backtop


Print 78 comment(s) - last by mars777.. on Jul 15 at 7:12 AM

Former ATI president and CEO resigns from AMD

AMD today announced Dave Orton, executive vice president of AMD, has resigned. The resignation is effective in the end of July 2007. Orton was the former president and chief executive officer of ATI before the merger with AMD. A new executive vice president has not been appointed yet.

“It is with mixed feelings that I am leaving AMD,” Orton said. “I am very optimistic about AMD’s future. I believe strongly in the strategies that brought AMD and ATI together and the talented employees of the ‘new AMD’ who are committed to winning in the market by delivering the best possible solutions for customers.”

Orton was one of the “key drivers in the successful integration of AMD and ATI,” according to AMD President and Chief Operating Officer Dirk Meyer. “With his integration work complete and the successful launch of key graphics and chipset products earlier this year, the time was right for Dave to take his personal and professional life in a different direction,” he added.

Despite the “successful integration” of AMD and ATI, the company continues to hemorrhage cash, with debts mounting up fast and an estimated $1.1 billion in the bank. The company also posted a first quarter loss of $611 million USD.  



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

wa
By ixelion on 7/10/2007 2:45:07 PM , Rating: 2
quote:
and only $1.1 billion in the bank


that sounds wrong to me




RE: wa
By kamel5547 on 7/10/2007 3:03:31 PM , Rating: 2
March 31st 2007 balance shows the following:

Current Assets
Cash And Cash Equivalents 991,000
Short Term Investments 176,000
Net Receivables 751,000
Inventory 938,000
Other Current Assets 331,000

Their cash in the bank is probably at most 1.1 billion unless they've borrowed money since then. (there is a 2 billion bond offering that was in the works but I'm not sure if it was placed or how much of that was eaten up on cap-ex)


RE: wa
By Spivonious on 7/10/07, Rating: 0
RE: wa
By Spivonious on 7/10/2007 3:09:54 PM , Rating: 2
Oops I meant another two quarters like they already had. Let's hope Barcelona doesn't tank like R600 did.


RE: wa
By smitty3268 on 7/10/2007 3:46:22 PM , Rating: 1
Most of that 600 million dollar loss was buying ATI, and they clearly aren't going to do such a massive deal again anytime soon. So even if Barcelona tanks worse than R600 they'll be around a lot longer than 2 quarters.


RE: wa
By MartinT on 7/10/2007 4:16:32 PM , Rating: 3
The first quarter loss had only a little to do with ATi's acquisition or performance, but was largely due to a huge shortfall in AMD's processor volumes, combined with shrinking ASPs. (AMD stuffed the channel in Q4, and had to sell through that volume first, before being able to move more product in Q1.)


RE: wa
By Chernobyl68 on 7/10/2007 5:10:54 PM , Rating: 2
I just hope AMD can return to profitability before they run out of cash


RE: wa
By TomZ on 7/10/2007 5:17:23 PM , Rating: 4
Here, I corrected that for you:

I just hope AMD can return to profitability before they run out of credit


RE: wa
By smitty3268 on 7/10/2007 10:31:13 PM , Rating: 1
Do you have a source for that, because I was pretty sure that wasn't the case? Could certainly be wrong, though.


RE: wa
By mmarq on 7/10/2007 7:51:59 PM , Rating: 6
quote:
so basically if they have another two years like they already had, they'll be bankrupt


The situation is not famous for sure but, they wont go away anytime soon.

AMD can play that game of 'price cut' because since they transitioned to a 300mm die area process they can compete with Intel on price because 'brisbane' is almost 20mm2 smaller.

http://www.chip-architect.com/news/2007_02_19_Vari...

Besides, considering the scaling factores in the above site, based on the actual implementations and not the ideal, that gives a medium scaling factor for logic+sram of 0,59 they would be very foolish to sell now that they have an advantage, because they too can *EASELY* play the MCM game.

