AMD: What happened?

[sequoia464]

Bulldozer wasn't just a slight miss-step, it was poorly thought out from both the architectural standpoint as well as the execution. Opting to go with a gate-first approach on an architecture requiring over 4.5ghz stock clocks on a brand new 32nm fab process and an ugly amount of slow cache doesn't point to just a GloFo problem, it actually points directly to the engineers. There's good reason people were saying the sky was falling, and it's because they're very very far behind as far as IPC goes.

What, if anything, will or can AMD do at this polnt to improve the architecture?

 

[Idontcare]

What, if anything, will or can AMD do at this polnt to improve the architecture?

On a piece of paper, write down everything that is different between bulldozer and Sandy Bridge. They can start with that.

 

[Obsoleet]

Ahh, the old "AMD is good enough" argument. This was probably true at one time, as they offered good performance for the price at the low end. I dont think even that is true now, as a Sandy Bridge Pentium can compete (at least in gaming) with a AMD quad core.

Shhhh! You can't say things like that. Remember how the P4 was so slow it couldn't even be used, and that a single P4 build could heat an entire factory? Now we don't need 'fancy Intel quad-cores' or even fast Intel dual-cores when a PhII will do just fine.

Not the same at all...

And you two would be wrong. Up until a year ago, a family friend was using a P4 Willamette. It browsed the web just fine.
I kept it going on 640MB RAM and a well-maintained, GPU accelerated, unbloated system. Ages after the 'crime' that was Willamette were long over.

Considering BD is comparatively better, or at minimum equal to the P4 at the time, I'd say it's fast enough in the age of overpowered CPUs and underpowered GPUs (in APUs).

It's not an argument, don't be a tool. It's honesty. I don't own one, wouldn't for a long time, but the chip really is pretty fast. If you stare at benchmarks all day I guess not, but I bet the guys who are using one aren't wishing they had a faster CPU. It's faster than a lot of the CPUs used by people on this forum, including myself.
Lets get some sense. We need a few less i7 big-rig spenders, and a few more real techs at Anandtech who can screw their head on straight.

 

[DominionSeraph]

Still I have to give kudos to Intel. Despite no serious competition from AMD, they continue to crank out steal after steal like quad core 3+ GHz CPUs for under $200! Unreal. Unlike AMD, Intel learned their lesson with PIII/P4. Rather than sitting idle at the top taking a break while well in the lead, they continue to innovate and push faster and better chips and better the industry even when nobody asked for them and without much serious competition forcing them to do anything at all. We could easily still be sitting at Core 2 Duos/Quads for $900.

And exactly how could Intel sell C2D's for $900 when the Phenom II 555 sells for $80, or C2Q's for $900 when the 955 sells for $110?

And if Intel loves you so much, why don't they sell you the i7 2700k for the price of an i5 2300? Or the i7 3960X for the price of a 3930k?

Face it, Intel's pricing is still dictated by AMD. They can't raise the price of their quads without leaving a large gap in the budget range. If Intel raised the price of all their quads by $100 sales of the Phenom II X6 and Bulldozer 8-cores would go through the roof. So Intel is stuck keeping the tail of its quads down in AMD range, which limits them to reasonable increases in the immediate 10-20% performance range above that (because you're not gonna find many who will pay 100% more for a 10% boost), which means they can only go crazy with their halo product.
Without competition I'd expect them to really start filling out the $400-$2000 range in the next few generations.

 

[IntelUser2000]

Without competition I'd expect them to really start filling out the $400-$2000 range in the next few generations.

CPUs have generally become cheaper over time as manufacturing costs go down and volumes shipped go up. It's also unlikely that will happen because many factors make it different from say, 15 years ago. Back then PC markets grew so fast, they could cannibalize their own products and it wouldn't matter as much. They also bet on lower prices over time on increased PC penetration. That won't happen with $400 processors as low end.

 

[DominionSeraph]

That won't happen with $400 processors as low end.

