Sign up for our Technology Newsletter 
AMD Offers More Radeon Chips
AMD has announced four Radeon HD 8000M series GPU chips sporting its latest Graphics Core Next (GCN) architecture.
AMD’s GCN architecture made its first appearance in the firm’s ultra high-end mobile chips, however next month’s CES show will see the firm show off laptops featuring four Radeon HD 8000M series chips. The firm’s four Radeon HD 8000M chips are pitched at the mainstream and gaming laptop markets, though the company said that Asus already has a laptop announced that will use the chips in an ‘ultrathin’ design.
AMD’s Radeon HD 8000M series sees the firm split three chips with the same number of stream processors and memory clock speeds that scale up to 1.25GHz but differentiated by their core speeds. The firm’s Radeon HD 8500M, Radeon HD 8600M and Radeon HD 8700M all have 384 stream processors but are clocked up to 650MHz, 775MHz and between 650MHz and 850MHz, respectively.
Topping AMD’s present range is the Radeon HD 8800M, which has 640 stream processors and is clocked at between 650MHz and 700MHz, while its GDDR5 memory is also clocked at 1.25GHz. All of the firm’s chips, by virtue of being based on the GCN architecture, support DirectX 11.1.
AMD said it will launch three more chips in the Radeon HD 8000M series in the second quarter of 2013. According to the firm’s roadmap, two of those chips will sit above the Radeon HD 8800M in terms of performance while the third will straddle somewhere between the Radeon HD 8600M and Radeon HD 8700M.
AMD was tight-lipped on the power figures for its chips, saying that full details of its Radeon HD 8000M series chips will appear at CES, where its partners will tip up with laptops sporting the chips.
TSMC To Boost 28nm Production
TSMC is able to make chips using 28nm process technology at a speedier pace that it originally anticipated. This means that the chipmaker will likely be able to meet demand for existing orders and start accepting new designs.
TSMC promised to increase its 28nm capacity to 68 thousand 300mm wafers per month by the end of the year. It did this by ramping up fab 15/phase 2 to 50,000 300mm wafers a month. According to the Taiwan Economic News it looks like the outfit managed to beat its own projections, which should be good news for customers like AMD, Nvidia and Qualcomm. Well not AMD of course. It just told Globalfoundries to stop making so many of its chips so it can save a bit of money.
But it looks like TSMC is flat out. In November the fab 15/phase 2 processed 52,000 wafers. When combined with fab 15/phase 1, TSMC should be able to process 75 – 80, 000 300mm wafers using 28nm process technologies this month. TSMC produces the majority of 28nm chips at fab 15, which will have capacity of more than 100,000 300mm wafers per month when fully operational.
AMD Shows Piledriver Opteron
AMD’s Piledriver rollout is all but complete. With Trinity in the mobile and desktop space, new 3300 and 4300 Opterons are bringing the new architecture to data centers.
The Opteron 4300 series offers six different parts, in quad-, six- and eight-core flavours. Stock clocks range between 2.2GHz and 3.5GHz, with TDP’s in the 35W to 95W range. The cheapest Opteron 4334 costs $191, while the priciest 4332HE comes in at $501. The 3300 series consists of three quad- and eight-core SKUs, priced at $125 to $229. The pricing of both series is pretty aggressive.
But what’s next for AMD? Well things should be eerily quiet on the server front in 2013. Abu Dhabi, Seoul and Delhi/Orochi C should last throughout 2013 and even a good part of 2014. That’s when we can expect some major changes, as AMD transitions to 28nm and goes about transforming its Opteron lineup.
Future Low Power CPUs and APUs (as AMD calls them) should replace Dehli/Orochi-C in 1P and dense server markets, but AMD is also planning “Client APUs for market enablement,” and this sounds a lot like ARM-based low voltage parts. Of course, in the high end AMD plans to stick with big Steamroller cores, but mid-2014 is a long way off.
Radeon 8000 HD Expected In Q2
Since we only have one month left in 2012 and due the fact that graphics companies rarely announce something big in December, it is obvious that Radeon HD 8000 slipped to 2013.
Our well informed industry sources are confirming that the next generation, based on Sea Islands, architecture is coming in 2013 and some of them dare to say that it will be Q2 2013 rather than Q1 2013. Some people were expecting to see the cards in Q1 2013 but even according to AMD’s own roadmap Sea Islands, the new GPU architecture with HAS features was scheduled (or should we say delayed. Ed) for 2013.
AMD has already communicated this schedule loud and clear in its February 2012 roadmap update, and even then it killed hopes that Sea Islands or HD 8000 cards are coming in very late Q3 or Q4 2012, as was previously expected.
We won’t get into any specific details like the 8000 branding, or die sizes as we simply don’t know this at the time being. It’s safe to say that these cards will end up faster than 7000 series and at similar TDPs to the previous generation, all manufactured in 28nm.
nVidia Speaks On Performance Issue
Nvidia has said that most of the outlandish performance increase figures touted by GPGPU vendors was down to poor original code rather than sheer brute force computing power provided by GPUs.
Both AMD and Nvidia have been using real-world code examples and projects to promote the performance of their respective GPGPU accelerators for years, but now it seems some of the eye popping figures including speed ups of 100x or 200x were not down to just the computing power of GPGPUs. Sumit Gupta, GM of Nvidia’s Tesla business said that such figures were generally down to starting with unoptimized CPU code.
During Intel’s Xeon Phi pre-launch press conference call, the firm cast doubt on some of the orders of magnitude speed up claims that had been bandied about for years. Now Gupta told The INQUIRER that while those large speed ups did happen, it was possible because of poorly optimized code to begin with, thus the bar was set very low.
Gupta said, “Most of the time when you saw the 100x, 200x and larger numbers those came from universities. Nvidia may have taken university work and shown it and it has an 100x on it, but really most of those gains came from academic work. Typically we find when you investigate why someone got 100x [speed up] is because they didn’t have good CPU code to begin with. When you investigate why they didn’t have good CPU code you find that typically they are domain scientist’s not computer science guys – biologists, chemists, physics – and they wrote some C code and it wasn’t good on the CPU. It turns out most of those people find it easier to code in CUDA C or CUDA Fortran than they do to use MPI or Pthreads to go to multi-core CPUs, so CUDA programming for a GPU is easier than multi-core CPU programming.”
AMD’s Roadmap Leaked
According to the latest AMD desktop roadmap, published by DonanimHaber, the Steamroller architecture could be delayed, which means Piledriver cores will power AMD’s third-generation APUs.
So what does this mean for consumers? Well, Richland APUs might not be quite as good as expected. AMD could optimize the x86 cores and go for more powerful graphics, but it’s hard to get excited about the next generation.
Vishera parts will also stick to Piledriver cores, backed by discreet Radeon 7xxx and 8xxx series graphics. However, we will see a new architecture in the low end. Kabini is apparently on track to launch next year, with Jaguar cores, refreshed graphics and an all new system-on-chip version, with an integrated on-chip Fusion controller hub (FCH).
What’s more, AMD will also offer quad-core Kabini parts, and who could say no to a dirt cheap E-series APU with four cores, good graphics and a ridiculous TDP?
AMD Makes More Cuts
There have been some rumors that AMD plans to drop prices on its older generation APUs as well as some AM3 Athlon II CPUs and as of yesterday, the new price list confirmed those cuts. Unfortunately the price cuts are not that impressive on the FM1 side but some AM3 Athlon II CPUs have seen a price reduction of over 30 percent.
The full list includes a total of six A-Series APUs and twelve AM3 Athlon II CPUs. The most impressive price cut in the A-Series was on the A4-3300 APU for 21.7 percent, or from US $46 down to US $36. The flagship A8-3870K got cut down by 9.9 percent while A6-3670K was cut down by far less impressive 3.7 percent.
The most impressive price cuts on the AM3 Athlon II side is on the Athlon II X4 640 and the Athlon II X2 265. The Athlon II X4 640 got cut down from US $98 to US $67, or 31.6 percent, while Athlon II X2 265 got cut down from US $69 to US $48, or 30.4 percent.
Two Athlon II CPUs, the X3 445 and the X4 638, were removed from the price list as they were most likely discontinued.
You can check out the full list and price cuts over at CPU-World.com.
Is The x86 Falling
According to Mercury Research, worldwide shipments of x86 parts saw a sharp decline in Q3. Researchers claim the drop was the biggest seen in more than a decade, 9 percent year-over-year.
Despite the drop, Intel still has something to brag about. Intel’s share hit 83.3 percent, up from 80.6 percent sequentially. AMD’s share dropped to 16.1, down from 18.8 percent, while VIA garnered a 0.6 percent share.
Mercury Research analyst Dean McCarron told PC World that both AMD and Intel experienced declines, but AMD took more of the hit than Intel.
“AMD was simply hit by what OEMs saw in the markets… and hitting the brakes,” he said.
What’s more, the third quarter is supposed to be traditionally strong for x86 chipmakers, thanks to the back-to-school shopping frenzy. However, x86 CPU shipments dropped 4 percent in Q2, followed by 9 percent in Q3. Things aren’t looking good for Q4, either.
“The key is how the macroeconomic situation is, which is not looking good for the next couple of quarters,” McCarron said. “Hopefully things will improve next year.”
ARM Goes High-End
Nvidia is itself an ARM chip licensee that has seen significant design wins with its Tegra 3 system-on-chip (SoC) processor, however the firm doesn’t see ARM based servers being able to do heavy lifting in server tasks for two years. Sumit Gupta, GM of Nvidia’s Tesla Accelerated Computing business unit said that even with GPGPUs, ARM based servers are not yet able to provide the computing power needed to drive high performance servers.
Gupta said, “Performance of these ARM cores is still not where it needs to be for servers. It is getting there; the new ARM64 [processor] is going to get it part of the way.” However he did say that eventually ARM SoCs could hit X86-like performance levels. “One day I think ARM will at least get to similar performance levels as X86 performance. The belief is that over the next one or two years these ARM SoCs will be good enough for cloud applications and web serving. I think it will take some more time to be good enough for accelerated computing.”
As for Nvidia using its Tegra chips to push work to the firm’s GPGPUs, a scenario that would make the firm’s accountants very happy, Gupta said he was surprised at the level of interest from developers and questioned the need for powerful CPUs. “We did a small development kit called Karma that has a Tegra 3 and a Nvidia GPU, [and] I was shocked by the number of those kits that have been sold. The interest in this ARM plus GPU is far larger than even I expected. If the GPU can do dynamic parallelism, it becomes more independent than how powerful CPUs do you need? I believe the first thing that will happen is that people will start using lower performing [Intel] Xeons […] then at some point when these Atom based processors become available they might use that, and when ARM64 is available they’ll use that.”
AMD Goes Piledriver
AMD has released its Piledriver desktop
processors codenamed Vishera.
AMD showed off Vishera at IDF last month, an overclocked chip running at 5GHz. Now the company has taken the wraps off its eight-core Vishera chip, a processor that it hopes will finally bury memories of its disappointing Bulldozer Zambezi chip.
AMD’s Vishera processors will continue to use Socket AM3+, meaning it is a drop-in upgrade for those customers lumbered with Zambezi processors.
The firm announcing four models all branded with the FX moniker. The low-end Vishera chip is the quad-core FX-4300 clocked at 3.8GHz boosted up to 4.0GHz, with 4MB of Level 3 cache.
The firm has kept feature parity throughout its Vishera FX range aside from core count and total Level 3 cache. Therefore AMD’s six-core FX-6300 still sports the same 1MB of Level 2 cache per core but has a total Level 3 cache of 8MB and is clocked at 3.5GHz that can be boosted up to 4.1GHz.
AMD’s top two Vishera parts, the FX-8320 and FX-8350 sport eight cores and have 8MB of Level 3 cache. The difference between the two chips is their clock frequencies, with the FX-8320 running at 3.5GHz and boosted to 4.0GHz while the FX-8350 is clocked at 4.0GHz and is boosted to 4.2GHz.
The firm’s decision to clock its FX-8320 and FX-8350 so closely is largely academic, as all Vishera chips feature an unlocked multiplier. AMD even plays up the overclockability of Vishera and touts 5GHz as being reachable with water cooling. Insiders have even said it can reach 5GHz with strong air cooling.
As for AMD’s Piledriver architecture, the firm claims it offers improved branch prediction and improvements to Level 2 cache efficiency and scheduling. Overall the company is sticking to its longstanding line that Vishera is a 15 percent performance increase over Bulldozer, and while that might well be true, Bulldozer was so far behind its competition in single-threaded performance that a 15 percent gain is needed simply to achieve parity, let alone a lead.