A 4 core MCM like Penryn at 45nm process will be 107mm2 x2 = 214mm2

Extrapolating a 4 core Barcelona at 45nm process will be
283mm2 x 0,59 = 167mm2

so:
A 2 core Penryn at 45nm process will be 107mm2

a 2 core Phenom at 45nm process will be 83,5mm2

again:
Extrapolating from Penryn that has 6.0mm2/MB for the SRAM a 8 core MCM Shangai(the K10,5 next revision for Q2/08) with 16MB L2/L3 at 45nm process will be 167mm2 x 2 = 334mm2 + (6x8)mm2 = 382mm2

To compete Intel will have to have a MCM with 4 Peryns that will be 4x107mm2 = 428mm2 (hugely expensive) or develop the very long process of making a 4 core chip.

Being compatible without doubt with AM3 and socketF a 8 core single chip from AMD shangai will be like a bucket of water full of ice down the Intel head.

http://en.wikipedia.org/wiki/AMD_K10

I'm not agains Intel and 'Yes people dont care of what is inside the package', but Barcelona is not only the number of cores in a single die, it is a superior technology that know is maturing, is relatively slow and cache starved, but will absolutely "kick ass" with its 8 core MCM when it arrives... and it will most probably arrive because its such a natural a small step a 2 chip MCM.

Bottom line is that AMD will be able to always offer the same number of cores for less money. In the meantime until Q2/Q3 2008 45nm process, Barcelona dont have to be much more performant than Penryn, it will only have to be equal and price competitive.


RE: wa
By zsdersw on 7/10/07, Rating: -1
RE: wa
By Kietsche on 7/11/2007 6:32:22 AM , Rating: 3
Great way to miss the point of a thoughtfully written post.

Your post wasn't a sentence either, since it didn't start with a capital letter. In addition, the English language is governed by usage, and has a continually expanding (particularly so in recent years due to the technology industry - would you be one of the people who said "robot" wasn't a word before it was in the OED?) lexicon in response to this. "Performant" gets "about 2,490,000" hits on Google UK as of today. Lastly, if you didn't know what it meant, it's pretty easy to look up.


RE: wa
By zsdersw on 7/11/07, Rating: -1
RE: wa
By Andrwken on 7/10/2007 11:26:31 PM , Rating: 4
Interesting argument, but it is pretty clear that Intel has always held the advantage in process and speed of implementation, not Amd. And if you believe Intel has just started pondering a native quad core, that is naive. Also, volume is probably more critical to a healthy business model in a price war as this, which last I checked Intel is what 80% marketshare? Intel will always be able to sell cheaper and maintain profit due to volume alone. So, until that changes I believe your statements are a bit hopeful.


RE: wa
By mmarq on 7/11/2007 1:09:05 AM , Rating: 2
Yes Intel as the bulk advantage of volume, but i believe this is not much a price war the way i sense you imply, AMD is not losing money per chip sell, in a desperate attempt to retain market share... AMD can target the more varied market segments better, and always did it from a long time ago...

Its not always bulk volume or maximum performance attainable by >1000$ factory overclocked chips the only thing that matters... its efficiency also.



RE: wa
By Andrwken on 7/11/2007 2:15:03 AM , Rating: 3
Yes efficiency in process can give them a big boost, but if performance is not up to snuff, then your efficiency gives you a very modest profit on the midline and nothing on the bottom end, just like now. The Barcelona ramp is a problem in that when they do get up to the promised speeds, they may be dealing with an entirely new architecture from Intel again, which initially looks to eliminate the last of Amd's technical superiorities, like IMC. And if Amd is not losing money per chip, then they have to be losing on lack of volume (idle facilities). They aren't bleeding cash for nothing.


RE: wa
By mmarq on 7/11/2007 1:20:22 PM , Rating: 2
quote:
but if performance is not up to snuff, then your efficiency gives you a very modest profit on the midline and nothing on the bottom end, just like now


Yes, but an interesting fact in IT spin, is that that performance measure is always taken from top of the lines and not from the bottom.

AMD as of yet dont have a quad core in the market to compete, period. So lets see about price/watt per performance points.

Yes the core 2 is a superior design than the K8, but if you were able to compare 2 CPUs with equal cache size at the same clock speed i believe that Intel would win in integer performance for only say 10%, and be almost equal in most floating point benchs, because AMD hypertransport platform is superior.

