AMD lanserer 90 nm-prosessorer

[Knick Knack]

Det sies ingenting om hvordan de oppnår disse 10 prosentene. For alt vi vet kan CPU'ene utvikle mer varme under samme betingelser som 130nm-utgavene men få besparelsen fra underklokking.

 

Jeg velger å ignorere tallet 10% inntil jeg ser hva forutsetningene er eller til jeg leser noen gode tester på dette.

Jeg synes det virker noenlunde fornuftig å tolke det dithen at de oppnår samme ytelse med 10% effektreduksjon. Om chipene er underklokket blir mer en definisjonssak. Ingen mobile prosessorer er per i dag frekvens begrenset. Alle er begrenset av et effekttak kunstig satt for å gjøre dem brukbare som mobile prosessorer. Dermed er alle per definisjon mer eller mindre underklokket. En kan vel også vente at den prosentvise effektreduksjonen vil variere noe for forekjellig speed grade. Endret 18. august 2004 av Knick Knack

[snorreh]

Detsies at AMD oppnår ca 10% reduksjon i effektforbruket på sine første 90nm produkter. Det er jo ikke ille, men fortsatt et godt stykke unna de glade skalerings dager på 90-tallet.

 

http://news.com.com/AMD+enters+the+90-nano...ml?tag=nefd.top

31 Watt på Oakville ser bra ut, det er en merkbar bedring iforhold til Odessa på 35 Watt som brukes i dagens low-power Mobile Athlon 64 prosessorer. 3000+ basert på Oakville skal ifølge ryktene lanseres snart med samme strømforbruk som dagens 2800+, eller kanskje lavere? Det lover uansett bra med tanke på Lancaster som kommer tidlig neste år, som visstnok er en low-power Mobile Athlon 64 med 1MB L2 cache og et strømforbruk på bare 25 Watt. Inntil da vil Opteron x40 EE fortsatt være den AMD64-prosessoren som forbruker minst strøm med sine 30 Watt.

Endret 18. august 2004 av snorreh

[snorreh]

AMD sneaks strained silicon into chips

Advanced Micro Devices has begun to incorporate a form of strained silicon into its chips, a design twist that will let the company increase the performance of its processors.

 

The strained silicon is being incorporated into all of AMD's 90-nanometer chips, which the Sunnyvale, Calif.-based company has just started shipping to PC makers. The technology also will be added to 130-nanometer chips that will be released this quarter, an AMD representative said Thursday. (The nanometer dimensions refer to average feature size on the chips. A nanometer is a billionth of a meter.)

[snorreh]

AMD and Strained Silicon

While AMD has had papers out in the past that outlined possible strained silicon strategies, it is interesting to see that they implemented this technology without a lot of fanfare. It is a fairly significant step forward for AMD, and the introduction of this technology at both the 90 nm and 130 nm nodes allows for both product lines to advance in performance over what can be expected without strained silicon. The use at 130 nm allows one more faster speed bin to be achieved, and keeps AMD in the performance lead over Intel throughout the rest of the year.

[snorreh]

AMD's 90 nm Strained Products

AMD let leak some months ago that they were going to use some strained silicon features in upcoming products.  The first announced product that does feature this technology is the Athlon FX-55, which is their recently released 2.6 GHz Athlon 64 processor made on the 130 nm SOI process.  There are several ways to strain silicon, and it appears that AMD is currently only using the "stretched" feature.  Compression is not yet used (something that Intel has had very good success with), but it is something that AMD will eventually integrate into their design.  What many were wondering is if strained features were integrated into AMD's latest 90 nm cores?

 

While the 130 nm line was the first to receive the strained treatment, I have just received word that pretty much all Athlon 64 starts now feature strained silicon.  This includes both the 130 nm and 90 nm lines.  While the initial 90 nm Athlon 64's did not have strained silicon, it wouldn't make much sense for AMD to integrate strained in only certain wafer starts and not in others.  The fab business just doesn't like to work that way.  It can be assumed that all Athlon 64 processors by Q4 will have strained features, and these include both 130 nm and 90 nm product lines.  Eventually more strained features will be added to the design mix, as both stretching and compressing silicon leads to better performing and lower power parts.

 

AMD takes advancements like these one step at a time.  Unlike Intel, which utilized a series of strained design features in the Prescott processor on the 90 nm node, AMD has integrated a single strained feature in its current lineup.  Now that AMD apparently has this wrinkle down, we can expect to see more strained features to be integrated in the future.  Most likely the 2.6 GHz and 2.8 GHz 90 nm parts expected in January will only feature stretched silicon, but products after that point will probably utilize more strained features.  From all indications, AMD's 90 nm SOI process is holding up very well, and the overall design of the Athlon 64 has contributed to the success of their 90 nm process (AMD didn't have to push transistor speed to such a degree as Intel did with the Prescott to achieve good performance).

