Jump to content
SceptreCore

AMD conFusion? Forget the hype!

Recommended Posts

perhaps since in multiple runs of a gpu benchmark you can expect to see ~3% variance

maybe they ran it till they got one they liked but it would be easier to just cheat outright

maybe amd just got lucky in those tests

or maybe the amd platform can perform a bit better in some gpu limited situations

I'll probably be getting the 8150. If a few more get some BD chips... then us Atomicans can do a little review of our own... and we can make sure of all settings and results.

 

I've decided to go with the 8120. Should technically be able to get similar clocks as an 8150 for less dollars :D Hopefully I can pre-order one through work soon.....

Share this post


Link to post
Share on other sites

G-relk, I suspect no one is going to buy the FX-4100 unless they are really hard up for money.

Share this post


Link to post
Share on other sites

G-relk, I suspect no one is going to buy the FX-4100 unless they are really hard up for money.

I don't know - if they are just a cut-down 81XX, then realistically, as long as one core or whatever is cut from one of each of the BD modules, then performance could potentially be fairly good as seen earlier in that article where this was done and single and dual thread performance went up a slight amount.

Share this post


Link to post
Share on other sites

http://www.xbitlabs.com/news/cpu/display/2...zer_Fiasco.html

Ex-AMD Engineer Explains Bulldozer Fiasco: Lack of Fine Tuning.

 

Engineer: AMD Should Have Hand-Crafted Bulldozer to Ensure High Speed

[10/13/2011 11:21 PM]

by Anton Shilov

 

Performance that Advanced Micro Devices' eight-core processor demonstrated in real-world applications is far from impressive as the chip barely outperforms competing quad-core central processing units from Intel. The reason why performance of the long-awaited Bulldozer was below expectations is not only because it was late, but because AMD had adopted design techniques that did not allow it tweak performance, according to an ex-AMD engineer.

 

Cliff A. Maier, an AMD engineer who left the company several years ago, the chip designer decided to abandon practice of hand-crafting various performance-critical parts of its chips and rely completely on automatic tools. While usage of tools that automatically implement certain technologies into silicon speeds up the design process, they cannot ensure maximum performance and efficiency.

Automated Design = 20% Bigger, 20% Slower

 

"The management decided there should be such cross-engineering [between AMD and ATI teams within the company] ,which meant we had to stop hand-crafting our CPU designs and switch to an SoC design style. This results in giving up a lot of performance, chip area, and efficiency. The reason DEC Alphas were always much faster than anything else is they designed each transistor by hand. Intel and AMD had always done so at least for the critical parts of the chip. That changed before I left - they started to rely on synthesis tools, automatic place and route tools, etc.," said Mr. Maier in a forum post noticed by Insideris.com web-site.

 

Apparently, automatically-generated designs are 20% bigger and 20% slower than hand-crafted designs, which results in increased transistor count, die space, cost and power efficiency.

 

"I had been in charge of our design flow in the years before I left, and I had tested these tools by asking the companies who sold them to design blocks (adders, multipliers, etc.) using their tools. I let them take as long as they wanted. They always came back to me with designs that were 20% bigger, and 20% slower than our hand-crafted designs, and which suffered from electro-migration and other problems," the former AMD engineer said.

Inefficiencies in Design?

 

While it is unknown whether AMD used automatic design flow tools for everything, there are certain facts that point to some inefficient pieces of design within Bulldozer. Officially, AMD claims that the Zambezi/Orochi processor consists of around 2 billion transistors, which is a very large number.

 

AMD publicly said that each Bulldozer dual-core CPU module with 2MB unified L2 cache contains 213 million transistors and is 30.9mm2 large. By contrast, die size of one processing engine of Llano processor (11-layer 32nm SOI, K10.5+ micro-architecture) is 9.69mm2 (without L2 cache), which indicates that AMD has succeeded in minimizing elements of its new micro-architecture so to maintain small size and production cost of the novelty.

 

As a result, all four CPU modules with L2 cache within Zambezi/Orochi processor consist of 852 million of transistors and take 123.6mm2 of die space. Assuming that 8MB of L3 cache (6 bits per cell) consist of 405 million of transistors, it leaves around whopping 800 million of transistors to various input/output interfaces, dual-channel DDR3 memory controller as well as various logic and routing inside the chip.

 

800 million of transistors - which take up a lot of die space - in an incredibly high number for various I/O, memory, logic, etc. For example, Intel's Core i-series "Sandy Bridge" quad-core chip with integrated graphics consists of 995 million.

 

While it cannot be confirmed, but it looks like AMD Orochi/Zambezi has several hundreds of millions of transistors that are a result of heavy reliance onto automated design tools.

The Result? Profit Drop!

 

As a consequence of inefficient design and relatively low performance, AMD has to sell its eight-core FX series processors (315mm2 die size) for up to $245 in 1000-unit quantities. By contrast, Intel sells hand-crafted Core i-series "Sandy Bridge" quad-core chips (216mm2 die size) for up to $317 in 1000-unit quantities. Given the fact that both microprocessors are made using 32nm process technology [and thus have comparable per-transistor/per square mm die cost], the Intel one carries much better profit margin than AMD's microprocessor.

 

AMD did not comment on the news-story.

The price of laziness has been paid?

 

 

http://www.xbitlabs.com/news/cpu/display/2..._Processor.html

Analyst Condemns AMD for Not Employing ARM amid Inability to Design Leading High-Performance x86 Processor.

 

AMD Has Serious Product Problems - Market Observer

[10/13/2011 09:36 PM]

by Anton Shilov

 

The inability of Advanced Micro Devices to address the rapidly growing market of smartphones and media tablets due to the lack of ARM-based solutions in its portfolio amid inability to design a leading-edge high-end x86 central processing unit may lead to sinking into irrelevancy, according to a semiconductor market analyst. Without competitive chips the company will be unable to meet its financial goals going forward.

 

"We maintain our view that AMD is being structurally squeezed by poor competitive positioning and longstanding operational struggles. AMD is on the verge of sinking into irrelevancy as ARM-based competitors gain share in low-end computing and Intel extends its advantages in performance and manufacturing,” wrote Daniel Berenbaum, an analyst with MKM Partners, reports Forbes.

 

AMD is concentrating around the so-called x86 everywhere strategy that dictates the company to tailor x86 architecture for all types of devices, from handsets to high-end servers. Even though the absolute majority of ultra-portable devices are powered by chips with ARM architecture due to its dramatically lower power requirements, AMD asserts that ARM is incapable of reaching performance levels of x86. At the same time, the company is risking to miss the smartphone/media tablet bandwagon as it is only beginning to design x86 chips for the latter and even has no plans for the former.

 

The company's latest high-performance x86 processor micro-architecture code-named Bulldozer failed to outperform Intel Corp.'s latest Core i-series "Sandy Bridge" chips in client applications, which naturally rises concerns about AMD's ability to score lucrative design wins with its latest FX-series central processing units. Moreover, if AMD's chips fail to outperform Intel's products in server applications as well, the company may not be able to regain the high-margin server market share.

 

One of the company's pearls is its ATI graphics processor design unit that manages to introduce competitive solutions. In fact, AMD is one of the two remaining developers of discrete graphics chips for personal computers with about 50% market share. The company can integrate its latest graphics technologies into Fusion accelerated processing units with its latest x86 Bulldozer cores to provide a decent mainstream offering. Unfortunately, AMD has been plagued by product delays for many years now and while analyst are confident in AMD's technologies, they do not believe in consistent execution.

 

"It’s difficult to remember the last product AMD launched on time, and it is now evident that, even aside from manufacturing challenges with partner GlobalFoundries, AMD’s technology roadmap is severely lagging. Third-party reviews indicate that the performance of new products based on the Bulldozer architecture is disappointing – this means that AMD will likely remain a bystander in the ongoing data center build cycle (which has accrued significant benefit to Intel), and will now also miss a window to compete in consumer PCs," added Mr. Berenbaum.

Share this post


Link to post
Share on other sites

2 very interesting articles there mark

 

I have to agree with that :-/

 

Though I will say this, Intel has oodles more money to play with. Wasn't their marketing budget almost as much as AMD's R&D budget?

Share this post


Link to post
Share on other sites

I'm gonna have to read these later. I'm moving house... and it could be up to two weeks before I'm back on the interwebs.

 

Cheers guys.

Share this post


Link to post
Share on other sites

Well that would explain a lot towards why bulldozer is less than expected. At the very least AMD released something even if it was a large disappointment, so now they have a working model of their architecture maybe they can redesign/refine the design by hand in order to reduce the problems with the SoC design.

 

Mind you 20% would be good but still wouldn't beat intel or even match it in single threads it would however have made it less of a disappointment and a good advance on thurban

Edited by UberPenguin

Share this post


Link to post
Share on other sites

Why do people keep saying AMD should enter the ARM market, they have enough trouble competing with 2 competitors, imagine if they had to go against dozens?

Share this post


Link to post
Share on other sites

http://www.insideris.com/amd-spreads-propa...yee-speaks-out/

 

Here is a quote from the AMD ex employee himself, who posted few comments about the whole situation some time ago:

 

On paper bulldozer is a lovely chip. Bulldozer was on the drawing board (people were even working on it) even back when I was there. All I can say is that by the time you see silicon for sale, it will be a lot less impressive, both in its own terms and when compared to what Intel will be offering. (Because I have no faith AMD knows how to actually design chips anymore). I don’t really want to reveal what I know about Bulldozer from my time at AMD.

What did happen is that management decided there SHOULD BE such cross-engineering ,which meant we had to stop hand-crafting our CPU designs and switch to an SoC design style. This results in giving up a lot of performance, chip area, and efficiency. The reason DEC Alphas were always much faster than anything else is they designed each transistor by hand. Intel and AMD had always done so at least for the critical parts of the chip. That changed before I left – they started to rely on synthesis tools, automatic place and route tools, etc. I had been in charge of our design flow in the years before I left, and I had tested these tools by asking the companies who sold them to design blocks (adders, multipliers, etc.) using their tools. I let them take as long as they wanted. They always came back to me with designs that were 20% bigger, and 20% slower than our hand-crafted designs, and which suffered from electromigration and other problems.

That is now how AMD designs chips. I’m sure it will turn out well for them [/sarcasm]

 

BTW, you ask how AMD could have competed? Well, for one thing, the could have leveraged K8 and the K8 team’s success and design techniques instead of wasting years of time on a project that eventually got cancelled using people that had never achieved any success. It took Intel years to come out with Nehalem, and AMD could have been so far ahead by that point that they’d have enough money in the bank that they wouldn’t have to accept a low-ball settlement offer in the antitrust suit and they wouldn’t have to sell off their fabs.

Gives a totally different perspective on why Bulldozer failed, doesn’t?

Share this post


Link to post
Share on other sites

Sad story TBH. Although it's very much in hindsight (never really thought this before) IMO K8 was probably a better basis to move forward from. It never had the kind of 'epic fail' built in that say Netburst did - and in fact Netburst is arguably the last attempt by someone to design a new chip topology from the ground up. Intel fixed that failure by developing the P3 design into Core, then Core 2, then Nehalem and now Sandy. Surely the relatively under-resourced AMD should have doen the same with their fantastic K8.

 

Wouldn't STARS + some cache basically have been competitive with Nehalem? What then if they'd spent Bulldozer time and money on evolving that topology? I dunno, I'm very relutant to comment on shit that I could never do in a million years, but when an ex-AMD engineer's saying the same thing...

Share this post


Link to post
Share on other sites

To be honest, I don't a complete tweak of the Stars architecture would have helped anyway. Look at Llano and how it compares against Sandy Bridge. Completely nothing could save it. This is coming from someone that has used bloody Stars in the past few years. Plus given the problems in the 32nm, it would have been a nightmare to do. I don't think the K10 has any flexibility left. It was all-out speed or nothing for that architecture.

 

But to be honest, we have two completely different situations. Intel went for improvement after improvement with an architecture they know has the flexibility (the P3 architecture is actually quite modular when I look at the original tech specs). AMD didn't for their K8 and K10 architecture so they had to do a new architecture. We know that new architecture would have teething problems and hopefully with the tweaks promised, we might see Bulldozer show off its strengths.

Share this post


Link to post
Share on other sites

Sad story TBH. Although it's very much in hindsight (never really thought this before) IMO K8 was probably a better basis to move forward from. It never had the kind of 'epic fail' built in that say Netburst did - and in fact Netburst is arguably the last attempt by someone to design a new chip topology from the ground up. Intel fixed that failure by developing the P3 design into Core, then Core 2, then Nehalem and now Sandy. Surely the relatively under-resourced AMD should have doen the same with their fantastic K8.

 

Wouldn't STARS + some cache basically have been competitive with Nehalem? What then if they'd spent Bulldozer time and money on evolving that topology? I dunno, I'm very relutant to comment on shit that I could never do in a million years, but when an ex-AMD engineer's saying the same thing...

but if you think about it, they did use the massive money pool from Athlon64x2 to start the Turion and then BD, and when BD needed more Dev time, they pushed into further K8 development (which they eventually pulled off quite well) and continued BD in the background... BD has been ongoin for a long time, definately long before the X4's came out... So really, i think they already HAVE done that step of furthering a previous chip, and this is kind-of a really-really delayed netburst equivalent. I think mostly due to the SOC/Design-Consolidation-with-ATI move discussed earlier is quite plausible too, and i think ur right that further rev's will regain a lot of that ground (will just come down to how much improvement IvyBridge brings i guess?)

Share this post


Link to post
Share on other sites

pretty much, yeah!

im looking forward to the potential for another turion-style turnaround though! i think the *could* still pull it off if price right & improved fast enough

 

i think sceptrecore drew a comparison earlier that it took like, 12 months last time to turn around the turion so --heres to hopez--

Share this post


Link to post
Share on other sites

From all the articles I've read, including Mark's it's pretty evident that Bulldozer's design is fundamentally flawed. AMD could surprise us but there's no fundamental reason to assume the design will turn around?

Share this post


Link to post
Share on other sites

From all the articles I've read, including Mark's it's pretty evident that Bulldozer's design is fundamentally flawed. AMD could surprise us but there's no fundamental reason to assume the design will turn around?

You never know! I mean Intel's Hyperthreading was pretty flawed when first released, they only got it right later in later CPUs past the P4. OK, it did take a little while ...

Edited by Jeruselem

Share this post


Link to post
Share on other sites

From all the articles I've read, including Mark's it's pretty evident that Bulldozer's design is fundamentally flawed. AMD could surprise us but there's no fundamental reason to assume the design will turn around?

You never know! I mean Intel's Hyperthreading was pretty flawed when first released...

 

Agreed. The next core revisions, piledriver and steam roller, should show some good speed ups as they iron this new architecture paradigm out.

 

I also wonder, now that they've got it out the door, whether they'll turn back to manual tweaking to get back the lost potential performance, power draw etc?

 

 

I don't know that I'd call Bulldozers design per se flawed, it's more the execution of putting it together (the automated stuff) that perhaps has fouled up it's potential. If what that ex-AMD engineer is saying is correct.

Share this post


Link to post
Share on other sites

Don't forget that rumour of the aborted architecture before the rush to bring out k10. So a lot of dollars disappeared there. I really think for all the stuffing around they have done they could have designed a truely competitive Phenom III. And yeah while you say that AMD has enough trouble competing with only two competitors and so why go into ARM market, as somebody once said 'The best way to loose a fight is not to turn up' well AMD have finally shown up to this fight with bulldozer, and they showed up to the tablet fight with z01.

 

The fact that AMD totally bypassed this bulldozer core for their Trinity project and went with stars and then piledriver tells me that they may have fixed a lot with the power consumption issues and clockspeed fail. They also say trinity will have 50% more cpu power than llano. llano in laptops runs at 1.4ghz so we are looking at the equivalent of a 2.1ghz phenom quad core. That's far from horrible. lets not dismiss the fact that both llano and trinity are built on the same manufacturing process, so there is not going to be a significant extra transistor budget here, trinity's graphics seem to be superior to llano's which probably means that it takes up more die area than llanos graphics. which in turn means that the piledriver CU's take up less space than the stars/husky cores they replace.

 

As has been said above. We have gone the whole way from Pentium pro to Core i7 as an evolution and we have gone from k6? through to k10.5 as an evolution. Intel had their go at starting fresh with p4 and what did they end up with out of it. Faster memory controllers and a mean branch predictor. Which they then stuck back into a p3 to make a core 2 duo. As long as AMD can survive this then they may end up with a stronger processor on the other side of it.

 

Really when you think of the evolution of these two cpu's and the fact that they were most similar through the p3 and athlon days. its no wonder new architectures don't work and the software is built for those processors. AMD probably would have been best of continuing the low core count phenom series for the desktop and the bulldozer series for the server space. but that would be too hard for AMD to maintain, it's obvious where AMD getting it profits from and they want to keep them in the server space.

Share this post


Link to post
Share on other sites

Any new architecture is practically a minefield in terms of manufacturing and architecture bugs. That's guaranteed. I don't think any remembered how the initial Pentium IV architecture was. Initially, it was a flop similar to Bulldozer given it was slower than the Pentium III before it was given tweaks and die shrinks. After that, it was considered capable of taking on the Athlon64. Then the lessons learnt in the Pentium IV was plain to see (the memory controller and as well as branch predictors) in the current Core i series.

 

The same thing could be said of the original Phenom which flopped but it was giving a few tweaks which lead to the Phenom II. I'm guessing it'll take Piledriver to really get things going. And AMD isn't losing that much in the server space which is what Bulldozer is designed to be competitive at.

 

EDIT: Looking at the articles, I do think Trinity will be interesting as it'll show what AMD can do really.

Edited by sora3

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×