AMD Starts Shipping First Bulldozer CPU

MrSeb writes "After an awfully long wait, AMD has finally begun shipment of its Bulldozer-based Interlagos (Opteron 6200) server-oriented CPU. If you believe AMD's PR bots, it is the world's first 16-core x86 processor. Unfortunately, and possibly because of reports that AMD is struggling to clock its Bulldozer cores to speeds that are competitive with Intel's Core i7, there's no word of the 8-core desktop-targeted Zambezi CPU. If AMD doesn't move quickly, Intel's Sandy Bridge-E will beat Zambezi to market and AMD will lose any edge that it might have."

Comments

Sandy Bridge-E

[Anonymous Coward]

If AMD doesn't watch out their mainline $200 processor will be made obsolete by Intel's $1000 EXTREME CPUs!

Re:Sandy Bridge-E

[CajunArson]

You may laugh but think of it this way... if that $1000 Gulftown CPU from March of 2010 can still beat an 8 core Bulldozer that comes out 19 months later, then you would only have to realize a marginal benefit of about $1.75 per day to make it economically worth your while to have bought the "overpriced" Gulftown chip. (that includes the cost from Intel motherboards that tend to be more expensive and the extra RAM for a triple-channel configuration). Nevermind the fact that 6-core chips have been sold for $600 for some time as well. I can think of a bunch of professional applications that can easily show a $1.75 / day benefit from the extra cores. Maybe not for playing games, but for a lot of real applications.

  Bulldozer should beat the consumer-level SB chips at perfectly threaded integer benchmarks, but it remains a very open question if it will be able to beat the almost 2 year old Gulftowns at the same tasks, and it is an almost foregone conclusion that it won't beat the 6 core SB-E chips at those tasks. Factor into account the 315 mm^2 die size of EVERY Bulldozer (not just the 8 core ones, but the cheap 4 core ones too since AMD only has 1 die design) and the immaturity of AMD's 32nm process and things could be expensive for AMD on the desktop. That's why it makes sense to ship the server chips first where AMD has some hope of getting higher ASPs.

Re:Sandy Bridge-E

[aliquis]

If AMD doesn't move quickly, Intel's Sandy Bridge-E will beat Zambezi to market and AMD will lose any edge that it might have.

Ok, so nothing will be lost?

(No, I'm not trolling or flaming, personally I would get a new machine to play Starcraft II, AFAIK SCII only seem to use two cores. AMD themselves has claimed their chip (best consumer chip?) would be similar to 2600k in performance. 2600k is quad core vs octo core for the AMD chip. Considering the work load the 2600k will still outperform the AMD by a lot. Also with socket 2011 we talk quad channel memory instead of dual channel and even back in 1156 vs 1366 days and with SCII only using two cores the 1366 chip beat the 1156 chip. So obviously memory bandwidth was an issue even at two cores. So I'm just stating the obvious. Intel is the leader and will remain the leader for the time being.)

Re:Sandy Bridge-E

[Anonymous Coward]

I've (honestly) just been asked what our expected budget requirements are for hardware for the next year. Please inform me where I can go to use the patented Intel time travelling technology so that I can retroactively use things before I decide to purchase them.

Re:Sandy Bridge-E

[rbarreira]

There's no evidence to indicate that AMD's "mainline" $200 CPU will be much better than the existing "mainline" $200 2500K that's out right now

There is some, depending on your application of course. If computer chess analysis is your thing, you would see benchmarks results like these, where the $189 Phenom II X6 1100T beats the $219 Intel 2500k.

So AMD already has CPUs which are price-performance competitive, surely Bulldozer shouldn't be worse in terms of price-performance.

Re:Sandy Bridge-E

[sgt scrub]

the immaturity of AMD's 32nm process and things could be expensive for AMD on the desktop.

That is true, and as you point out, for the desktop. Machines in a data center are cooled so the number of cores is a better measure of functionality. If you build machines that run multiple VM's, which is usually the case, that cheaper 6200 will not only outperform Intel's Gulftown and more likely be preferred when adding more machines to the data center even over the SB-E chips. If AMD can get a better footing in the "cloud" infrastructure they might make enough to move to a die size smaller than 32nm, which is REALLY what they must to do.

Re:Sandy Bridge-E

[bhcompy]

That $200 is still a fuzzy number given the motherboard prices. You can get a very good AMD chipset(880/890) for $100 and have all the latest features(USB3, SATA3, etc) while being forward compatible for quite a while(manufacturer dependent, but Bulldozer(AM3+) is compatible with AM3 chipsets with BIOS updates). With Intel, you're still paying more for the equivalent and next year you'll need to get another motherboard to upgrade that processor because of constant socket changes.