* that would take a C2duo with *1M L2* at 2,4 to 3,0 Ghz to compare with brisbane *

http://www.hardocp.com/article.html?art=MTIzMywxMi...

quote:
...new architecture from Intel again, which initially looks to eliminate the last of Amd's technical superiorities, like IMC


I believe that was already addressed in c2duo and the tradeoff is, for having an IMC than the L2 cant be so big.

I believe that HT3.0, Direct connect 2.0, will be superior to CSI when CSI arrives.

Second i dont believe that the intension of FUSION is to put a full GPU integrated with the CPU, but to make an instant on, double, triple, quad, penta... crossfire(in asymetric form). Since the CPU will have GPU function it can be like any other GPU... meaning booting without a GPU graphics card or a software layer framebuffer(for servers), and 'instant on' crossfire.

Expect CrossFire 2.0 to be nothing much more than Direct Connect 2.0 with some drivers and a special HT3.? for connection with HTX slots. Since also the GPU will have CPU functions for floating point and SIMD/MIMD, expect those superior design schematics showing a 4 SMP all fully connected around HT and DC(for servers), to show also 2 CPUs and 2 GPUs instead.

Yes, i believe that AMD will end up abandoning PCIe slots in their chipsets for graphics, and adopt HTX slots instead.

One of the few great advantages of AMD will still be their platform. Intel was/is foolish by not adopting hypertransport. Transporting for the present time if they did, the C2duo on HT will have absolutely killed AMD.


RE: wa
By mmarq on 7/11/2007 1:38:03 PM , Rating: 2
A last detail more directly related with this topic...

I believe that was one of the major reasons AMD bought ATI. Otherwise Intel and market pressure would make no good GPU 'manufacturers' to design and make graphic cards for the HTX slot. Co-processores are already here for HTX slots, but no good graphic card will be anytime soon if AMD dont take that matter in their own hands.


RE: wa
By zsdersw on 7/11/2007 1:44:46 PM , Rating: 2
quote:
Yes the core 2 is a superior design than the K8, but if you were able to compare 2 CPUs with equal cache size at the same clock speed i believe that Intel would win in integer performance for only say 10%, and be almost equal in most floating point benchs, because AMD hypertransport platform is superior.


HyperTransport is "superior" only in multi-socket systems, and even then, it earns the "superior" adjective only in systems above 2 sockets.

quote:
I believe that HT3.0, Direct connect 2.0, will be superior to CSI when CSI arrives.


Based on what? Your AMD fanaticism?


RE: wa
By mmarq on 7/11/2007 4:08:36 PM , Rating: 2
quote:
Based on what? Your AMD fanaticism?


Nothing of the kind. Only because HT, DC are around for 5 years, and is acknowledge superior design to shared buses and not only in servers.

Too much spin, no real numbers...
http://www.eetimes.com/showArticle.jhtml;?articleI...
http://interconnects.blogspot.com/2007/06/nudging-...
http://www.eetimes.com/news/semi/showArticle.jhtml...
http://www.eetimes.com/news/latest/showArticle.jht...

... i dont buy it, not because is Intel but because i despise brainwashers. So implying otherwise without numbers cannot be called Intel fanatism, yes ?


RE: wa
By zsdersw on 7/11/2007 7:14:18 PM , Rating: 2
Lots of things have been around for 5 years.. that are far from immune to being leap-frogged by the competition. HyperTransport's length of time in the marketplace does not mean that any potential competitor cannot possibly be as good or better. There is no basis for you to conclude that CSI won't be at least on par with the rendition of HyperTransport available at the time of CSI's arrival.

And yes.. only in 4S+ servers and workstations is HyperTransport a truly soar-away superior architecture. In everything else (laptops, desktops, 1S and 2S workstations/servers) there is plenty of competitiveness from Intel to muddy the water on HyperTransport's supposed "superiority".


RE: wa
By mmarq on 7/13/2007 2:38:38 PM , Rating: 2
quote:
Lots of things have been around for 5 years.. that are far from immune to being leap-frogged by the competition.


