backtop

Print E-mail del.icio.us 83 comment(s) - last by eye smite.. on Dec 17 at 11:38 PM
AMD's candid Mario Rivas discusses AMD's next-generation processors, and some of the problems with its old ones

AMD has had more than its share of problems over 2007 with possibly the worst issue being with its new Phenom and Opteron processors.  With the current problems with AMD's K10 processors, now the question becomes, "Will AMD make its deadline for the next processor?"

ChannelWeb Network sat down for a phone interview with AMD Executive Vice President of the Computing Products Group Mario Rivas for some more information on the bug and how AMD plans to recover from the torrent of bad press that has resulted.

Rivas says that the bug started as an observation and it wasn’t until mid-November that it actually turned into a more serious bug. Rivas also said that the company tried to do BIOS workarounds and patches with a 90% success rate.

In a closing comment, Rivas details some heavy information about the company's next-process chips. "We have 45nm on the way. We will have initial samples also in January. I'm fairly confident that those puppies are going to boot, and then we can have a follow-up conference call and I'll tell you, 'The sucker is booting.'"

Typically, when a processor is first taped-out, a operating system boot is one of the markers of a successful design.  Intel's 45nm Penryn, for example, loaded the Windows XP operating system on the first spin.  11 months later, the processor began shipping for volume.

According to AMD’s John Pellerin, director of logic technology and development, AMD plans to ramp production of 45nm  chips in the first half of 2008. Pellerin says that AMD is more concerned with finding customer applications for its processors rather that a process race with rival chip makers. 

Pellerin said in an interview during the International Electron Device Meeting that AMD’s 32nm high-k parts were in the development phase showing that AMD is already looking down the road. The company hopes to learn from the problems encountered with its early 45nm process.

AMD public relations claims the company has not announced plans for high-k metal gate on its 45nm node.  The company's roadmap plans to debut high-k metal gate in 2010 with 32nm, with the option of introducing it on the 45nm node at that time.

Rivas closes, "We also have 32nm advance work in SRAMs, which as you know is the initial step. So we will be a fast follower again, and as long as we have architectural advantage, our 45nm will be as good as the other guy's 32nm."

Intel announced its 32nm test shuttle just three months ago in September 2007. If AMD does have a 32nm test shuttle already, the year differential for 45nm may shrink considerably with AMD's 2010 architecture launches.


Comments     Threshold


This article is over a month old, voting and posting comments is disabled
blind devotion
By Andypro on 12/14/2007 2:11:37 PM , Rating: 4
Rivas closes, "We also have 32nm advance work in SRAMs, which as you know is the initial step. So we will be a fast follower again, and as long as we have architectural advantage , our 45nm will be as good as the other guy's 32nm."

Guess what? You don't.




RE: blind devotion
By retrospooty on 12/14/2007 2:26:16 PM , Rating: 2
Either he is living in the past (Athon 64 over P4) or has intimate knowlegde of the future. Likely its the past ;)


RE: blind devotion
By erikejw on 12/14/2007 7:46:43 PM , Rating: 5
Currently Intel is smoking AMD in every possible way.
Their Core2 architecture is great.

However I don't think AMDs Phenom is that bad as everyone else
says(I would never buy one since Intel is better however).

Penryn is just so much better than K8 but K10 is big step towards being competitive again. K10 is about 10% slower clock for clock. The large problem is the speed of the ship, 2.3 GHz doesn't cut it.

AMD has always been slower when they ramped new architecture s due to their less resources. They got the K8 dual to work at 3.2 GHz at 90nm. Now they have 2.3 GHz at 65nm which is abysmal at the time. I expect them to ramp up the speed quite quickly during 2008 since the architecture itself is designed to reach higher MHz than K8. Will it be enough?
Intel can release faster CPUs whenever they want, they have headroom.

I don't expect them to overtake Intel in any way but I expect them to get much closer. If they get like 2 speed grades below Intel they should be able to charge quite good for their chips and actually make some money and hopefully get in the black this year or at least the 2008H2.


RE: blind devotion
By retrospooty on 12/15/2007 9:30:30 AM , Rating: 2
I agree with you there. If AMD can get the cache bug fixed and speeds up Phenom can be awesome (well, its awesome now, its just that Core2 is more so).

The big obstacle coming now is Nehalem. Its due in late 08 and moves to internal memory like AMD chips (which is AMD's one big advantage).


RE: blind devotion
By Zurtex on 12/14/2007 2:27:26 PM , Rating: 3
AMD's current 65nm parts almost run clock for clock with Intel's current 45nm parts. But that's not the point, he's talking then not now, which is as you can imagine what AMD need to be doing.

Really do hope AMD learn from their mistakes and pick themselves off the ground and make another good run at Intel.


RE: blind devotion
By Master Kenobi (blog) on 12/14/2007 2:30:18 PM , Rating: 2
Clock for Clock AMD does not run "almost" with Intel's 45nm line. They don't even run "Clock for Clock" parity with Intel's 65nm line, what makes you think they can compete with Nehalem on 45nm? Wishful thinking on your part indeed.


RE: blind devotion
By Zurtex on 12/14/2007 2:41:13 PM , Rating: 2
I'm sorry, but did I miss something?