 

AMD definitely has a smoother road ahead of them as compared to Intel.  With the Prescott Intel threw a whole slew of technologies into the design and manufacturing of that product, and they essentially ran out of tricks to get it to go faster and consume less power.  Intel has obviously seen this dead end, and have cancelled the 4 GHz Pentium 4.  It is also very uncertain if Intel will release a 3.8 GHz Pentium 4.  The new strategy for Intel apparently relies on other architectural enhancements such as 2 MB of L2 cache, higher FSB speeds, and eventual dual core processors.  AMD has been able to match Intel's parts until this point with 130 nm products, and only now have they started to release 90 nm parts that reach 2.4 GHz.  AMD is still ramping up performance on their 90 nm line, and they have yet to integrate many of the performance features that Intel's 90 nm process currently possesses.  Compressed strained silicon is the next step, and there are many features left to integrate.  The first 90 nm products off of AMD's line were focused on good yields and low power consumption.  This allowed AMD to have good initial yields and to sell these products for revenue (as compared to trying to achieve high speed at the expense of poor yields).  This gave AMD engineers a chance to work the kinks out of their 90 nm process without the added worry of squeezing every ounce of performance from the new process.

 

We can perhaps get a good feeling where AMD is heading in terms of clock speeds for the Athlon 64 on 90 nm.  If we look at the current range of 130 nm products, we see the initial speed of 2.0 GHz with the 3200+ released one year ago.  The current speed champ for the 130 nm process runs at 2.6 GHz (the FX-55).  We can safely assume that AMD will eventually reach 3 GHz with 90 nm products, and by that time we will also see the introduction of the desktop dual core processor.  With their current roadmap, AMD is not relying on brute clockspeed to achieve their performance goals.  While manufacturing is always a primary concern in processor design, AMD has done its best to limit the negative effects that a poorly performing process might entail.  Luckily for AMD their 90 nm process appears robust, and they can continue to tweak Intel's nose with faster and more competent products.

 

AMD is in an envious situation at the moment.  Not only do they have one of the fastest desktop processors in the world, but they are steadily increasing their 90 nm output.  This gives AMD the ability to produce more chips than they have previously, and also allows them to address a larger percentage of the market.  With the original 130 nm Athlon 64 design, AMD could only theoretically supply around 20% of the market with all of their manufacturing lines maxed out.  With 90 nm and the latest designs, AMD is in a position where they could theoretically address up to 28% of the market.  If AMD continues to keep one step ahead of Intel, then by Q2 of 2005 they could acquire around 23% of the market (I believe their current marketshare is around 18%).  From the roadmaps I have seen of Intel and AMD, AMD should continue to stay one small step ahead of Intel.  From a desktop application and gaming perspective, AMD certainly holds the performance crown.  Intel still holds the performance lead in many audio and video encoding applications.  Now that AMD will have PCI-E based motherboards on the market shortly, the feature race will tighten even more.  As AMD scales the Athlon 64 in clockspeed, then the encoding advantage that Intel holds will start to diminish.  Once both companies have their dual core products on the market, then things will probably get mixed up again and we will see a huge shift in relative performance between the two.

[snorreh]

AMD Athlon 64 Revision E adds SSE3 Support

we noticed something very interesting when looking at the latest AMD roadmaps while in Taiwan: the 90nm chips listed as San Diego and Venice (Athlon 64 FX and Athlon 64 respectively) claimed SSE3 support as a feature, but the current 90nm chips do not have SSE3 support.

 

We went around to quite a few manufacturers asking what the difference was between the 90nm chips shipping today and the 90nm chips that supposedly feature SSE3 support, and unfortunately we were left with very little information - until we learned to ask for the right thing.

 

Internally, AMD's San Diego and Venice CPUs are referred to as nothing more than Athlon 64 Revision E chips. Revision E includes even more bug fixes and performance improvements over those we found in Revision D, including support for the 13 new instructions that were added with Prescott, more commonly known as SSE3.

 

The performance enhancements that go along with Revision E chips include some optimizations to the Athlon 64's memory controller. The more optimized memory controller improves bandwidth efficiency with regards to unified graphics memory accesses; given that the only type of graphics that uses system memory (and thus the on-die memory controller) is integrated graphics, it's safe to say that the Rev E chips will offer better integrated graphics performance.

 

The first Revision E CPUs will begin shipping in early 2005. AMD also has plans to introduce an Athlon 64 4200+ towards the middle of 2005; they are not listing whether the part will feature a 512KB or 1MB L2 cache, but it will most likely run at 2.6GHz. The Athlon 64 FX-57 is also listed on the roadmap as a 2H-05 part, it's specifications are also unclear but we'd expect it to be a 2.8GHz part with a 1MB L2 cache. Both the Athlon 64 FX-57 and Athlon 64 4200+ appear to be 90nm parts built on the San Diego and Venice cores, respectively.

 

AMD's 2005 roadmap did not specifically list anything faster than the 4200+, although the classic "> 4200+" was present on the roadmap to indicate potentially faster parts.