Yes nothing is immune to leap-frogging in the IT industry. But where are the numbers about latency, bandwidth that shows that at least Intel intents to do it with CSI ? ( a link please)

quote:
And yes.. only in 4S+ servers and workstations is HyperTransport a truly soar-away superior architecture.


NO. Actually HT1 sucks a little at 4 way, and is no improvement at all at 8 way because of bus contention. It doesn't have enough bandwidth to deal with the cache coherency protocol of 4 CPUS all connected to each other much less 8. HT3 has the potential of more than the double the bandwidth and with less latency. It can be pushed to more than 25Gb/s with 16 lanes ~12,5Gb/s in each direction(in and out).

C2D shared bus at 1066Mhz can punch 8,5Gb/s aggregate in all directions. DDR2 at the same speed can punch ~14Gb/s (12,8 at 800Mhz) one direction at a time. Talking about huge waste of memory bandwidth(8,5-14=-5,5GB/s) and bus contention .

HT1 can punch around ~12Gb/s with 16 lanes or ~6Gb/s per direction. But here stops any comparativeness because of the IMC, HT has to pull/push comparatively much less data trough the FSB. A huge benefice IMO

Penryn shared bus at 1333Mhz can only punch 10,6Gb/s aggregate in all directions. DDR3 at the same speed can punch ~24GB/s. Even a more huge waste of memory bandwidth(10,6-24=-13,4Gb/s) and higher bus contention.

HT3 can match DDR3 at 1333MHz, with no theoretical bus contention. And it only has more comparativeness because if you think of SLI or Crossfire, than AMD has a huge advantage.

And that is why Intel move to DDR3 is *PURE MARKETING BS*. They are on the catch up and losing. Can CSI change this to the point of being better than HT3 ?... It would be a huge leap-frogging, and i doubt it very much.

http://techreport.com/reviews/2007q2/core2duo-e675...


RE: wa
By TSS on 7/11/2007 1:49:25 PM , Rating: 2
i'd think that intel doesn't want to completly kill off AMD since beeing a monopolist isn't exactly benificiairy at present, definatly not with the EC lingering about. it'd be hard to imagine, in any case, that both intel's and AMD's strategy's for the long term are anything but complicated.

there is the problem of cache sizes though, and more importantly what it does for intel. if the cache's are equal, then the chips would perform similar. it might even happen that the core2duo would perform worse then the good ol' AMD64. but their not, and intel has of yet crammed alot of cache into their chips simply because they can with the scale of chips their on. the point beeing however that once the IMC is used by intel aswell the cache sizes can decrease dramaticly (the reason why amd doesnt really function alot better with more cache) and more area can be used for say computing power or it can be left off to increase profits. or, keep the profit the same and sell at a lower price.

now i dont have much besides logic to help me out here, so i could very well be wrong. i don't have a clue how much space a IMC takes up on the die either. but i'd figure it's less then the ridicolous amount of cache intel is adding. in any case, the way a IMC affects AMD may be different from what we think.

the decicive battle won't be fought this year in any case. if i'd have to guess which point in time will be critical for either compagny, it'll be what AMD's move is going to be after barcelona. once barcelona's up and running and the refresh comes about to take on nehalem, it'll be more like it is now. but if the next CPU that needs to take on the nehalem refresh dissapoints, and this is just a matter of *hom much*, AMD might be done for.

but i do belive then it will be because of their own fault, not intel's. AMD has shown more then once in the past how to improve performance by creative thinking, and i'd like to think they can do it again. but you can't blame the competition if you lost by lack of innovation.


RE: wa
By Andrwken on 7/11/2007 12:52:31 AM , Rating: 3
quote:

A 4 core MCM like Penryn at 45nm process will be 107mm2 x2 = 214mm2

Extrapolating a 4 core Barcelona at 45nm process will be
283mm2 x 0,59 = 167mm2

so:
A 2 core Penryn at 45nm process will be 107mm2

a 2 core Phenom at 45nm process will be 83,5mm2


Also, you are comparing 45nm Amd chips to 45nm Intel chips that will never compete directly against each other, Intel will have the Nehalem architecture out before Amd ever shows a 45nm chip.

Not trying to come off too pro-Intel here, but this post has some flaws in its logic.


RE: wa
By CyborgTMT on 7/11/2007 2:03:56 AM , Rating: 1
Both the Nehalem and the Deneb are scheduled for a 2H 2008 release.


RE: wa
By Andrwken on 7/11/2007 2:21:12 AM , Rating: 4
If recent history proves anything, it would be very suprising to see a 45nm deneb in retail by the end of 2008. Don't get me wrong, nothing is saying that intel will have another barn burner with Nehalem, but in the last two years, their timetables have been spot on or early. You just can't say that for Amd as of late.


RE: wa
By mmarq on 7/11/2007 2:11:09 AM , Rating: 3
Nehalem ?

Obviously not this one
http://www.theinquirer.net/default.aspx?article=74...

And if is this information, than how came can anyone call this interesting news!!
http://www.xbitlabs.com/news/cpu/display/200604281...

Ah! the same old good spin, so typical of the IT industry, noise and lot of 'names', but very few technical details.

So
Penrym is basically a die shrink with 2MB more L2 cache

Nehalem would be a somehow enhanced design on a CSI interconnect.

Interesting times 'we' have ahead. But Nehalem will be 2 core native right ? How would you expect a 4 core 2 chip MCM to compete with a 8 core 2 chip MCM shangai. ?

For one thing i hope that the press wont start posting that that is a unfair match 'apples to oranges', because i'm very interested in the maximum performance benchmarks of those beasts.

Than there is ZRAM. If this paper is correct
http://www.hotchips.org/archives/hc18/2_Mon/HC18.S...
can represent for the expected 6Mb L3 cache of the K10.5(?) Shangai, a maximum size of 36Mb (6x) in exactly the same die area and with about 50% less power consumption, and that in spite of the eDRAM being slower than SDRAM could represent a huge performance gain equivalent to another die shrink!...

Well... we'll see!


RE: wa
By Andrwken on 7/11/2007 2:34:04 AM , Rating: 3
No, that nehalem on that 2003 inquirer link was scrapped for the Core architecture. The new one looks to be a native quad that they will mcm to eight core with hyperthreading on each core. 16 simultaneous threads. Like we need that by next year, lol. I would guess that is what Amd's direct competition to their 45nm will be in 08. The penryn will be competition to the 65nm Barcy's.

Here's a link for you :)
http://www.hothardware.com/articles/Intel_Penryn_a...


RE: wa
By mmarq on 7/11/2007 3:41:19 PM , Rating: 2
Very interesting link thanks...

So it seems i was wrong in my suspicions and Penryn is more than a die shrink, capable of equal or better performance than Barcelona at the same clock speed, and able to achieve much higher clocks... AMD is in trouble if they cannot scale the clock of their Barcelona.

(hmm... IBM managed to get pass 5.0 Ghz(so they claim) in a bigger chip (Power 6), in the same almost exact 65nm process of AMD, why cant they get somehow much closer to that ?)

quote:
The new one looks to be a native quad that they will mcm to eight core with hyperthreading on each core. 16 simultaneous threads


So it seems that its a native quad core.

# Scalable performance: 1 to 16+ threads, 1 to 8+ cores, scalable cache sizes
(very good, and with L2/L3 too... hardly to tell how it compares to decoupled, clustered, speculative multithreading of reverse-hyperthreading rumors and other interviews and slide clues of AMD... if they are true... :
http://citeseer.ist.psu.edu/rd/69795829%2C57813%2C...
http://citeseer.ist.psu.edu/rd/69795829%2C445663%2...
http://citeseer.ist.psu.edu/rd/69795829%2C306699%2...
http://citeseer.ist.psu.edu/rd/69795829%2C227934%2...
http://citeseer.ist.psu.edu/rd/69795829%2C659643%2...
Universitat Politècnica de Catalunya, Barcelona... hmmm... speculative anyway)

# Scalable and configurable system interconnects and integrated memory controllers
(doubt CSI could take on HT3.0, Direct Conect 2.0 at first time)

# High performance integrated graphics engine for client
( Intel dont have hypertransport neither a crossfire like scheme... would they buy Nvidia for not being inferior to FUSION? )

Remarkable year 2008 will be... and if AMD dosent execute much faster than today, and i hope they will, we will have an absolute monopolist... and i would not like to buy my next 8 core CPU at 2000$... thats the only reason not an anti-Intel sentiment.


RE: wa
By mars777 on 7/15/2007 7:12:30 AM , Rating: 2
quote:
(hmm... IBM managed to get pass 5.0 Ghz(so they claim) in a bigger chip (Power 6), in the same almost exact 65nm process of AMD, why cant they get somehow much closer to that ?)


Because the Power architecture is in-order vs the out-of-order of modern multipurpose cpus made by Intel and AMD.

THis makes them more than 20% slower clock-per-clock basis on few-thread operations, which they hide by using some form of hyperthreading.

Anyhow they are not targeted to me and you. Simple as that :)


RE: wa
By vignyan on 7/11/2007 2:46:45 AM , Rating: 2
Dude.. you are going paranoid with your AMD fever. Anyways, correcting all the flaws with your posts in this thread....
First.. The retooling to 65nm and 300mm wafer sizes cost AMD a 2bn hole. and that too has to come back to them from the selling the processors for a profit.
Penryn is not just a die shrink plus a 2MB addition to Core2 architecture... it has advaced SSE4 instructions that help in media encoding and tech computing performace... Also the ATA(Application Targetted Accelerators) instructions in the SSE4... thats not just it.. with all this it has a 25% lower power consumption than the best core2 today.
Another thing is that Nehalem is an architecture... like Core and Core2... not to signify the number of cores. Get ready to be suprised by the way nehalem is going to be introduced... Scalability is the key. You can configure from 1-8+ cores, scalable cache sizes, IMCs (yes IMC s ), system interconnects, integrated graphics engine...

Be Amazed... and SEE...

PS: Yes... i am an Intel fanboy... but only because of the technology that they are innovating and sharing... Used to be AMD's... but now am Intel! ;)


RE: wa
By SmokeRngs on 7/11/2007 10:35:46 AM , Rating: 2
quote:
it has advaced SSE4 instructions that help in media encoding and tech computing performace


In my experience, it normally takes 2-3 years for a noticeable amount of software to incorporate optimizations which make use of one of the iterations of SSE. I haven't really looked recently, but I haven't seen much in the way of advertising of SSE3 optimizations.

Basically, optimizations such as this rarely show any improvement in performance until later down the road. AMD will have SSE4 long before there is widespread use of the instructions so it's unlikely there will be any actual performance gains at Penryn's release due to SSE4.

I'm interested to see how Penryn actually does. I applaud any efforts to increase the efficiency of my computing hardware.


RE: wa
By KnightProdigy on 7/11/2007 10:59:19 AM , Rating: 2
It's funny how AMD and ATI are heading up the on-die graphics arena. Intel revealed their idea for it only 6mo after AMD.
Intel with a better architecture?
LOL
Thats why they have been playing catch up to AMD since the K6 right?

Sure, as soon as Intel gains the speed crown or releases 'new' information, all the fanbois come out trying to make it sound like Intel 'had it' all along.
I used to be an Intel fanboi until they AXED the potential of the Copperhead FORCING their customers to buy the insanely overpriced higher end procs.

Your advanced SSE4 and ATA instructions are only to try to keep up with OLD 3D Now tech. Intel has always sucked on the media end of things and are STILL behind AMD in that market. Gotta look at the bigger picture here.

As far as the 25% less power consumption...true, but in the low power consumption arena, AMD still has them beat in performance/watt (I could be wrong but Penryn cannot even run full speed @ 45w or below).
The new interconnects are only to compete with HT3, and from the specs, it looks to me like even that has come up short. In the desktop and consumer market, analysts have stated that 8 cores will show little to no performance increase over 4 (similar to a RAID affect). After 4, the architecture and innovation needs to come to par to see a significant speed increase. Considering the AMD vs Intel release date information, AMD may have that covered, since they have scheduled a NUMBER of releases in the next year as Intel, well, not so many.