Phenom generally performs about 6-8% worse than Conroe with the odd benchmark than an equivalently clocked Core2Quad, with the exception of a few benchmarks where it performs a little better. Right? Penryn doesn't really add major performance and I could be wrong but it looks to me that it adds stunning extra performance of 9% or so mainly in areas where Phenom used to look a little competetive.

If not Anandtech needs to revise its benchmarking articles. That kind of difference is only made a big deal about when it's 2 different companies, if it's the same company releasing news products with that difference everyone laughs.


RE: blind devotion
By Master Kenobi (blog) on 12/14/2007 2:57:34 PM , Rating: 5
So let me sum up in your own words.

6-8% Slower than Intel's 65nm year old chips. (Conroe)

9-15% Slower than Intel's 45nm month old chips. (Penryn)
(This is subjective since to fix errata, AMD had to dump 10-20% performance to make them not lock up)

New Phenom and Barcelona processors won't be out until March (According to AMD) and all existing ones were yanked from shelves due to this error.

July timeframe Intel will be launching Nehalem and the new QuickPath platform.....

Call me skeptical, but where in here does AMD look like it's making progress? It's about to be a process node AND an architecture behind just like it was 2 months ago. AMD won't be rolling out 45nm until mid 2009 (or late 2009 given their track record) when Intel will be launching a 32nm shrink on the Nehalem platform. AMD seems to be hedging its bets on Fusion at this point to turn them around.


RE: blind devotion
By grenableu on 12/14/2007 3:01:54 PM , Rating: 2
AMD obviously has to put a positive spin in things, but you really need to have your head in the sand to not see how grim things are starting to look for them.


RE: blind devotion
By Andypro on 12/14/2007 3:04:18 PM , Rating: 3
I agree with everything you said sans the Nehalem launch. No way Nehalem is ready for launch by July. I have speculation from an Intel insider that it simply won't be ready for a full launch until Q4.

Despite that "hiccup," nothing AMD is doing leads me to believe they'll have anything close to competition in 08 or 09. They're re-rolling out K8 ffs.


RE: blind devotion
By rninneman on 12/14/2007 3:37:34 PM , Rating: 1
quote:
No way Nehalem is ready for launch by July. I have speculation from an Intel insider that it simply won't be ready for a full launch until Q4.


The operative term there is speculation. Based on Intel showing a functioning Nehalem system at the fall IDF and their recent track record, let's just say I would not be betting against them launching when Intel says it will.


RE: blind devotion
By Targon on 12/14/2007 5:47:20 PM , Rating: 3
At this point, we don't know how much of the performance of Phenom has been lost due to the cache issues. It may be that AMD isn't just losing the 10 percent or so caused by the BIOS workaround for the bugs, but is also losing another 10 to 20 percent in addition to that.

At this point, it is too early to be sure just how far behind AMD is due to the bugs in the K10 launch processors. This isn't being blinded by "fanboiism", it is more about not knowing enough about what is broken in the current K10 processors to know what their true potential is.


RE: blind devotion
By vortex222 on 12/16/2007 10:02:42 PM , Rating: 2
I AM an AMD fanboy. Have been for a long time, and I always try to give them the benefit of the doubt whenever i can. I still run my socket 939 X2 4600 and looking to upgrade my video to an ATI 38xx video card. I think its great value for what i get.

But at this point if i were looking to buy a higher end platform I would be a total lunatic to go with an AMD quad core. C2Q has advantage in performance, thermals, and especially overclocking (thats not that big of concern for me however). I cant even justify Phenom on cost-per-performance.

If the Phenom was more competitive in price, and more comparable in performance I would most definatly go for it because its AMD.

Coming from an AMD Fanboy who always hopes they will put out the best product they can, The only thing i don't like about the Core2 is the fact its made by Intel.


RE: blind devotion
By theapparition on 12/17/2007 12:41:15 PM , Rating: 2
Clock for clock, Phenom was 5-15% slower than Penryn before Phenom BIOS updates. After the BIOS bug-fix update, it gets worse at 15-25% slower.


RE: blind devotion
By Zurtex on 12/14/2007 3:10:09 PM , Rating: 5
I'm not a raving fanboy I said:

"Really do hope AMD learn from their mistakes and pick themselves off the ground and make another good run at Intel."

I don't see why you need to pick a fight, 6 - 8%, 9-15%, these under benchmarks where Intel does better (yes if you read the benchmarks there are a couple of things the AMD chips did faster) I mean give me a break. Those kind of figures don't move mountains, it's in Intel's process that it holds massive advantage not their architecture (certainly when we're talking about clock for clock, or performance per watt or heat dissipation)

I'm not saying Intel don't have a huge advantage, but let's call a spade a spade and not mix things up.

Nehalem will launch, it's an exciting change for Intel, but we'll see benchmarks when we see benchmarks. I think if anything looking at the history of Intel vs. AMD it's never best to count the chicks.


RE: blind devotion
By 16nm on 12/14/2007 7:56:21 PM , Rating: 3
quote:
it's in Intel's process that it holds massive advantage not their architecture (certainly when we're talking about clock for clock, or performance per watt or heat dissipation)


I think this is a fair statement. K10 and Conroe are pretty similar and they are more or less the same clock-for-clock. It's Intel's superior processes that gives them the greater efficiency, performance per watt and cooler operation. Plus, it doesn't hurt the great overclocking abilities of Intel chips. AMD have a good 90nm process, but they really need to keep working on 65nm, IMO.