Re:Sandy Bridge-E

[V!NCENT]

Talk to a server farm when Intel is putting money in the development of Coreboot and we'll talk.

For now AMD is superior, because a server reboot requires about 1/100th of the time that it takes an Intel CPU farm to get back up due to horrible BIOSes. The more motherboards you have, the longer it takes due to serialized bootup. Ouch... Massive ouch...

Downtime versus marginable CPU speed... And less cores...

Re:Sandy Bridge-E

[rbarreira]

On what applications? Post benchmarks...

Re:Sandy Bridge-E

[serviscope_minor]

It is a negative feedback loop for them - by failing to create competitive technology

You missed the bit about how they create competitive technology and never get any customers because intel has a huge monopoly.

Also, for cramming flops into U's the quad socket AMD 6100s beat out anything Intel has to offer in the same segment.

Re:Sandy Bridge-E

[beelsebob]

And in the vast majority of tasks the X6 is slower, including some very multithreaded ones:

http://www.anandtech.com/bench/Product/203?vs=363

Re:Sandy Bridge-E

[Sloppy]

Is the 6200 16 physical cores or just 8 dualthreaded ones?

Part of what makes Bulldozer interesting, is that there is no quick straight-forward soundbite answer to that question. It's 16 less-than-complete cores, or 8 more-than-merely-dual-threaded cores.

I look forward to the day when some profiler nerd figures out that most cost-effective ratio is to make a chip with 3 memory busses, 17 integer units and 11 floating point units, with 23 sets of registers. /proc/cpuinfo will say it's a 23-core machine so that'll be the soundbite answer, and then some math twit will whine that it's really only an 11 core machine. And everyone will be both wrong and right simultaneously.

Re:Sandy Bridge-E

[Anthony Mouse]

The real problem for AMD is that they didn't expect Intel to turn it around so quickly with Conroe. They had an excellent design and they expected to continue extracting reasonable margins from it in order to fund Fusion development, but when Core 2 came around they lost that.

The thing is, Bulldozer is a great direction for them. It will not beat Intel's best at single-thread performance, but it isn't supposed to. What it's supposed to do is offer better performance per watt and per rack unit for common server workloads, which it very much does. And that gets them back into the server room with reasonable margins and keeps AMD alive to continue developing new products.

The real interesting thing is that I'm not sure Intel has a good way to respond to it. SB is designed to be the fastest processor for single threaded performance, and it does that. The trouble is that the server rack is where the high margins are and most of those people don't care about single thread performance, they care about the things that AMD has designed for. So Intel ends up with a bad choice: Choice one is that they can try to adapt SB to be more like Bulldozer, i.e. SMT with more execution units so that you can get nearly two cores worth of performance out of substantially less than two cores worth of silicon and power budget, with the consequent hit to single threaded performance. Or choice two is to continue developing SB on its existing goals for the desktop and for the relatively small number of non-threadable server applications and then dump a billion dollars into developing an entirely new processor to meet Bulldozer for the lucrative parallel server market.

Re:Sandy Bridge-E

[serviscope_minor]

Except that all came about when AMD came around with the Athlon 64's,

Well yes, and that really wrecked AMDs revenue. The were number on on terms of absolute performance and price/performance, but a distant two in sales.

The massive market fixing done by Intel meant that AMD was unable to put nearly as much as they could have done to develop future processors. Perhaps as a result (we will never know), Intel was able to catch up. It takes several years to develop a new microprocessor architecture.

That was also not the first time that Intel had to pay AMD because of illegal market fixing.

Re:Sandy Bridge-E

[kimvette]

Boot time vs. run time is what? 1:500000?

You obviously don't run Windows, or at least don't install the patches. ;)

Re:Sandy Bridge-E

[Chris Burke]

Well yes, and that really wrecked AMDs revenue. The were number on on terms of absolute performance and price/performance, but a distant two in sales.

It did more than that. It wrecked their entire financial structure.

For a brief period, AMD's market share was fab limited. So AMD built a huge new fab, acquiring several billion dollars in debt in the process. No big deal, given the possible revenue the extra capacity could supply. Fabs are expensive. It's part of the game.

Once the fab came online, though, AMD's market share barely moved. Their market share was now Intel-limited. This is bad, because while their fabs were formerly full and thus efficient and profitable, now they were largely empty. An underutilized fab is a money sink fab, above and beyond the debt from building it and the depreciation of the equipment.

This more or less directly led to them spinning off their manufacturing division as Global Foundries. Their debt structure no longer matched their prospects, and it was going to be very expensive to finance the next fab which they would have to build to stay in the game.

There were certainly missteps by AMD, too. They did underestimate how quickly Intel would adapt. They had been developing their own from-scratch high-frequency K9 architecture, but it was canceled late in development. Instead they just tweaked K8 and doubled the cores to four to create Phenom, which had big enough problems that when rev C fixed them they re-launched the brand as Phenom II.

It is hard to say what would have happened had Intel not engaged in market manipulation. One thing I think can be said -- Intel, too, has had their missteps, but they're less damaging because they have the financial stability to absorb them. Maybe AMD would have made the same mistakes, and still been behind Intel just as much, but if they'd been able to attain the the market share they could have, they would have had the money to ride out the problems, and been able to keep their fabs, and a lot of engineers that were laid off.

The price/performance ratio.

[Chrisq]

What will be interesting is the price/performance ratio compared to the Intel chips. This chip will be typically used in server farms, and this will be at least as important as the raw power - though obviously there is an overhead in running more servers. AMD has usually been ahead of Intel, and it still is on most mid-range and low-end chips, but it has started to fall behind at the high end.

What is the FPU performance of these things...?

[serviscope_minor]

Does anyone know the FPU performance of these things?

So comparing a 16 "core" 'dozer to a 12 core magny-cours:

The number of parallel integer (and memory addressing) threads has gone up from 12 to 16.

The number of FPUs has dropped from 12 to 8.

The new FPUs are now twice as wide with the AVX instructions.

So, two threads share one wider FPU now. If it's hard keeping an FPU full, then this should make better use of the hardware. It seems that if your code does well for parallel, scalar FPU work already, then there may be a performance drop.

If you have trouble filling the FPU for scalar work, then this should give better utilisation of less hardware. There's a possible performance increase if your utilisation is currently under 67%. Since the two core units can feed the FPU independently, there is a little latency hiding now. This could help even if there are two completely independent processes using the FPU at the same time.

I suppsose the reasoning is that there is often fine-grained parallelism to be had, and the problem of fine grained parallelism and keeping the FPU full are often independent. So AVX will improve performance there.

So, it seems that the peak FPU performance has increased in the ratio of 16/12.

The actual performance could be all over the place. It will be interesting to see.

The other thing is that these are now single chips with 8 bulldozer units on and 16ish cores. Perhaps AMD will go and make more MCMs like before, giving 32ish cores per socket :)

It is all about the die size

[sgt scrub]

Unfortunately, and possibly because of reports that AMD is struggling to clock its Bulldozer cores to speeds that are competitive with Intel's Core i7, there's no word of the 8-core desktop-targeted Zambezi CPU.

If you increase the clock on the CPU you have to cool it. Reducing the die reduces the amount of cooling that needs to be done. AMD is not able to shrink their die. Yet.

Re:It is all about the die size

[Anonymous Coward]

Yeah, Intel's vast capital reserves mean they typically have a generational lead in process tech, and get the increased efficiency / decreased temps of a die shrink "for free". AMD have to out-innovate them just to produce an equivalent CPU, let alone a better one. Unfair, but that's how (near-monopolistic) business works I guess.

Re:It is all about the die size

[shugah]

Well it certainly helps GlobalFoundry to have more than one client. Apparently GF is still scaling up (and down) it's sub 32 nm processes and facilities and will be able to give TSMC some serious competition soon. AMD can only benefit from this competition.

For the future, I think This is what we are going to see from AMD. An integrated, heterogeneous MPSoC architecture with open (or at least standardized) on chip interfaces that might allow a mix and match CPU Core, GPU, PIO and Memory. Sort of taking Fusion one step further.

Is high performance really an issue?

[gman003]

I've never really considered AMD the manufacturer to look towards when looking for high-performance stuff. In my mind, at least, they're the "dirt cheap and good enough" side - I bought a triple-core Phenom for about the price of a low-end Core 2 Duo a year or two back. They've always had the best performance per dollar. Sometimes, yeah, they did even have the best absolute performance, but Intel's back in the lead again.

High performance just isn't a very profitable market segment. Gamers and high-end servers will buy it, but that's not where the big market is. The big market is desktops and laptops - stuff where a 4gHz sextuple-core processor is overkill. A business machine will work fine with half that - and with AMD's price advantage, they've been moving in on business and desktops. Supercomputers might also be enough to sustain the company - they buy by the thousands, and AMD's power efficiency and multi-core design has usually been attractive to the few in that business. There, performance per core isn't nearly as important as cores per watt.

That said, I'm not surprised that AMD is (supposedly) having issues meeting their targeted clock rates. Pre-release info pegged the top desktop processor at 4.2gHz - a record for an x86 processor. The last to get close to that was the last few Pentium IV HTs at 3.8gHz. AMD's top processor to date only reached 3.7ghz (Phenom II X4 980BE), and that was after years of refining their process. AMD set their sights too high, and is having problems for it.