OP of this thread is correct. Do your homework before arguing with someone.
Guess we will have to wait and see.


RE: wa
By zsdersw on 7/11/2007 12:02:41 PM , Rating: 2
quote:
Thats why they have been playing catch up to AMD since the K6 right?


You couldn't be more wrong.

quote:
Intel has always sucked on the media end of things


Again, you couldn't be more wrong. Even in the P4 days, Intel was either ahead or very competitive in media encoding. It was a common bright spot in the benchmarks for the P4. Core 2 didn't give anything back to AMD in media encoding (or anything else), either.


RE: wa
By mmarq on 7/11/2007 4:39:56 PM , Rating: 1
quote:
Thats why they have been playing catch up to AMD since the K6 right?

You couldn't be more wrong.

Well since the K6 no, but since the K7 and K8 yes. And AMD can blame itself, and the huge influence of Intel upon the spin bullshitting press for not being able to dent a much much larger dent on Intel market share.


RE: wa
By zsdersw on 7/12/2007 12:04:07 PM , Rating: 2
quote:
but since the K7 and K8 yes.


Incorrect. Core 2 has places Intel in the lead. Now it's AMD's turn to catch up.


RE: wa
By zsdersw on 7/12/2007 12:33:17 PM , Rating: 2
has *placed* Intel in the lead, I meant to type.


RE: wa
By mmarq on 7/11/2007 4:23:31 PM , Rating: 2
Dont take me wrong

But what does an Intel fanboy doing working as an accountant for AMD!... (The retooling to 65nm and 300mm wafer sizes cost AMD a 2bn hole )

Much of the problem is not 'funboyism', anyone has preferences, but is the BS spin that floats around by the ton.


RE: wa
By defter on 7/11/2007 2:36:17 AM , Rating: 2
You forgot that scaling isn't linear and dual core Phenom with L3 cache will have about >60% of the Barcelona's die size.

For 65nm, die sizes are about following:
4MB L2 Conroe: 143mm^2
2MB L2 Allendale: 112mm^2
2x512KB L2 Brisbane?: 125mm^2


RE: wa
By mmarq on 7/11/2007 5:25:26 PM , Rating: 2
No, i was not taking linear approaches. I was taking as measure, factores taken from real implementations(penrym), not the hipotectycal ideal conditions... so the site implies. Nevertheless the error margin should be much less than 10%.


RE: wa
By mmarq on 7/11/2007 6:48:45 PM , Rating: 2
Good point. So the numbers are

At the 65nm process.

4MB L2 Conroe: 143mm^2

3MB L2/L3 Phenom: ((283/2)+11,2) = 152,7mm2 =>(11,2mm2/MB for L3)

2MB L2 Allendale: (143-(2x12,4) = 118,2mm2 =>(12,4mm2/MB for L2)

1Mb L2 Runa: ((283/2)-(11,2x2) = 119,1mm2

Yes here AMD will have to compete on performance. But since the Barcelona implementation is so cache starved, IMO, we can expect that that 1Mb additional L3 makes a lot of difference from a linear extrapolation, and Phenom X2 will be able to easily compete with Conroe, but with Penryn that will be another story unless clocks go much higher. Can Runa compete with Allendale in performance? we'll have to wait and see the benchs but my idea is yes but no with the 45nm equivalent.

I've scrap Brisbane because according to his roadmap http://www.vr-zone.com/?i=5078 there is a hole and nevertheless it will only be for 45W or 38W parts centered around barbone and HTPC where GPU/video codec acceleration is more important.

Strange that their 45nm seems to have a much better theoretical scalability. How complicated tradeoffs they had to have done between cache, clock and functions, and how that stall them.

Bottom line is that they have a better design than core 2, perhaps even better than the Penryn enhanced one. How they execute and how it scales its the big difference.

We'll see.


RE: wa
By Spivonious on 7/12/2007 9:43:06 AM , Rating: 2
I'm not sure why you quoted me...perhaps because you have strong feelings towards AMD and truly believe in the future success of Barcelona, therefore causing AMD to not have another two years like they've already had.

From scanning through your post it seems you're saying that Barcelona will compete in the market because it's smaller and therefore cheaper to produce.

The problem is that you're comparing an 8-core version of Barcelona with an 8-core Penryn. By the time AMD has a competing processor, Wolfdale will be out and AMD will be behind in the performance game again.

Customers who only see price won't be buying a $300 Barcelona anyway. They'll be picking up the remaining stock of Athlon X-2 chips for <$40. I hate to say it, since competition is good for the market, but AMD is on the way out. And most satisfying is that Intel did it by creating a better product, not simply pricing AMD out of business.


RE: wa
By mmarq on 7/13/2007 3:13:37 PM , Rating: 2
quote:
The problem is that you're comparing an 8-core version of Barcelona with an 8-core Penryn. By the time AMD has a competing processor, Wolfdale will be out and AMD will be behind in the performance game again.


I don't want to trow ice water but that seems a little hopeful. Yes Intel has always executed much faster, if for nothing else because they have much more resources but i believe if AMD really wanted bad they could have a MCM with 8 cores before the end of this year.

A 566mm2 monster with more than 130W TDP. Yes we'll be talking about perhaps a 2000$ chip, but concerning TDP Intel would still win with its 3.0 Ghz Xeon X5365 with 150W TDP. I'm talking of lack of decency!...

http://techreport.com/reviews/2007q2/intel-v8/inde...

If in drop_in_place parts with many of actual Opterons systems, it could prove to be a huge sell success, because a 4P 2 way server could be converted to a 16P 2 way with very little modification. And the press spin they could generate with that !... if they drop for the Bench Industry a fascn8 workstation !...

In that sense i agree with you. Because if it were Intel in the same position of AMD, the first thing they would put out now in August would be that monster.


RE: wa
By jdun on 7/11/2007 1:59:51 AM , Rating: 2
If they really have 1.1 billion left in the bank then they will go bankrupt next quarter. AMD interest payment is $100 million a month, I believe. In order to stay afloat they must have a minimum of $600 millions in the bank to make payroll and interest. AMD is not forecast to make a profit for this year. In fact it is expected they will lose 2 billion by the end of this year.

Good thing I shorted this company.


RE: wa
By halbhh on 7/11/2007 2:13:19 PM , Rating: 2
amatuers should not short stocks, especially not stocks like this one


RE: wa
By OrSin on 7/10/2007 3:36:51 PM , Rating: 3
Can we stock options. In a merge like this The company being bought give hugh stock options if the old case stays for period of time. I bet he walks away with 50-100 million in his pocket.


RE: wa
By Desslok on 7/10/2007 3:04:26 PM , Rating: 2
Wasn't there an article back before the Athlon came out that said several accounting firms were keeping a daily tally on AMD's cash outlays to see when they would go under?


RE: wa
By webdawg77 on 7/10/2007 3:21:15 PM , Rating: 3
I think he meant the word only as in sarcasm guys! Don't always take things too literally :).


RE: wa
By ebakke on 7/10/2007 5:20:17 PM , Rating: 1
He most certainly did not.


RE: wa
By MartinT on 7/10/2007 4:13:22 PM , Rating: 1
It is wrong. And the culprit is the $2.2 billion convertible bond offering AMD has completed since their last financial report at the end of Q1. In the end, it was supposed to add about $1.6 billion to the cash available, but as AMD has surely lost a couple of hundred million dollars in Q2, too, I'd say it'd be best to wait for next Thursday's Q2 report for an update on AMD's cash situation.

If I had to guess, I'd put it around $2.5 billion.


"If you look at the last five years, if you look at what major innovations have occurred in computing technology, every single one of them came from AMD. Not a single innovation came from Intel." -- AMD CEO Hector Ruiz in 2007

Related Articles
AMD's Large Debt Mounting Up Fast
April 29, 2007, 1:54 PM
AMD Posts $611 Million Net Loss in Q1
April 19, 2007, 9:13 AM
AMD-ATI: A Done Deal
July 24, 2006, 5:00 AM













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki