time in advance.

 

[coolitic]

Maybe now they can catch up?

I'm fine with their gpu's (though still prefer nvidia), but when it comes to their cpu's intel is running circles around them.

 

[childofthekorn]

Firepros aren't meant for gaming, its more for design work, static images and mesh type of deals. Its a workhorse in different avenues but doesn't stand up to the dynamic works of video games. Its more of a workstation graphics (autocad, 3d modeling/animation, etc).

It's funny how this reputation sticks to workstation cards. The only differences between Firepro and consumer cards are: higher price, more RAM (with ECC), non-crippled double-precision arithmetic, slightly lower clock speed, and driver optimizations for CAD-type apps.

I doubt the driver optimizations actually hurt game performance much, if any. And while NVidia used a slightly different core for workstations/servers (that was more optimized towards GPU workloads), AMD hasn't been doing that.

See: http://www.tomshardware.com/reviews/firepro-w8100-workstation-graphics-card,3868.html (AMD FirePro W8100 Review: The Professional Radeon R9 290) and check the specs for yourself. If you look at the gaming benchmarks and simply scale the workstation cards' results by the clock speed ratio between the workstation and corresponding consumer cards, you get almost exactly the same results.

This proves that the impact on gaming is simply the 13% and 7% lower clock speeds than their consumer counterparts.

If you get nearly exact same results, have fun with it at the higher cost.

 

[bit_user]

If you get nearly exact same results, have fun with it at the higher cost.

I'm not even sure what's you point, here. Nobody is saying you should buy a workstation card for the purpose of playing games. I was just pointing out that it's no longer true that workstation cards are particularly bad at games.

Which is not to say I consider this point particularly relevant to... anything.

 

[bit_user]

What I can't understand is why they did not enter the tablet/phone arena. The future in computing is in that direction. PC's on the floor are being replaced by laptops and tablets. It's not a mystery why AMD is struggling.

Isn't that why Dirk Meyer got the axe - because the board was unhappy with the lack of a strategy to address this market?

Since then, AMD has been swiftly embracing ARM. For both server and mobile. But it takes time, and I'd guess the mobile space is already so crowded that they've been prioritizing servers, in hopes of establishing a lead, there.

 

[bit_user]

Not sure I would go so far as to say they have a monopoly, even in the Windows PC world you do have AMD as an alternative

In both legal and practical terms, a monopoly needn't be absolute, to be considered a monopoly. I think Intel's dominance over the PC market clearly qualifies as a monopoly, though I'm a bit fuzzy on the exact formal definitions.

I think action by the DoJ is long overdue. I wish they'd make Intel sell off its manufacturing business, or at least put them under independent management and force them to open up to both customers and competitors.

That said, I'm not an Intel hater. The last AMD CPU I owned was a K6. I just think Intel has gotten too powerful and someone needs to level the playing field for the sake of us all.

 

[somebodyspecial]

I've said before AMD was going down, and that management was who was taking them there (not the employees, always single out management). Of course they've laid off 30% of the engineers in the last 3-4 years so no shocker when crap products come from here on out (meaning everything after console has been affected, which we see) with 1/3 the people to do REAL work. I said consoles wouldn't save them, but rather it would detract from CORE products. Here we are. Nvidia passed saying this verbatim, stating winning consoles would detract from their core products (so true). It's not really rocket science to predict this stuff.

I don't think they can recover from consoles with no profits to R&D their way out of this fiasco. They will go down slow (progressively falling behind on gpu like cpu, see maxwell vs. tonga), or be bought. Freesync won't be as good as gsync, they're just hyping it to stop gsync sales for a few more months (arguably their best plan, I'd do the same I guess). Losing share in gpu will only get worse if more gsync monitors hit (coming shortly now) and pricing drops some as freesync comes out and shows why they were NOT testing tons of games at CES. They will continue to get squeezed by ARM-Intel race to the middle, and by NV in gpu who can afford to R&D on multiple shrinks, take some risks etc (lawsuits, gsync, car crap etc) while AMD just can't in cpu/apu or gpu (never mind needing to mount an attack on mobile). It's sad, but AMD has been screwed pretty much since paying 3x the price for ATI. Management has destroyed AMD in a decade, it just took years to see it all play out. The stock will probably see $2 before it sees $4 again if ever. They need to sell now while their IP is still worth something, before they are too far behind NV & Intel. A company with a few billion in cash for R&D after buying AMD could turn the company around quickly just by making sure their products keep coming on the latest process nodes/updated quicker, etc for a while.

 

[bit_user]

Freesync won't be as good as gsync

Why? What makes GSync better?

 

[childofthekorn]

If you get nearly exact same results, have fun with it at the higher cost.

I'm not even sure what's you point, here. Nobody is saying you should buy a workstation card for the purpose of playing games. I was just pointing out that it's no longer true that workstation cards are particularly bad at games.

Which is not to say I consider this point particularly relevant to... anything.

Cost to performance ratio would say otherwise while in context of gaming and comparing the quadro/firepro series to the ones built for gaming.