RE: blind devotion
By JumpingJack on 12/14/2007 8:49:46 PM , Rating: 2
10% difference clock for clock, that is not pretty similar:

http://www.tomshardware.com/2007/11/19/the_spider_... (compare 2.4 Ghz to 2.4 Ghz)...

It get's worst... even the 2.6 Ghz part does not achieve parity with the q6600: http://www.bit-tech.net/hardware/2007/12/13/amd_ph...


RE: blind devotion
By xti on 12/15/2007 2:31:30 AM , Rating: 2
10% isnt going to matter to any of the major customers they have. Even server customers dont care about speed, they rather take lower and stable. thats why the barcelona flop is something that matters, and the 10% isnt going to bother anyone other than geeks.

now...are they making money off these 10% slower-chips?


RE: blind devotion
By JumpingJack on 12/15/2007 2:56:53 AM , Rating: 1
Hmmmmmm.... depends on the price and power would it not... regardless of what you can percieve (though 10% is clock for clock, intel's top bin will out shoot AMD's top bin by much greater)....

However, in server... this is where performance matters the most... and the all more critical Perf/Watt...

In retail (the Best Buy/CompUSA -no defunct) I agree with you... AMD still has 50% of the retail market... but they aren't making money there, at the moment, obviously. Retail is tricker because costs are so important on thinner margins.

In the high end, AMD is just not there... like I mentioned the 10% number is just clock for clock... Intel is clocking up much higher and the popularity shows in the revenue achievements... compare Intel this year to AMD this year.


RE: blind devotion
By crystal clear on 12/15/2007 3:41:07 AM , Rating: 2
The price wars have left "No profits" for AMD on these chips.

Marketshare or profitshare ?

If you were an IT manager-you choose the best available servers at the best possible price(for you).

Would you take the risks of choosing Barcelona if any tier 1 or even 2 OEMs having anything to offer.

When it comes to buying decisions-there is no place for fanboyism,rather hard core financial & reliability issues etc takeover in your purchasing decisions.

"Beware of little expenses; a small leak will sink a great ship."


RE: blind devotion
By leexgx on 12/15/2007 5:43:27 PM , Rating: 2
but if thay allready have am2 servers it cost them an alot more to upgrade to intels rimm ram (what ever there new one is called) and cpu

then just putting Barcelona chip in and haveing 2 more cores

new servers core 2 probley better option now but for upgradeing curent amd systems to K10 chips seems better cost


RE: blind devotion
By Spuke on 12/15/2007 10:42:51 PM , Rating: 4
quote:
but if thay allready have am2 servers it cost them an alot more to upgrade to intels rimm ram (what ever there new one is called) and cpu
Businesses don't swap servers every 3 to 6 months like the enthusiast community does. We update our servers every 3 to 5 years or longer. I still have an "old" Athlon 64 2800 server that's JUST now getting updated and that's only because we need more onboard SATA ports and more RAM. We buy based on features, price, and proven stability not outright speed. That measly 15% performance difference doesn't justify the price difference between Intel and AMD. Our servers run fine on X2 5000's and below and the "extra" money can be used for more hard drives or more memory or a better power supply, etc. BTW, this is a corporate environment with 6000+ users.

Also, there will be no Phenom's or any of the new stuff from anyone in our servers. They simply haven't been on the market long enough and we're not guinea pigs.


RE: blind devotion
By bfellow on 12/17/2007 4:57:26 PM , Rating: 2
Why in the world are you using X2 5000 instead of Opterons for servers? Using single socket servers?


RE: blind devotion
By vortex222 on 12/16/2007 10:19:04 PM , Rating: 2
I think on a given process its fair to say AMD always had the best chip per NODE in the end. However Intel getting there first and usually focusing on the next advancement was the reason for that.

AMD simply did process jumps 'the budget way'.

250nm. Katamei p3 vs Pluto.
180nm, Coppermine and Willimite vs Palamino.
130nm, Northwood and Banias vs Barton and Sledgehammer. (albiet barton got beeten up bad by northwood)
90nm were at Prescott, Smithfield(?) and Dothan vs Toledo and Windsor.
at 65nm it may be fair to say Intel will be the first to be ahead in that respect. We will see what Barcelona will do at the end of its life before its shrunk.


RE: blind devotion
By 91TTZ on 12/14/2007 10:31:39 PM , Rating: 2
quote:
it's in Intel's process that it holds massive advantage not their architecture (certainly when we're talking about clock for clock, or performance per watt or heat dissipation)


This is incorrect. The process is completely independent of performance per clock cycle. That's entirely due to the architecture. When they shrink the process, they're making the exact same CPU, only smaller. Since performance per clock is due to the architecture, making the chip smaller will do nothing for that.

It will make it run cooler, and be cheaper to produce. Also, chipmakers usually roll out architectural improvements when they do a process change, but the increased per-clock performance is due to those architectural changes and not due to the process.


RE: blind devotion
By Zurtex on 12/14/2007 11:00:39 PM , Rating: 2
o...k, people seem to have misread what I was trying to say.

I know: "The process is completely independent of performance per clock cycle. "

That's what I meant, I meant that their process allows them ramp up large clock speeds and keep moving towards the next die shrink...


RE: blind devotion
By eye smite on 12/17/2007 11:38:31 PM , Rating: 2
Lets talk about blind devotion a moment. I've been on amd products since the k6-2 with 3d now and loved every minute of it. I've built athlon systems, athlon t-birds, athlon xp's, athlon 64's and athlon x2's. Do you think for one instant that intel that burnt me yrs ago with the pentium can win me back? I'd rather run a Via cpu. I don't care about fastest, what I care about is product support and intel doesn't have it. Lets introduce a new architecture, chipset or instruction set every 3 months so that the guy that just spent alot of money can be obsolete and spend it all again, that's intel's approach and has been for years. I guess everyone's forgotten you wouldn't have all this wonderful performance if amd hadn't beaten intel over the head for 3 yrs with athlon64. All you guru's of the tech world, do you still have your P4's running and if so is it more than just a file server at your house? I have more need for performance than most of you here because I run med research units for world community grid, but guess what my ahtlon 64's and x2's do it just fine. What do you do that's so important with your core 2, game?


RE: blind devotion
By maroon1 on 12/14/2007 4:23:54 PM , Rating: 1
quote:
Phenom generally performs about 6-8% worse than Conroe with the odd benchmark than an equivalently clocked Core2Quad


65nm Core 2 Quad performs 10% faster than equivalently clocked Phenom

http://www.tomshardware.com/2007/11/19/the_spider_...


RE: blind devotion
By Mitch101 on 12/14/2007 5:40:31 PM , Rating: 2
AMD's 45nm chips are K10.5 the current Phenom is K10. Its possible that K10.5 is on par or slightly faster than the current cores however thats Marketing and we know if Marketing can find one benchmark that shows superior then it must be superior.

I'm not holding my breath after the K10 disappointment I am getting a Wolfdale or Yorkfield in the next 2-3 months.


RE: blind devotion
By KristopherKubicki (blog) on 12/14/2007 5:49:47 PM , Rating: 2
From what I've read, Shanghai is just an optical shrink of Barcelona -- with some fixes/tweaks I suppose.

I would be hesitant to give it even a half designator.


RE: blind devotion
By Mitch101 on 12/14/2007 5:56:33 PM , Rating: 2
I suspect its those fixes and tweeks that will give it just enough to inch ahead of the Intel design but most probably it will be a 50-50 split. Certain benchies for Intel and certain benchies for AMD.

Intel will probably still overclock further.

All I know is I am in the market for a CPU and Yorkfield is on my radar after a few AMD 939 chips.

I would be curious to know how many chips AMD could sell if they made a 939 6400+ chip. But thats probably too late now.


RE: blind devotion
By Regs on 12/14/2007 6:54:47 PM , Rating: 2
They released the Barc thinking it was a finished product. As in it had no problems.

So the real question is, does AMD even know what the problems are with the K10? Sure, the fixed L3 cache will help, but there might be something else the architecture is not taking advantage of.

Could be another Williemate vs Northwood debacle, which we all hope for.

Either way, I hope someone at anandtech will make an article of what AMD changed in each stepping. That's if the end result leads to much better gains for AMD.


RE: blind devotion
By 16nm on 12/14/2007 7:50:03 PM , Rating: 2
quote:
From what I've read, Shanghai is just an optical shrink of Barcelona -- with some fixes/tweaks I suppose.


Yes, absolutely. I would hope they fix the L3 cache bug! LOL. Can you imagine hearing about this bug on the new Shanghai chips? Ugh- That would be extremely frustrating for those of us hoping for the best for AMD...


RE: blind devotion
By ajfink on 12/14/2007 8:56:22 PM , Rating: 2
The cache bug should be fixed in the next round of 65nm silicon. I doubt it will slip to 45nm. As for 45nm, I would be surprised to see more SSE4 implementation and perhaps a few different L3 cache tweaks, but not much else.


RE: blind devotion
By crystal clear on 12/15/2007 4:50:43 AM , Rating: 2
AMD in its recent "analyst day meeting" has admitted that "Barcelona is not a manufacturing problem but rather design flaws that has caused the delays".

OPENING COMMENTS (from President and COO Dirk Meyer)

"We've done one thing very poorly, Namely, We haven't delivered our quad-core products consistent with our plan. Last month in the final stages of validation we uncovered a design error...

In reality Barcelona is just not ready for release yet,it will ttake a full quater (1Q 08) to get ready for release.

Dont expect general availabity in 1Q 08-push it to the 2Q 08.

We need "perspicacity"
perspicacity \pur-spuh-KAS-uh-tee\, noun:
Clearness of understanding or insight; penetration, discernment.

Note this-

During the company's third-quarter earnings conference call, AMD's President Dirk Meyer said: "We're building 45 nanometer microprocessors as we speak...[and will] be starting our production ramp of 45 nanometer processors in the first half of next year."

Dont you see/note he is just f?????

artifice: ....

Now as you say-Shanghai is just an optical shrink of Barcelona !

When Barcelona is itself delayed (1 full year from summer 2007 to summer 2008)then what optical shrink can there be.

Dont you see-

Built by design and artifice, it fell apart in confusion and chaos.


RE: blind devotion
By JumpingJack on 12/14/2007 8:11:11 PM , Rating: 2
Well, this is utterly false.


RE: blind devotion
By Polynikes on 12/14/2007 10:21:26 PM , Rating: 2
quote:
Really do hope AMD learn from their mistakes and pick themselves off the ground and make another good run at Intel.

So do I. Competition makes for a wealth of choices for us, the consumers. I mean, a Q6600 for $266? I'm in heaven.


RE: blind devotion
By 91TTZ on 12/14/2007 10:26:02 PM , Rating: 2
quote:
AMD's current 65nm parts almost run clock for clock with Intel's current 45nm parts.


That's a meaningless statistic, since the performance per clock cycle is dependent on the architecture, not the process. Making the chip smaller does not alter its per-clock performance in any way. It may enable them to clock it higher, but it's still going to do the same amount of work per clock cycle.


RE: blind devotion
By Master Kenobi (blog) on 12/14/2007 2:27:29 PM , Rating: 2
I think the better question is what makes AMD think they can fast track from 45nm to 32nm. Their 65nm process sucks, and the bugs need to be worked out there, shrinking will just cause more problems. I think this exec is oversimplifying the harsh reality.


RE: blind devotion
By Zurtex on 12/14/2007 2:35:02 PM , Rating: 2
And that's the point, it does seem to be their process more than anything that particularly sucks. Hope they've learnt something from their mistakes that they've not shared with us yet or something from their work on die shrinking the graphics chips.


RE: blind devotion
By Jack Ripoff on 12/14/07, Rating: 0
RE: blind devotion
By KristopherKubicki (blog) on 12/14/2007 3:01:11 PM , Rating: 3
While I don't usually get in the middle of these, Theo de Raalt is quite possibly one of the most egotistic self-promoting people I've ever (virtually) met.

His rant was long discredited by others who have a little bit more knowledge about processor errata, including Linus himself.

All processors are "riddled" with bugs by Theo's definition. The difference, of course, is that AMD is now 3 months behind their schedule because of the TLB bug.


RE: blind devotion
By Jack Ripoff on 12/14/07, Rating: 0
RE: blind devotion
By jhtrico1850 on 12/14/2007 4:34:13 PM , Rating: 3
http://aceshardware.freeforums.org/viewtopic.php?t...

Seriously though, Mr. de Raadt is just whining that he wrote poor code and now he's being forced to change it. You can read it all in the forums at RWT...

David Kanter of Real World Tech


RE: blind devotion
By KristopherKubicki (blog) on 12/14/2007 4:47:10 PM , Rating: 2
Linus replies somewhere in that thread too.


RE: blind devotion
By KristopherKubicki (blog) on 12/14/2007 4:51:48 PM , Rating: 2
Hell we even wrote an article about it:
http://www.dailytech.com/article.aspx?newsid=7897


RE: blind devotion
By KristopherKubicki (blog) on 12/14/2007 4:46:41 PM , Rating: 3
quote:
Let me get this straight: is giving your code away completely freely rather than licensing it under a viral license that requires something back in compensation being "egoistic"?

It's clear you've not had much personal contact with Theo.

OpenSSH and OpenBSD's funding from the U.S. DOD and UPenn was yanked because Theo protested the war in Iraq. I don't disagree with his stance, but it's very clear that OpenBSD is "his" project, and anyone who disagrees with him gets conveniently tossed aside (packet filter, the NetBSD fork, etc).

That's not open source, that's Theo-source.

Matt Dillon's response speaks for itself. Of the 50 bugs Dillon and Raadt list, they only agree that 2 or 3 are exploitable.

But let's not forget either that these are very smart guys, and even though the errata has been posted since June, nobody has ever produced a working exploit based on this errata.

No, this is CPU errata as usual. I've taken enough classes in software security, hardware security, operating system design and exploit design to know otherwise.

Here's the funny part in all this. Intel documented problems it had with TLB in its errata, and 5 months later AMD recognized it had a similar but more serious bug. These were one of the "2 or 3" Theo listed as exploitable.


RE: blind devotion
By Ringold on 12/14/2007 9:28:29 PM , Rating: 1
quote:
is giving your code away completely freely rather than licensing it under a viral license that requires something back in compensation being "egoistic"?


Have you ever met communists? Real, red-blooded political communists? They are no more noble, no more humble, no less inherently evil or egotistic than the rest of us. In fact, that they believe so ardently in a single ideology makes many some of the most arrogant people I've ever met. That goes for other ideologies as well, but just because his ideology is "free" software means absolutely nothing. Man is an animal, the license attached to his labors doesn't change that he's an animal and has all the flaws everyone else does.


RE: blind devotion
By JumpingJack on 12/14/2007 9:59:12 PM , Rating: 2
K10 Errata
http://www.amd.com/us-en/assets/content_type/white...

K8 Errata
http://www.amd.com/us-en/assets/content_type/white...

C2D Errata
ftp://download.intel.com/design/processor/specupdt...

All processors have bugs that are triggered under unique conditions or a very rare sequence of events. Some are more serious than others, normally by the time a processors is released, AMD or Intel have sufficiently tested the processor that even with the 'bug ridden' processors, they are perfectly stable and capable.

In some cases, the bug is general enough, severe enough in the consequences (system hang), and occurs at a high enough frequency, that processors will be delayed from launching...

The TLB errata, while unfortunate, was one that AMD's quality control process did not catch and they allowed the launch to happen with this bug. Subsequently, they halted shipment of mission critical parts (i.e. server) but let the desktop users deal with it ... most will see a lockup and curse MS anyway.


RE: blind devotion
By Spuke on 12/15/2007 10:56:25 PM , Rating: 2
What's wrong with the previous post that ranks it a one? That's good info in that post. If it gets ranked down further, I'll repost it.


RE: blind devotion
By JumpingJack on 12/16/2007 12:49:55 AM , Rating: 2
Many people do not like to see the dirty laundry hung out in the wind, hence people vote it down. :)

My point in post errata documents from both companies was simply to show both companies deal with obscure bugs, that 99% of the time are meaningless....

It's the 1% that escape and wreak havoc that generate the horrid PR... Pentium FDIV bug is one example, this TBL bug is another. In the TBL bug case the timing could not be worse.


RE: blind devotion
By JumpingJack on 12/16/2007 12:54:16 AM , Rating: 2
Well that stunk, let me try again...

Corrected:

Many people do not like to see the dirty laundry hung out in the wind, hence people vote it down. :)

My point in posting these errata documents, lists from both companies, was simply to show both AMD and Intel deal with obscure bugs, that 99% of the time are meaningless....

It's the 1% that escape the labs and wreak havoc that generate the horrid PR... Pentium FDIV bug is one example, this TLB bug is another. In the TLB bug case the timing could not be worse.


RE: blind devotion
By Screwballl on 12/14/2007 2:34:51 PM , Rating: 1
quote:
Guess what? You don't.


guess what... they do

The current C2D/Quads are faster (but still not more efficient or an architectural advantage) than AMDs current because they are still using the older but still more efficient Athlon64 design. Look at Athlon/Athlon64 versus Pentium 4/EE/D. AMDs 130nm versus Intels 90nm (with 130nm faster, and more efficient) and then the eventual jump to 90nm to help maintain dominance at the time. Look at the numbers, a 90nm processor with the same TDP as a 65nm proc? Which is more efficient? The 90nm. So this means the 65nm should have a much lower TDP from AMD than it does from Intel. Even AMDs ACP (average) of 55W is much lower than Intels (average) TDP ratings of 65-135W.
Phenom is a small step up but with 45nm coming, we will see if AMD can start to close that gap again and be on top like they were in the Athlon days. Plus with 32nm already where it is at it will be another interesting few years (after a boring past 2 years).
Fanboys have no say in this, if AMD can pull this off then Intel is in real trouble AGAIN.


RE: blind devotion
By Screwballl on 12/14/07, Rating: 0
RE: blind devotion
By MrTeal on 12/14/2007 2:52:04 PM , Rating: 2
The clock for clock performance crown is currently in Intel's hands, not AMD's. Take a look at the benches, the 2.4Ghz 9700 is soundly beat by the 2.4Ghz Q6600. The 9700 is AMD's new processor design.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?...

Yes there is a process difference there, but the smaller node in and of itself won't improve performance. It can lower TDP and help ramp up clock speeds, but it won't affect clock for clock performance.


RE: blind devotion
By Screwballl on 12/14/2007 3:16:37 PM , Rating: 1
It is very tough to compare clock for clock performance nowadays with onboard memory controllers, multicore systems, procs with adjustable clock speed per core, applications that still cannot take advantage of more then one core at a time and other core advances. Yes Intel is currently faster than anything AMD currently has out but as fast as this technology is evolving, AMDs 45nm may just pull another Athlon sleeper surprise. We are looking at a matter of months now before the next gen parts are going to be battling it out.


RE: blind devotion
By MrTeal on 12/14/2007 4:06:54 PM , Rating: 2
I would include the IMC in AMD's clock for clock performance. If anything, not including the benefit of it and just comparing the performance of the CPU itself would be worse for AMD. We'll have to see what sort of benefit it bring to Nehalem, but as it stands right now, at the same clock speeds, Intel's offerings are faster.

I'd love to see AMD make a big comeback with their 45nm process, but I'm not holding my breath. We've seen a small increase in performance from Conroe to Penryn, mostly in SSE-optimized code. If AMD gets a similar lift with tweaks going to 45nm, they might close the clock for clock gap that exists between Phenom and Penryn. Problem for them is, by the time AMD's 45nm fab is online, Intel's more mature 45nm will most likely have
1) higher yields, which will give more flexibility in pricing
2) higher volume, so better binning allowing higher speed grades
3) Nehalem.

Unless Nehalem is the second coming of Prescott, or is massively delayed, it will be very difficult for AMD to compete on anything other than price. Worst case for AMD, the new core from Intel is as big of a jump as the Athlon 64s were from Barton.


RE: blind devotion
By tinyface on 12/14/2007 11:49:42 PM , Rating: 2
Unlike Intel, AMD's first generation 45NM process will not use high K/metal gate. This means the leakage (gate to channel, source to drain) will be even worse and may prevent AMD achieving higher clock speeds. AMD's (IBM's actually) SOI (more expansive Si) would theoretically help limit gate leakage, but it doesn't appear to be working well for AMD at 65 nm node, and will remain questionable at next node given the struggle AMD has at 65 NM.


RE: blind devotion
By JumpingJack on 12/14/2007 8:45:32 PM , Rating: 2
It's not tough at all, you clock them with the same clock speed, you run application X on both, and which ever one completes the computational task faster is clock for clock faster.

There is not magic to this.... IMC on die or off die simply shuttle data to the cores.

Variable speed cores are only variable when the computer is idle once loaded they run full bore.

Most benchmarking sites turn off the power saving utilties anyway in order to get the best representation of computational performance.

AMD's 45 nm process will be like their 65 nm process in most probability since AMD is not taking steps to solve the gate leakage problem until 32 nm (or potential 1/2 way through 45 nm).


RE: blind devotion
By calyth on 12/14/2007 5:15:56 PM , Rating: 2
Yes, you're right. Clock per clock, the Phenom's at a disadvantage.
But with the tons of Phenom reviews that compares it to the Q6600 (the QX are simply beyond most people's budget, and sure, they're faster), a 2.6GHz Phenom part are generally on par - in which I mean that the differences are probably statistically insignificant. And honestly, even for encoding jobs, I tend to set them overnight, or when I'm at school or at work, so the differences are not that big of an impact.

Sooner or later, the current Phenom pricing would have to drop out of the sky, simply because their parts are not clock-per-clock compariable. For most people, yes even a guy like me who have OCed systems before, they look at the price-performance ratio, not just raw performance. Granted that had I known Phenom would turn out fairly poorly, I would've gone for C2D system, but as they fix things (and even the Intel fans better hope there are real competition), I'm quite lined up for a nice and easy upgrade.

There are nforce boards out there that have compatibility issues. There are preliminary previews that suggests Wolfdale may have compatibility issues.

There are issues that people should consider than just that 6-10% performance difference when they buy their systems.


RE: blind devotion
By 91TTZ on 12/14/2007 10:34:19 PM , Rating: 2
quote:
Yes there is a process difference there, but the smaller node in and of itself won't improve performance. It can lower TDP and help ramp up clock speeds, but it won't affect clock for clock performance.


Correct. Finally somebody gets it right.


RE: blind devotion
By Haltech on 12/14/2007 6:47:23 PM , Rating: 2
do you buy processors based on a performance jump or a 30W difference in TDP? Just curious.


RE: blind devotion
By tinyface on 12/14/07, Rating: 0
RE: blind devotion
By JumpingJack on 12/15/2007 2:57:54 AM , Rating: 1
Quite true...


RE: blind devotion
By crystal clear on 12/15/2007 3:05:04 AM , Rating: 2
NO your title of the comment "Blind devotion" is wrong.

This RIVAS guy has lot of REGRETS/Grumblings/Complaints/ etc about Barcelona/Phenoms.

Much of this stuff(REGRETS etc) was removed from the original interview to CRN.(I read it all)

His boss HECTOR RUIZ had all the bad stuff about AMD in the interview pulled down,days before the much publicized "Analyst day meeting".

Hector did some extreme HECTORING on RIVAS & on CRN -

-read the dictionary meaning of Hector-

2. (lowercase) a blustering, domineering person; a bully.
3. a male given name.
–verb (used with object) 4. (lowercase) to treat with insolence; bully; torment: The teacher hectored his students incessantly.
–verb (used without object) 5. (lowercase) to act in a blustering, domineering way; be a bully.


http://dictionary.reference.com/browse/hector

RIVAS & CRN are extremely FRANGIBLE !!!!

frangible \FRAN-juh-buhl\, adjective:
Capable of being broken; brittle; fragile; easily broken.



They are not K.K. of D.T. types.

In short -

RIVAS regreted just everything that happened at AMD's launch of BARCELONA & PHENOM.
He is a very disappointed person & extremely unhappy person at AMD.

He is "cacophonous"...............
harsh or discordant sound.


He is not alone there is cavalcade like him who have left & still leaving or plan to do so.

As for his statements,they are best described as-

tarradiddle: a fib; also, pretentious nonsense.

YES pretentious nonsense its the word to describe it.


RE: blind devotion
By eye smite on 12/17/2007 11:27:03 PM , Rating: 2
The first real quad core instead of 2 dual cores sandwiched together and you say they don't? Catch the clue train when it comes by.


AMD Fusion
By Shida on 12/14/2007 2:25:51 PM , Rating: 2
Is this guy hinting about AMD Fusion?




RE: AMD Fusion
By KristopherKubicki (blog) on 12/14/2007 2:27:31 PM , Rating: 2
I think everyone at AMD is always hinting at Fusion.

When I first checked, the CPU part of Fusion was a 45nm core. Not surprisingly, 45nm Nehalem is the target core for Intel's GPU + CPU projects.

I think waiting for 32nm on Fusion is not something the company can do -- if Intel gets there first they won't let go of that market.


RE: AMD Fusion
By Master Kenobi (blog) on 12/14/07, Rating: 0
RE: AMD Fusion
By crystal clear on 12/15/2007 6:42:14 AM , Rating: 2
Everyone at AMD were always hinting at Barcelona/Phenoms a year ago(2006/7).

You/we saw it all what happened.

Now they picked up FUSION.

The reason the company has missed on execution is that it has been trying to do too much at once. The reason that it's trying to do too much at once is that it has fallen behind competitively."
It would be wise for AMD "to be less ambitious, take it a bit slower, and execute cleanly."

What about the Bull dozer ?

AMD's bluff and bluster in the past has sown the current pessimism. Though not related directly to quad-core production, there were other troubling signs.

One was the silence on Bulldozer that had been touted heavily at the last analyst meeting as "designed to be the highest performing single and multi-threaded compute core in history."

Fusion & Bulldozer some joke -another paper launch ....or one day disappears from those Roadmaps.(until you catch them).

"We all know art is not truth. Art is a lie to make us realize the truth."


RE: AMD Fusion
By Shida on 12/17/2007 10:59:15 AM , Rating: 2
You know I have suspected something similar, and yeah it's with sympathetic incline: what if it's that AMD, with it's limited monetary financial resources, is just trying to save money and put it to further research on Fusion and 32nm?

In other words, they are cutting some corners here and there to then produce Fusion and stuff?

I don't know but it's something that I think is suspect during this time to AMD's lackluster performance.


cut to 32 nm
By flipsu5 on 12/15/2007 1:51:09 AM , Rating: 2
AMD's problem is its executives don't know how to manage the technology and products.

AMD is in a bad enough situation that it has nothing to lose to skip 45 nm and go straight to 32 nm in 2008. More development samples per wafer is not a bad thing. Especially if it made the immersion investment. Intel is still vested in dry lithography for current product and cannot insert immersion as quickly as AMD.

The architecture does not fulfill mainstream needs, mainly server market, which is low volume anyway. Can they find a killer GPU application?

How about full-depletion SOI or starting a bulk silicon line?




RE: cut to 32 nm
By crystal clear on 12/15/2007 2:14:58 AM , Rating: 2
quote:
AMD is in a bad enough situation that it has nothing to lose to skip 45 nm and go straight to 32 nm in 2008.


Read my comment-

Now can you believe these bold highly optimistic statements?
By crystal clear on 12/15/07,


By ChipDude on 12/15/2007 12:39:01 PM , Rating: 2
Looks like AMD is in full spin and confession these days and hopes by confessing to the widely already known screwups on Barcelona promises and Phenom follow disaster it will some how give credibility to their projections that future design and technology is all well?

It would be great to believe they are on track but what you are seeing right now is that the cost to do leading edge CPUs takes signfiant and sustained invesment in design, technology at the multi-billion dollar level for many years.

AMD underinvested and as a results is late on 65nm technology and design had issues. By any metric neither the design nor technology have been shown to be anything but traling edge. How can anyone believe without sustained investement and past successfuly designs and technology that they have the building blocks to be successful on 45nm let alone 32nm.

My prediction is they will be late on both 45nm and even more so on 32nm. You simply can't leap frog a generation. There are no short cuts when it comes to technology and AMD is the last company that can afford adding the required resources and money to try to short cut the development and ramp of a leading edge CPU design and technology.




By flipsu5 on 12/15/2007 5:29:55 PM , Rating: 2
It is because their management makes them a follower that they are always behind, and in a fast business, this lag can grow exponentially.

But as far as what they need for 32 nm, they already have the materials and lithography to produce test chips.


By flipsu5 on 12/15/2007 6:49:13 AM , Rating: 2
http://www.semiconductor.net/article/CA6510988.htm...

IBM and AMD already have separate 32 nm developments as well as collaborative development of 32 nm HKMG, both very important. The collaborative result process is usually lowest risk but is also of lower worth to each of the members; each member will tailor its 32 nm for its own market.

Of course a 2009 timeline is more realistic for 32 nm, but the point is a focus on 32 nm now (keeping an eye on 2008) would be the only chance for AMD to get out of its quagmire.


By crystal clear on 12/15/2007 8:59:11 AM , Rating: 1
Yes I agree with you on this.

Also-

The reason the company has missed on execution is that it has been trying to do too much at once. The reason that it's trying to do too much at once is that it has fallen behind competitively."
It would be wise for AMD "to be less ambitious, take it a bit slower, and execute cleanly."


From my comment earlier above.


process race
By casket on 12/15/2007 3:10:02 PM , Rating: 2
"Pellerin says that AMD is more concerned with finding customer applications for its processors rather that a process race with rival chip makers."
-- This is exactly the problem. The game is all about process rame... and AMD is losing this race. They better start taking this process race seriously... or they're dead.

It's alot harder to find customer applications when you have an inferior product. AMD should be working on 25nm... they need to start playing catch up... and quickly.




Speed kills
By casket on 12/15/2007 3:14:46 PM , Rating: 2
"AMD is the last company that can afford adding the required resources and money to try to short cut the development and ramp of a leading edge CPU design and technology."
-- Either you spend the money up front to try and get this edge... or you suffer with inferior products that sell for $100 / processor and fall even further behind.

They better up there resources and try to afford it... or else they are a dead-man-walking.




what's that smell?
By GeorgeOrwell on 12/14/07, Rating: -1
RE: what's that smell?
By DallasTexas on 12/15/2007 11:00:29 AM , Rating: 1
That was a great response to the article.

Both factual with entertaining sarcasm - a "5" ! When the children in here grow a pair, this site will return back to being the great site that it was. Now it's polluted with MySpace refugees.


RE: what's that smell?
By KristopherKubicki (blog) on 12/15/2007 9:41:55 PM , Rating: 1
I read all of GO's replies -- some are incredibly witty and well written. I think he only gets modded down for the rhetoric, but I always mod him up when I can.


"If a man really wants to make a million dollars, the best way would be to start his own religion." -- Scientology founder L. Ron. Hubbard











botimage
Copyright 2009 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki