Tech Link (Processor): Clovertown versus Barcelona

RickGeek of Geek.Com posted an interesting opinion about native versus mcm. In his post, the author cited many advantaged/disavantages of Clovertown MCM versus the competition's so called offering. And the truth isn't very hard to see, Intel has been shipping millions of quad core processors while AMD has yet to ship one. And I don't mean sample, Intel is shipping retail products, just in case you're not aware yet, and AMD is still delaying their Barcelona release. This processor better perform or it will be a major dissapointment even for a FanBoy like me...

OPINION


When I first heard about AMD's decision to make a native quad-core processor--that is a single piece of silicon comprising all four cores--I thought it was an excellent idea. I began to think of the advantages four-way HyperTransport has over the shared bus architecture of Intel's design. I also saw the significance in the larger shared cache, especially when all four cores are working closely on some particular task, and I didn't really see any downside. However, the more I think about it, the more I am realizing how much better Intel's solution is.

Clovertown is a single-package solution comprising two dual-core Core 2 processors within. It uses a shared bus architecture and is somewhat slower in off-chip communication, but because the die is so much smaller, they can create a larger shared cache for each dual-core, thereby making up for much of that loss.

AMD is not at the same place in manufacturing prowess that Intel is. It has been reported that Intel's yields are also very high, typically 90% or higher with some reports saying 95% or higher. AMD's yields have never (to my knowledge) been publicly disclosed, though I'm going to make a probably very incorrect assumption that they're around 80% for current processors. Even assuming they were 95% for current processors, we know that the number of potential defects goes up by the square of the surface area, so as we move from K8's current 183 mm^2 dual-core 65 nm 512 KB L2 processors, we would see a significant degredation in potential yield due to an additional 100 mm^2 per die alone. That also does not include the potential loss of high-binning parts due to even small defects not making the chip unusable, but rather making them less clockable at higher speeds.

We now know that Barcelona's die will be 283 mm^2, and that Clovertowns are comprised of two 143 mm^2 dual-core dies. Using my Wafer program, we find that AMD could fit a maximum of about 212 Barcelonas per 300 mm wafer, while Intel could fit about 430 dual-core Core 2 processors on a single 300 mm wafer. Now, assuming a 90% yield for Intel, that's about 388 dies per wafer, or 194 Clovertowns. In order for AMD to make as many Barcelonas per 300 mm wafer, its yield would have to be 92%, and by all accounts that's not something AMD will be able to do 1) based on its unconfirmed history of lower yields, 2) based on the natural increase in defect rate due to the larger die size and 3) that defects of any kind which do not render the chip inoperable will render it operable only at slower speeds.

AMD is facing several realities here with its native quad-core design, and it would seem that it does not have the manufacturing prowess to pull it off. Some at Intel even admitted recently that the company would have a hard time doing a die that large with high yields. Note also that Intel's Itanium at its largest on an 180 nm process technology was about 473 mm^2, quite a bit larger. I believe that also might explain why Intel targeted the highest-end markets with Itanium--because those processors sold for so much that very low yields would still be profitable.

As we stand today, with a pricing war going on between Intel and AMD, this native quad-core does not seem to provide enough financial incentive for AMD to have gone with it, even with its potentially much higher performance and scalability.

I see no two ways about it; I think AMD has made a significant mistake in going with the native quad-core design. Had it gone with an approach similar to Intel's, the company could've already had single-package quad-core parts out the door, and their yields would've been greater, their processing power would've only been slightly slower, and they could've been more price competitive.

As it stands, I just don't see the advantages of a native quad-core design. However, I'd very much like to hear your opinions.

Please post your thoughts below.

Source:Clovertown versus Barcelona

Misc: Clarification on Barcelona Benchmark

Last week, I blogged about a V8 8-core system beating a Barcelona 16-core system. However, at the link posted, I am a bit lost since there was no mention of the Barcelona benchmark on the YouTube video, and there was no link about Barcelona hard numbers from the source. Anyway, I just got another email last Saturday (sorry, I just got time to blog it now) that apparently, there is indeed a Barcelona benchmark.

For those who haven't seen the YouTube video, check this out...

Tech Link (Processor): 8 Intel cores faster than 16 AMD cores!

I just got this info from a friend, apparently, uberplus err uberpulse managed to snag a benchmark of a whooping 16core-based Barcelona. I am not really sure I understand the video interview, but they compared a Penryn with Kentsfiled, I have yet to see why Barcelona was even mentioned. Check it out...

At the Microprocessor Forum, Intel demoed its V8 workstation running the POV-Ray benchmark. The machine is equipped with Intel's Workstation Board S5000XVN, 2 quad-core Xeons 5365, clocked at 3GHz and 16GB RAM. And the results are simply impressive: Intel scored over 4,900 pixels per seconds versus a little bit over 4,000 for AMD's 4 sockets quad-core (Barcelona) system. Again, this is an AMD 16 cores system versus Intel's 8 cores V8 machine.

If you take 16 cores at 1.8Ghz, divided by 2 and add 60% to get to 3.0Ghz, it gives POV-Ray score of 3600 for 8 cores. Meaning that at 3.0GHz, Barcelona still lose by quite some compare to an 8 cores Clowertown system.

"Why do you need 16 cores, when you can do better with 8. Our 8 core system is 30% faster than the 16 core machine AMD showed to the press yesterday. I just don't understand how they can claim to be 40% faster", said Francois Piednoel, an Intel engineer present at the show.

Well, now AMD has some explanation to do and sooner rather than later. Because if they can't figure out what happened with those POV-Ray results they showed us, that's the end of it. At the show, Intel also demoed a system with a 45nm Penryn quad-core processor (shipping by year end, about the same time than Barcelona!) that is 40% faster than the top of the line quad-core generation, the Core 2 Extreme processor QX6800. Wow!

Source:8 Intel cores faster than 16 AMD cores!

Rant: ATI Radeon HD 2900 XT Performance Benchmarks

Ok, this is not a Tech Link for a reason, and it's just my random rant. I happen to stumble upon this AMDtech news bit (yes, yes, I know it's DailyTech) from XS. It is about AMD's upcoming video card, and no, that's not my rant, but before I post my rant, let me quote part of the news bits:

Onto the benchmarks. The tests were conducted on an Intel D975XBX2 BadAxe2, Intel Core 2 Extreme QX6700 and 2x1GB DDR2-800 MHz. The operating system on the test system was Windows XP, with a fresh install before benchmarking each card. Testing of the AMD ATI Radeon HD 2900 XT was performed using the 8.361 Catalyst RC4 drivers, while the GeForce 8800 GTS used ForceWare 158.19 drivers.

I bolded the items which I wanted to emphasize. So here's the rant: why would AMD used a Bad Axe 2, of all the boards capable of CrossFire? And didn't ATi already have their very own chipset that supports CrossFire with an AMD CPU? Or if I am mistaken, I know there's a board for Intel that has an ATi chipset that can support Crossfire and quad core (Fatal1ty F-I90HD) so why not this one instead? And even if they really prefer i975, why not Asus for such demo?

I know that they can use anything they like, but I just find it odd that of all the possible alternatives, they end up with a competitor's board, and chipset, not to mention, processor.

Update (04/26/2007): Apparently, DailyTech is the one running the benchmark. I am really not sure, because their first paragraph is as clear as mud and can be very easily mis understood:

After several delays, AMD plans to launch its long-awaited R600 graphics processors. AMD is currently briefing select members of the press on its R600 architecture in the Tunisia, but there is no embargo date on the R600 for DailyTech -- we can show you benchmarks now.

Anyway, here's the update that makes it much clearer now:

* Our benchmarks for Half Life 2: Episode 1 showed an abnormal framerate for the NVIDIA GeForce 8800 GTS card that scaled with lower resolutions -- we believe there was a copy error. We reran the tests this morning and achieved 119.2 frames per second with the GeForce 8800 GTS.

Tech Link (Industry): AMD praying ‘Barcelona’ makes up for four-core mistake

Woah, they finally realized it's the performance that matters and not how the technology is implemented. If products performs, within the consumers' budget, and works like a charm, end users will grab them like hotcakes. I wonder why they never said MCM is better than their 4x4, that is, you need two dual cores to achieve quad core (cripes!!!!). Quotable quotes from the Register:

“If I could do something different, I wish we would have immediately done a MCM - two dual cores and call it a quad-core,” said Mario Rivas, an EVP at AMD, during a recent interview in Austin, “because, I guess, the market sucks it up.”

Expand for more...

By Ashlee Vance in Mountain View (ashlee.vance@theregister.co.uk)
Published Saturday 17th March 2007 00:05 GMT

Exclusive Bruised by a resurgent Intel, AMD wishes it had tackled the four-core era with a different approach. The chipmaker stands behind the technical merits of pumping out a so-called native four-core chip with all four cores on the same piece of silicon.

It, however, admits that Intel gained a major marketing edge by melding a pair of dual-core processors with a multi-chip module (MCM) when it released the "Clovertown" version of Xeon last year. That four-core chip allowed Intel to claim a server processor technology milestone ahead of AMD for the first time in about three years.


“If I could do something different, I wish we would have immediately done a MCM - two dual cores and call it a quad-core,” said Mario Rivas, an EVP at AMD, during a recent interview in Austin, “because, I guess, the market sucks it up.”

Before Clovertown, AMD enjoyed one of the more remarkable runs in server chip marketing and production. It beat Intel to 64-bit extensions for x86 chips and then nailed the release of mainstream, dual-core chips. Besides hitting these milestones, AMD clobbered Intel’s Xeon on overall performance and performance per watt - two of the server world’s favored metrics.

While a high-end part such as Clovertown misses the mainstream, it has proved popular enough with the most demanding customers and analysts to toss the technology and marketing edge back toward Intel. Recent sales figures show that Intel has regained server processor share from AMD, and Intel has shown leading performance on a wide variety of benchmarks.

So, AMD is going retro with its upcoming release of the four-core Barcelona chip, hoping to tap into the momentum of yesteryear.

“Barcelona is as much of event in the x86 world as Opteron was when we launched it,” Rivas said.

Other AMD executives have used this line in recent months, although we’re not sure the pitch fits.

The release of Opteron, as mentioned, gave AMD the first 64-bit x86 chip. More importantly, it turned AMD into a real contender in the server market for the first time.

Barcelona seems like less of a game changer. At the most basic level, it’s simply a better performing chip than today’s dual-core Opterons. AMD expects the processor to provide a 40 per cent surge on most software loads and a much higher boost on floating point-heavy software. Best of all, the chip slides right into existing systems.

AMD hopes to continue the socket compatibility with the four-core follow-on to Barcelona called Shanghai and the eight-core follow-on Montreal - a tidbit you may consider a Register exclusive.

El Reg: Will the upcoming chips Shanghai and Montreal be socket compatible as well?

MR: Yes. That’s the goal. We have not released detailed specs that I know of on those two devices, but that (strategy) has served us well.

Rivas declined to discuss what might happen with Bulldozer - meant to be a native eight-core chip - but hinted that “we have analyst day coming up, and I don’t want to share any more than I have to (before that).”

Channel Stuffing
While Barcelona's release may not match Opteron blow-for-blow, it will stand as a crucial product for AMD’s near-term success.

Sources tell us that the chip is cranking through software at an unreal clip in the labs, beating out AMD’s oft-cited 40 per cent performance gain figure.

Over the last couple of months, AMD executives seemed to change the language around Barcelona’s delivery date, making us wonder if the chip was suffering from a delay. Top staffers started talking up a “late Summer” release rather than a “mid-year” shipments as previously promised.

Rivas said the mid-year plan holds, and we’re looking for Barcelona to ship in July.

The chip can’t come fast enough for AMD’s channel partners who have suffered at the hands of Tier 1 demand.

“We got a little bit distracted,” Rivas said.

AMD has, in fact, blamed its channel shortfall for a recent revenue warning. Rivas said that an unexpected spike in demand for mobile processors along with Tier 1s gobbling up all the available Opterons hurt the supply of gear to the channel.

The channel excuse has rubbed us the wrong way because AMD has spent the last three years celebrating new programs for smaller partners, saying such efforts have been a huge success.

“It was a combination of a boom in our OEM customers and a mix issue that made us not look at our distribution perhaps as well as we could,” Rivas said. “And now we are (suffering) the consequences.”

Moving forward with Barcelona, AMD expects that it will have Opterons for all.

“We will have plenty of silicon” Rivas said. “We will make certain that everyone gets their fair share of the pie.”

Should Barcelona prove as successful as AMD hopes, then the company could be poised to gain server market share against Intel before beginning another game-changing phase.

AMD expects its ATI buy to pay serious data center dividends when the company shifts to 45nm chips in 2008. The chipmaker is looking to combine x86 processors with graphics engines to make products capable of cranking through scientific workloads with unprecedented performance. Such products will be complemented by third-party accelerators that plug right into Opteron sockets - an attack where AMD still enjoys a healthy lead over Intel.

So, come 2008, AMD should gain a shot at seizing the marketing high-ground once again rather than agonizing over missed opportunities. That is if all goes according to plan . . . ®

Source:AMD praying ‘Barcelona’ makes up for four-core mistake