I said "filling out the $400-$2000 range", not, "ignoring the <$400 range." And what is this "15 years ago" nonsense? Price gouging has happened whenever one side or the other has had a substantial performance lead. The Athlon 64 FX's weren't $100 chips.

 

[frozentundra123456]

And you two would be wrong. Up until a year ago, a family friend was using a P4 Willamette. It browsed the web just fine.
I kept it going on 640MB RAM and a well-maintained, GPU accelerated, unbloated system. Ages after the 'crime' that was Willamette were long over.

Considering BD is comparatively better, or at minimum equal to the P4 at the time, I'd say it's fast enough in the age of overpowered CPUs and underpowered GPUs (in APUs).

It's not an argument, don't be a tool. It's honesty. I don't own one, wouldn't for a long time, but the chip really is pretty fast. If you stare at benchmarks all day I guess not, but I bet the guys who are using one aren't wishing they had a faster CPU. It's faster than a lot of the CPUs used by people on this forum, including myself.
Lets get some sense. We need a few less i7 big-rig spenders, and a few more real techs at Anandtech who can screw their head on straight.

You are totally missing (or just ignoring) the point. I never said AMD wasnt "good enough" in most cases. The point is that Intel offers better performance for the dollar at every price point now. It makes no sense to me to pay more for the same performance.

 

[frozentundra123456]

And exactly how could Intel sell C2D's for $900 when the Phenom II 555 sells for $80, or C2Q's for $900 when the 955 sells for $110?

And if Intel loves you so much, why don't they sell you the i7 2700k for the price of an i5 2300? Or the i7 3960X for the price of a 3930k?

Face it, Intel's pricing is still dictated by AMD. They can't raise the price of their quads without leaving a large gap in the budget range. If Intel raised the price of all their quads by $100 sales of the Phenom II X6 and Bulldozer 8-cores would go through the roof. So Intel is stuck keeping the tail of its quads down in AMD range, which limits them to reasonable increases in the immediate 10-20% performance range above that (because you're not gonna find many who will pay 100% more for a 10% boost), which means they can only go crazy with their halo product.
Without competition I'd expect them to really start filling out the $400-$2000 range in the next few generations.

AMD wasnt above gouging nearly 1000 dollars per chip when they had the performance lead. And you are picking very isolated examples when you compare 1000 dollar Intel chips to sub 100.00 AMD chips. And the 2500K which can be had for equal or less than Bulldozer mops the floor with it in nearly every application. You could actually argue that Bulldozer is overpriced now because it is "8 cores", and some people dont realize a 4 core intel chip beats it in 90% of applications.

And the only way Intel is dictated anything by AMD is because of anti-trust considerations. With their volume and cash on hand, Intel could easily drive AMD out of the CPU business if they wanted and the government would not intervene. Actually, they almost have anyway, because by their own admission, AMD is no longer competing directly with intel in the high end (and I would say in the mid range either).

 

[Red Storm]

They're trying to compete against a company that has so much more money, R&D, and production advantages, they can't afford to make any mistakes (and they have).

 

[PingSpike]

AMD did the same thing Intel did that allowed AMD tp stomp a mud hole in Intel with beating them to the 1 GHz race with the TBird and then dropping the Athlon 64 FX bomb. They got over confident and sat on their asses and tried to stretch out, rehash, repackage, and resell old technology over and over again as new products thinking they had the cat in the bag while the competition was busy redesigning from ground up. Once you're on the bottom it's hard to recover due to lack of income for R&D for said ground up redesign, then it's only about survival.

That's what I saw. Intel stumbled and AMD pounced on them, then promptly forgot all about that and went on to make the exact same mistake Intel did pretty much immediately after. I have to blame management honestly. You can blame the engineers but it was a pretty long time line and didn't their engineers also put out the Athlon, XP, A64 and X2? Those were all solid to game changing products produced by a company that was financially way weaker than Intel. If you have the wrong people on it or they're going in the wrong direction its up to management to rectify that.

I can understand why they would want to stretch out and suck as much juice out of their investment as possible. AMD has seemingly been perpetually in debt and constantly playing catchup. But that just plain wasn't an option. Intel could afford to learn from its own mistakes, AMD couldn't. They had to learn from Intel's one. And from where I'm standing, they didn't.

People talking about the pricing by Intel are sort of right and wrong I think. AMD was not above a bit of gouging when they had a clear top position. I recall the lowest end Athlon X2s on socket 939 refusing to go below $200 for a pretty damn long time. I don't even blame them much, they had the best product and weren't they up against Pentium D in those days? They could command a premium because there just wasn't an alternative.

I don't really think Intel is being as gracious as you might think with their pricing though. Part of me thinks the whole thing is a fuck you to AMD but I bet there are two less emotionally charged reasons for their new friendlier pricing strategy: The market has changed and they want to win the long war in graphics.

They can't sell $500 chips these days like they used too. Performance needs for most users just will not demand that price. I like to game and I have to admit that my last few CPU and GPU upgrades have essentially been pointless. I didn't need the performance but bought in because it was cheap. For some I'm sure they really need it but its probably harder to sell to most. If the 2500K I rock was $500 I never even would have considered it. Hell, at $200 I waffled a lot. Imagine the OEMs who are selling $500 laptops and good enough email boxes all day.

And its no secret that Intel isn't a juggernaut in graphics. Their integrated has gotten better, that's true. But its still nothing impressive and there is still no powerhouse solution, integrated or discrete. They're behind on performance, drivers and clout. Their game changing larrabee probably turned into an internal revelation about how far behind they really are. So they're using the tools they have to catch up. With ATIs fates attached to AMDs, they'll club them in the CPU space to keep the company broke and hope they run out of cash for R&D. With Nvidia they'll leverage the fact the company will not tie itself to AMD if they scorn them (since they're their main competitor) and close their own gates and hope they die on the vine. Maybe they'll buy Nvidia later when they're cheap. Maybe they'll develop their own and try to catch up, but for now they can slow and weaken any competitors in that space to create an opportunity later.

 

[formulav8]

On a piece of paper, write down everything that is different between bulldozer and Sandy Bridge. They can start with that.

I hope your not serious

 

[Red Storm]

Intel doesn't want AMD to go away. They want them alive and kicking, but just barely.

 

[pelov]

What, if anything, will or can AMD do at this polnt to improve the architecture?

The way I see it is that AMD made a judgment call 4-5 years ago and assumed that software progression would keep up with and even out-pace an ever-upward-scaling core count in the computing world that would try to feed the software that favored moar coars over IPC. It was a piss-poor call. Somewhere along the way somebody should have said "Hey, wait a minute. Maybe we were overly optimistic in our assessment." Instead, they stuck with it and the delays were both due to GloFo as well as an attempt to cover up the poor performance -- I think that's why it's got so much damn cache and was about 2 years too late.

I think they can certainly do quite a bit to patch it up, but what it's going to require and the amount of time that's needed is another story. They wanted to hold IPC the same with Bulldozer but missed that mark by -10%. In certain cases, a 2500K beats an 8150 by 40-50% in single-threaded performance. That's absolutely atrocious and they quite simply don't have a chance in hell of closing that gap with Piledriver nor Steamroller. I'm guessing they'll try their best by addressing the uberslow cache speeds, bump up stock clocks and that will allow Piledriver to do well in highly-threaded workloads and get close to or even slightly beat the 2600K (remember, that was the whole intent of CMT anyway) but will still display its weaknesses in low-threaded applications but by a smaller margin

What will be really interesting is to see if there was something that was missed and can be improved within the fetch/decoder within a module so that -20% performance hit within-module can be mitigated. I think any changes for the better on that end will help them tremendously as far as highly-threaded workloads go. Basically, attempt to make CMT look less like CMT as far as benchmarks go but still advantageous on the fab and $$ side.

edit:

Just want to add that if you see Piledriver stock clock speeds increased drastically (and judging by the Trinity APUs that does look to be the case) I'd be extremely wary of any significant IPC increases. What they did was lengthen the pipeline and increase the latencies (the cache being the biggie here) so the high clock speeds was something that Bulldozer required in order to make up for that intended IPC halt. Higher clock speeds at lower TDP likely means the cache speeds haven't been addressed and instead they're ramping up the clock speeds on a more refined process in order to make up some ground in per-core performance. Those would be the bad signs.

Unfortunately it seems that's exactly what AMD did. I ran the numbers on Trinity's clock speed in comparison to Llano (both 65W TDP variants and not including Turbo) and AMD's estimates pointed to a ~20-30% increase in PCmark Vantage, which roughly equated to the increase in clock speed from Llano to Trinity. In short, if we're to assume these xbit graphs and AMD performance estimates are accurate, we'd be looking at Trinity equaling Llano's IPC and that performance improvement would be a direct result of higher clock speeds. That's a good sign even though I just stated it's a bad sign (!) and in essence should have been what Bulldozer brought to the table because this was the initial plan for Bulldozer from the get-go, but it also paints a poor picture going forward because they'll HAVE to change their approach from clock speeds to IPC increases due to their slow progression and limitations on the fab front. It looks as if they took the easy road and just bumped up clock speeds with a minor IPC increase to overcome Bulldozer's initial shortcomings. It looks as if it's a step in the right direction for the short term, but long term they'll require a complete overhaul in approach.

The good sign (better sign, imo) would be a leveling off of clock speeds on a smaller node (not happening until 2013 at the earliest) and a decrease in TDP on CPUs rather than APUs. It's a bit more difficult to judge the APUs because of how strong AMD's graphical performance is and just how it would affect the entire chip. There's also the issue of whether we'll even see any more AMD CPUs for the desktop after Vishera. According to their slides on FA day they're planning to replace them all with APUs going forward.

adding even more to this
http://www.ilsistemista.net/index.p...omparison-whats-wrong-with-amd-bulldozer.html
Not very good grammar (seems to be foreign), but the dude does a good job of gathering info from many sources and puts it together really well.

 

[beginner99]

What the question really oughta be, is how pathetic it is that intel with all their resources, 10x the size of it's competitors, proven illegal tactics and cohersion with the defacto software provider of 95% of worldwide PC's and largest OEM's, taken 25 years and still haven't managed to outperform the competition by more than 10-20% in the traditional CPU space.

10-20% by using half as much power and with double the margin that AMD has. Anyway as long as Intel does not need to be better, the surplus of money will flow into the shareholders pockets and not into research...

 

[DominionSeraph]

AMD wasnt above gouging nearly 1000 dollars per chip when they had the performance lead. And you are picking very isolated examples when you compare 1000 dollar Intel chips to sub 100.00 AMD chips. And the 2500K which can be had for equal or less than Bulldozer mops the floor with it in nearly every application. You could actually argue that Bulldozer is overpriced now because it is "8 cores", and some people dont realize a 4 core intel chip beats it in 90% of applications.

And the only way Intel is dictated anything by AMD is because of anti-trust considerations. With their volume and cash on hand, Intel could easily drive AMD out of the CPU business if they wanted and the government would not intervene. Actually, they almost have anyway, because by their own admission, AMD is no longer competing directly with intel in the high end (and I would say in the mid range either).

Are you just stringing random words together? Because that's just a mess.
<---- Avatar related.

 

[Makaveli]

I don't think either Pentium 4's Hyperthreading or Bulldozer's Multi-threading was bad at all. It was hobbled by the overall architecture.

Pentium 4 was actually quite narrow, something that doesn't work well with SMT. It's hard to say something conclusive about Bulldozer, but it also went the narrow route. It seems the idea of taking out execution units for a minimal performance loss(due to low utilization rates) don't work so well in practice.

From what I read this was the reason. P4's front end was too narrow to gain anything from HT.

Which they corrected in Nehalem.

 

[Makaveli]

Can Intel/AMD quickly turn things around on a dime? I doubt it.
Intel is still coasting along very well on its Core architecture which debuted in 2006. It had bad years previously with its P4 line 2000-2008. The Athlon 64 X2's had their genesis in the Athlon, tweaked into the Athlon XP, then more tweaking into the Athlon 64's.

If I remember correctly the A64 didn't come until later and the P4 was competing fine with the Athlon XP. So I don't follow your 2000-2008 bad years quote at all.

And now that I think about it during those "bad years" Intel was still making more money than AMD even when getting raped by the a64.

 

[exdeath]

AMD wasnt above gouging nearly 1000 dollars per chip when they had the performance lead.

I paid around that for my FX-55.

 

[Makaveli]

I paid around that for my FX-55.

I don't know how you guys do it.

I have the money to burn but would never spend $1000 on a chip.

Been in this industry for far too long nothing ever stays on top long enough for you to even brag about it.

For me the $300 chip that I can overclock the shit out of to get 90% of the performance of that top end chip is a far better deal.

I guess some of us do crazy thinks like a guy I saw with 7970 Xfire for 1920x1080 gaming lmao.

 

[frozentundra123456]

Are you just stringing random words together? Because that's just a mess.
<---- Avatar related.

So now you are reduced to criticizing my sentence structure instead of making logical arguments that are on topic?

 

[frozentundra123456]

I don't know how you guys do it.

I have the money to burn but would never spend $1000 on a chip.

Been in this industry for far too long nothing ever stays on top long enough for you to even brag about it.

For me the $300 chip that I can overclock the shit out of to get 90% of the performance of that top end chip is a far better deal.

I guess some of us do crazy thinks like a guy I saw with 7970 Xfire for 1920x1080 gaming lmao.

Personally, I agree with you, but on the other hand one can spend several thousand dollars on a vacation, or spend 40k on a car that is far above what anyone needs. So in that context, maybe it is not so outlandish.

 

[Obsoleet]

You are totally missing (or just ignoring) the point. I never said AMD wasnt "good enough" in most cases. The point is that Intel offers better performance for the dollar at every price point now. It makes no sense to me to pay more for the same performance.

I knew what you said, but you also prefaced what you said by mocking my statement that "AMD is good enough". I knew we were on a slippery fanboy slope in the thread with the idea being tossed around (not specifically by you) that BD as a CPU is NOT 'good enough', or it's 'slow'. So I wanted to make that crystal clear, and I am glad we agree that Bulldozer is indeed a fast enough CPU in this era for most purposes.

The P4 era was a different ballgame, advancing CPU power then was a big deal, today, not so much. No ones going to turn down a faster CPU, but most are working on a better APU and to further integration.
Yet, I've even gotten those horrid P4s to work into the modern era. A notorious CPU that similar, yet opposite aisle fanboys attacked relentlessly.

There are clowns on both sides of the CPU aisle, it just so happens people are getting a little out of hand in their anti-BD zealotry. Tomorrow, it will be a blazing fast Intel CPU that didn't live up to expectations.

If you want the real scoop on the real story, the market letdown in the market has been Intel's integrated GPU performance, holding back a lot of progress for years now.
My wish is that IvyBridge and Haswell's performance as an APU are finally acceptable, because it certainly is not 'good enough' on my i7-640M. Bulldozer failed expectations considerably but it does do the job- this sorry excuse for an integrated GPU is simply unacceptable for use. I'd gladly give up some CPU power for more GPU.

 

[blackened23]

Buying ATI was a really smart move. Unfortunately after all this time they really have nothing to show for it. Who knows what's really going on in their labs as far as true CPU/GPU integration, but from what I am seeing on the market they are nowhere close to realizing their potential. I think someone like apple would have gobbled them up if they actually truly fused a gpu with a cpu core.

Buying ATI was smart. AMD however, paid *far* too much for ATI. They've been treading in mud ever since (cpu division)

 

[386DX]

If I remember correctly the A64 didn't come until later and the P4 was competing fine with the Athlon XP. So I don't follow your 2000-2008 bad years quote at all.

And now that I think about it during those "bad years" Intel was still making more money than AMD even when getting raped by the a64.

The P4 years wasn't as bad for Intel as people make it out to be. Sure in hindsight the architecture was a failure but at the time the CPU itself was never totally outclassed by the A64. The much higher speed of the P4 compared to the A64 allowed it to at least trade wins. P4 was stronger at FPU while A64 was stronger at Integer, so there were situations where one would choose a P4 over an A64 at the time.

Today the Bulldozer situation is much worst imo. There is literally no argument you can make for getting a Bulldozer CPU. Intel has a CPU that can beat Bulldozer at any single task you can think of... and often times for less money.

While some people would say Intel is gouging there customers, you really have to look at things in perspective. Excluding the super high end CPUs the value you are getting now for sub $320 Intel CPU's is fantastic compared to your choices back in the P4/A64 days. When you take into consideration inflation the value is even more so. I personally don't believe Intel gouges there customers. If you look at there margins on CPU's it been pretty much the same throughout there history, as process improves, and chips get cheaper to make there prices also go down. Sure there's still some $1000+ CPU's available but those aren't the norm and those are the ones that offset the lower margins of the sub $100 CPU's. If you look at AMD on the other hand they are the ones that have changed since the A64 days, there margins have gone way down on there CPU's once they tried to start a price war with Intel.

From a business perspective Intel is a much smarter company. Sure AMD will sell you a quad core CPU for $100 and make virtually nothing on it, Intel will just say go ahead you can have that market segment... if we can't put out a CPU at that price point with the margins we want we won't put it out.

 

[Fox5]

I knew what you said, but you also prefaced what you said by mocking my statement that "AMD is good enough". I knew we were on a slippery fanboy slope in the thread with the idea being tossed around (not specifically by you) that BD as a CPU is NOT 'good enough', or it's 'slow'. So I wanted to make that crystal clear, and I am glad we agree that Bulldozer is indeed a fast enough CPU in this era for most purposes.

The P4 era was a different ballgame, advancing CPU power then was a big deal, today, not so much. No ones going to turn down a faster CPU, but most are working on a better APU and to further integration.
Yet, I've even gotten those horrid P4s to work into the modern era. A notorious CPU that similar, yet opposite aisle fanboys attacked relentlessly.

There are clowns on both sides of the CPU aisle, it just so happens people are getting a little out of hand in their anti-BD zealotry. Tomorrow, it will be a blazing fast Intel CPU that didn't live up to expectations.

If you want the real scoop on the real story, the market letdown in the market has been Intel's integrated GPU performance, holding back a lot of progress for years now.
My wish is that IvyBridge and Haswell's performance as an APU are finally acceptable, because it certainly is not 'good enough' on my i7-640M. Bulldozer failed expectations considerably but it does do the job- this sorry excuse for an integrated GPU is simply unacceptable for use. I'd gladly give up some CPU power for more GPU.

Pentium 4 level performance is good enough... when you have enough dedicated acceleration hardware to go along with it. ARM cores and Atom cores are in the realm of the single threaded performance of the initial Pentium 4's (and as dual core, can probably hang with the middle of the line p4's), but having dedicated hardware for video, 3d, audio, etc, makes them far more viable than they would be otherwise. It's kind of weird, cpus got more powerful in the PC market, so no one wanted to pay for the dedicated add-in hardware. Vendors that tried to do all in one add-in cards for graphics + everything else quickly found themselves falling behind in all areas, yet in turns out this is really what the market wanted all along.