AMD's K10 Dead - Intel's Big Opportunity

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
I found a new article.

http://www.theinquirer.net/?article=27421

It seems that AMD's next-generation K10 architecture is dead. This is actually kind of funny really since everyone has been pointing fingers at Intel for delaying the introduction of Xeons with integrated memory controllers. Now it seems AMD is facing problems as well.

Without the K10, the AMD64 will definitely lag behind Intel's next-generation Conroe, Meron, and Woodcrest architecture. Already a Dothan at 2.56GHz Dothan is able to beat AMD's FX-55 and Intel's own Pentium 4 Extreme Editions in all 3d games as well as in most other benchmarks. It also did so while running at 58% lower power comsumption and the FX-55. And this is only a Dothan. Yonah itself will offer considerable improvement over Dothan with its improvements to multimedia and FPU performance. One can only imagine Conroe's potential.

To meet Intel's new architecture, AMD will only have a stop gap measure called the K8L. With the K10 or its replacement delayed until 2008, Intel will have an excellent opportunity to dominate AMD and regain market share over the next 3 years. Hopefully, the execs at Intel perk up and don't squander it.

The Dothan figures are from this article:
http://www.tomshardware.com/cpu/20050525/pentium4-21.html
 

Nights_L

Distinguished
Jan 25, 2003
1,452
0
19,280
Yeah!! let's see how Intel becomes the processor king again and we'll be paying twice amount of $$$ to buy an Intel Inside Processor! :eek:
<P ID="edit"><FONT SIZE=-1><EM>Edited by Nights_L on 11/03/05 11:10 PM.</EM></FONT></P>
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
Actually Intel is doing pretty well on its own pushing down processor prices. Just look at the Pentium 820. Granted the performance is not on par with say the X2 3800+ but the 820 is 31% cheaper. You can get the dual core 820 for only $245.19. The value of this already low priced processor will only increase as more dual threaded programs become available.

AMD on the other hand actually broke the record for the highest release price for a desktop consumer processor. When the FX-57 was released, AMD priced it at $1031. This went against its previous policy of underpricing against Intel's $999 Extreme Editions.
 

Kelledin

Distinguished
Mar 1, 2001
2,183
0
19,780
Without the K10, the AMD64 will definitely lag behind Intel's next-generation Conroe, Meron, and Woodcrest architecture. Already a Dothan at 2.56GHz Dothan is able to beat AMD's FX-55 and Intel's own Pentium 4 Extreme Editions in all 3d games as well as in most other benchmarks. It also did so while running at 58% lower power comsumption and the FX-55. And this is only a Dothan. Yonah itself will offer considerable improvement over Dothan with its improvements to multimedia and FPU performance. One can only imagine Conroe's potential.
Actually, taking Intel's own cherry-picked benchmarks and extrapolating based on the (very generous) assumption that Yonah's performance will scale perfectly with clockspeed, a future dual-core Yonah part at 2GHz falls a hair short of the X2 3800+--the lowest-end dual-core part AMD has <b>today.</b> We can probably expect much the same (perhaps worse) with Sossaman. The power consumption is impressive, but that's about it.

Not to mention which, Yonah's not really scheduled to scale much in clockspeed anytime soon. It will release at a clockspeed <i>somewhat</i> faster than current Dothans--2.13GHz IIRC--but that's hardly an impressive leap. Will it even have 64-bit extensions?

Conroe and Merom might get the performance crown back for Intel. Yonah probably won't though.

"You have been sh<font color=black>it</font color=black> upon by a grue."
 

rettihSlluB

Distinguished
Jun 5, 2005
296
0
18,780
Don't make a fool of your self.

This is what a member from the amdzone forums has to say about all this stuff:

None of this information is particularly surprising or disheartening to me. I had expected K10 to be 4 core and it appears that a 4 core design will be released. I expect this to be done on 65nm which is still a full process generation ahead of Intel. That is, I don't expect Intel to do 4 core until it can produce 45nm.

I don't see anything tragic here. K8 is working fine and AMD will make substantial additions to its architecture. These additions take time and resources. As long as these offerings remain competitive there is nothing wrong with moving K10 back to 2008. Merom is not a threat. Intel won't have a solid offering until Nehalem in late 2007. So, K10 should arrive at about the right time. Okay, look at this another way. Intel was supposed to deliver a very threatening product in the form of Whitefield. Whitefield was canceled. I imagine that this took pressure off of AMD and so they decided that a K8 derivative would be good enough to meet the challenges. I'm sure K10 is still alive in some form and will be out in 2008.





WOULD YOU LIKE TO BE A WELL FED SLAVE OR A HUNGRY FREE MAN?

<P ID="edit"><FONT SIZE=-1><EM>Edited by Bullshitter on 11/03/05 11:43 PM.</EM></FONT></P>
 

Era

Distinguished
Apr 27, 2001
505
0
18,980
ltcommander_data is a BS generator, so why bother?
There are people who can memorize lots of things, but understanding little.
You give one a banana, he will go at his keyboard and starts tapping it happy as a cucumber.
 

Atolsammeek

Distinguished
Dec 31, 2007
1,112
0
19,280
What funny is theinquirer Has not much info on this subject. Yet has tons of info on Intel Faults.

Hum Intel 5 to 6 Problems due to high power useage. To Heat and netburst.

vers

Amd

Hardly any info on any problem.

I think the main problem is ltcommander_data Who dont like the fact Amd better then Intel.
 

josh_1413

Distinguished
Jun 7, 2005
38
0
18,530
I agree 100%. Why rush? As if Intel has anything in the next year or two. They can just spend more and more time working on the next generation processor.
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
I take your point and I'll use THG's reviews with a grain of salt. However, in this case THG isn't alone. Every other overclocking review shows the exact same result. Namely the Pentium M architecture present in Dothan shows tremendous performance when overclocked.

Anandtech got the Dothan up to 2.56GHz.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=10

At that speed, the Pentium M is at the top or near the top of application and game benchmarks. The only weakness was in media encoding tasks. However, they do note that Yonah will correct those problems.

"All of the architectural improvements, outside of the move to dual core, involved SSE and floating point performance - the two major weak points of the Pentium M's present day desktop performance."

These results were likewise found by PC Perspective.

http://www.pcper.com/article.php?aid=133

"In nearly all cases, the Pentium M easily rivals the top end Athlon 64 and Pentium 4 processors, and in some cases beats them."

With Yonah fixing many of the problems of Dothan, the ability for Intel to generate a extremely high performance processor is there. The only limiting factor is that Yonah is suppose to be mobile based and as such it will be limited in its released clock speeds to save power and meet a 31W envelope. However, this won't stop Yonah from dominating the mobile market.

The real potential of Pentium M and Yonah can only be discovered in the desktop market, when the 31W power envelop is lossened. In regards to this, I believe that The Inquirer had an article a while back about the possibility of a Yonah Extreme Edition. If the Dothan can meet or AMD's FX's at the same clock speeds, Yonah should further expand that potential.

However, I do see Kelledin's point. Intel will probably never release the Yonah to the desktop market at higher desktop speeds. We will instead have to wait on Conroe and Merom which should offer still higher performance due to an expanded architecture. Merom and Conroe will double the L2 cache to 4MB, and a Conroe Extreme Edition will have 8MB. As well, the belt will be loosened for Conroe which will have a 1066MHz FSB to ensure the processor is saturated.

What I find most interesting is the fact that Dothan can reach its high performance with so few execution units. It only has 1 Complex Integer Unit, 1 Simple Integer Unit, and 1 FPU yet it can match the performance of AMD's K8 architecture which has 3 full (Complex) Integer Units and 3 full FPUs. This means that Intel is able to match AMD's performance with less than half the execution units. This hints at the performance potential that Intel could build into Conroe, Merom and Woodcrest should bring. I believe that current conservative estimates are that Conroe will hae at least 2 Full Complex Integer units, instead of the 1 Complex, and 1 Simple. The number of FPUs will double to 2 from only 1. Yonah is already adding full support for SSE2 to fix the shaddy SSE2 additions that were originally built into the Pentium M. Conroe will of course add support for SSE3. Execution performance will be further improved as the use of Micro-ops fusion is expanded.

Any performance gains will of course be matched with reductions in power and heat. Yonah and Merom of course have extrodinary power saving procedures, and Conroe and Woodcrest will no doubt be receiving most of those features as well. Speedstep will obviously be standard, but I think that the most critical thing will be the 65nm process's addition of sleeping transistors. In Yonah and Merom, sleeping transistors ensure that the units within the processor only switch on when needed. When the processor is idle, most of it would be shut down producing little power. On Conroe and Woodcrest, Intel will probably just inverse the process. Keep the units on by default to avoid the slight wake-up latency, but still allow the ability to shut of units when not in use. Slight reductions in power and heat would even be seen when the processor is at 100% load as even under full load, not all the transistors in the processor are in use at once and can thus be shut down.

I seriously doubt josh_1413's claims that "Intel has anything in the next year or two." I believe that he will be presently surprised with Conroe and family.
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
It' funny how bringing up AMD's problems makes one a fool while someone writing on the AMDZONE forums makes them correct. Or how presenting facts makes one a "BS generator" as Era so wonderfully puts it.

And for AtolSammeek, I noticed the fact that The Inquirer limits its criticism on AMD while focusing on Intel. However, its not that I don't like AMD. In fact its great that AMD has shaken up the market. What I don't like is exaggeration of things that people seem to enjoy inflicting on Intel.

For instance in Bullshitter's quote from the AMDZONE Forum. What do they mean by AMD's "65nm which is still a full process generation ahead of Intel"? It seems they mean that AMD"s 65nm process will allow them to make quad cores while Intel will have to wait for 45nm. This is far from the truth. In fact, a quad core based on Intel's next-generation Woodcrest is already in the works. It's called Woodcrest and is scheduled for release in Q4 2006. It will use the 65nm process. This has always been the case and has nothing to do with the Whitefield delays. Whitefield is supposed to be Cloverton's replacement in late 2007-2008.

For the comment that "Merom is not a threat." First of all, Merom is scheduled for H2 2006 and won't be competing with the K10 on its now delayed release in 2008. Secondly it is a mobile processor and shouldn't be lumped in performancewise with Whitefield, high-end server processor. I've already outlined the advantages of Merom and its architecture in my post above. Simply put Dothan already meets or beats AMD's K8 based FX at the same clock speeds, Merom and its family should further improve. Besides, the delay of Whitefield does not remove its threat. Tigerton, Whitefield's replacement is based on the same architecture. The difference is the absence of an integrated memory controller.

As well, Intel still has a trump card up its sleave. Up to now we've been talking architecture, while process benefits will also become significant. Intel's 45nm process will be a force to be reckon with. intel has finally solved the leakage problem. This means extremely efficient processors in terms of power consumption and heat production. It also means that large clock ramps can be made without concern for thermal density, since thermals would be very low.

According to The Inquirer,
"So, if you hear gushingly good things about 45nm coming from IDF, believe it. If you hear anyone pooh-poohing Intel and its process tech because of the debacle that was 90nm, just point and laugh. This one will be very very good."

And thats not Intel marketing talk, thats The Inquirer gushing. The same Inquirer that AtolSammeek saids usually has "has tons of info on Intel Faults."

What is even more significant is that this process is scheduled for an early 2007 release. K10 or its replacement of course won't be available until 2008 at least.

http://www.theinquirer.net/?article=25512

Another article backs up The Inquirer.

http://www.neoseeker.com/news/story/4896/
 

endyen

Splendid
Where to begin? That August article is aluding to fully depleated silicon on insulator. That was developed before prescott, was originally slated for use on 65 nanos, but has been pushed back to 45.
Yes, the Dothans have looked promising, we've been talking about Intel shifting thier chips in that dirrection for almost 2 years now.. There is some concern about how high they can actually go. The concensis is that Intel will not be able to market them above 2.8ghz, which just isn't fast enough.
You understand that we are a little jaded, because Intel has just not been able to come up with the goods for four or five years (aside from northwoodc which was pretty good). It seems like most of us came from Missouri, even though we've never been there.
 

pat

Expert
Ok, by this time, the H5N1 deadly influenza virus might have kill a lot of peoples. Then maybe terrorist attck would have stike on major part of this planet. You may die tomorrow in a car accident. Or being killed walking on the side of the road. You may paraplegic, in a team sport. No chance to recover. Or recover and enjoying life and not using you computer as often.

What about getting a life? Because, I seriously, don't care at [-peep-] all who will make the CPU that will power my computer in my next upgrade. I won't have choice. I'll take the one that has the best price/performance ratio. Hey, even if I'd like to, I simply can do nothing about that. Nor you..

So, looks at what is happening now. And would anybody would have think about having dual core, when P4 running at almost 10GHz were expected thank to netburst architecture?


Ahhh those peoples always looking to the future are making me sick! Do you realize that your are wasting your present time?

Now, I've wasted enough in this thread.

(\__/)
(=<b>˚</b>.'=)That was my bunny..terminator!!! Now terminated...
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
I can understand people's frustration with Intel. That's why I don't dislike AMD, they've been doing everyone a service by offering stiff competition and driving the market forward. I just try to remain optimistic about Intel, and from what I've seen of Intel's next-generation architecture, there is enough potential there to make me think my optimism is justified.

About the clocking, while its true that Dothan could only clock to 2.8Ghz using standard air cooling, I think the next generation architecture could do a bit better. Conroe and family has their pipeline lengthened an additional 2 steps to 14 steps so it should offer a bit more clocking headroom on the 65nm process. It is also my understanding that the power and heat reductions in the 45nm process available early 2007 should allow further headroom. I hope I'm correct on that belief. Intel is rushing toward a transition to 45nm despite the fact that 65nm is not even out yet. I believe the Merom will be the first to be replaced in 45nm. Merom will only be around about 6 months, then Penryn will be in 2007.

Even without significant clock speed increases, architechtural improvements can continue making performance increases. As I've mentioned Dothan can already match AMD's FX processors performancewise at the same clock speeds with half as many execution units, and with less temperature. With the extra room on the 65nm process and the 45nm process, Intel could easily increase its execution unit count to increase computing power.

I think that its important to note that the use of an integrated memory controller isn't as significant as it appears. When comparing to the Pentium 4 it obviously is, but Dothan is able to match the K8 using only an FSB. And the Dothan doesn't even have an 800MHz FSB available like the Pentium 4 does. Yonah is set to increase the FSB from 533MHz to 667Mhz, and with up to 1333MHz FSB available, the lack of an integrated memory controller shouldn't be a strong limiting factor on the Pentium M and Conroe architecture.

If someone can show me benchmarks that show the K8 consistantly outperforming Dothan at the same clockspeeds on a variety of activites I might be more conservative in my optimism for Intel. Of course, even if K8 can slightly outperform the Dothan at present, Yonah will be coming out in 2 months to correct many of Dothan's deficiencies.
 

rettihSlluB

Distinguished
Jun 5, 2005
296
0
18,780
What I find most interesting is the fact that Dothan can reach its high performance with so few execution units. It only has 1 Complex Integer Unit, 1 Simple Integer Unit, and 1 FPU yet it can match the performance of AMD's K8 architecture which has 3 full (Complex) Integer Units and 3 full FPUs. This means that Intel is able to match AMD's performance with less than half the execution units.
The real reason why the pentium M is so competitive against K8 it's because of it's huge low-latency L2 cache which really helps that type of processor architecture.

K8 can rape the pentium M in multimedia applications and even in games that are not cache-dependant thanks to it's powerful floating point engine.

WOULD YOU LIKE TO BE A WELL FED SLAVE OR A HUNGRY FREE MAN?
 

rettihSlluB

Distinguished
Jun 5, 2005
296
0
18,780
If someone can show me benchmarks that show the K8 consistantly outperforming Dothan at the same clockspeeds on a variety of activites I might be more conservative in my optimism for Intel. Of course, even if K8 can slightly outperform the Dothan at present, Yonah will be coming out in 2 months to correct many of Dothan's deficiencies.
<A HREF="http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=1" target="_new">There you go.</A>

I hope I didn't burst your bubbles. :D

WOULD YOU LIKE TO BE A WELL FED SLAVE OR A HUNGRY FREE MAN?
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
Thankfully, low multimedia performance is specifically what Yonah was designed to address. Decoder width for SSE instructions has been tripled, Micro-Ops Fusion has been expanded to include SSE and SSE2 instructions, and SSE3 instructions have been added. Thankfully, Intel has also redone the SSE2 implementation, which was poorly incoporated to the Pentium M from the Pentium III. As well, despite critism of the added latency of shared caches, Yonah's 2MB of L2 cache will maintain the same low 10 cycle latency as Dothan. These improvements should make Yonah more competitive with the K8 until Merom arrives with additional and redone execution units.

http://www.anandtech.com/tradeshows/showdoc.aspx?i=2366

I have a question regarding a feature of the Conroe architecture. What is "memory disambiguation"? I think its a type of prefetching but is it a new feature then?
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
Em. If you turn to page 10 you'll see a Pentium M clocked at 2.56GHz is in the top 3 performing processors in 6 out of 8 charts. Out of that it is the fastest twice. Similarly in the Game benchmarks on page 11, the Pentium M is in the top 2 in both benchmarks. Being first in Doom 3 and second in Wolfenstein. That was what I meant by the Pentium M being able to perform equavalent in many cases to AMD's K8 since at 2.56GHz it performs similarly to the 2.6GHz FX-55. Of course it is weak at media encoding which I've replied to above.

Sadly, they didn't include a 533MHz FSB 2GHz Pentium M to compare to the 2GHz AMD64 3200+ although the 400MHz FSB Pentium M is pretty close to it in many cases. The 2.13Ghz Pentium M that they do include is clearly superior to the 3200+ even in media encoding, although thats mostly due to higher clock speed.
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
Why wusy, I believe that's the nicest comment you've ever said to me. It's certainly better than "He'll be raped by me before that happen, I assue you."

I of course understand waiting until Yonah is released. Thankfully, its due early January. Hopefully we see leaked reviews before that. It would either be a Christmas present for AMD or Intel, depending how it goes.
 

kenbosley

Distinguished
May 12, 2004
6
0
18,510
On the high end you are correct in terms of what is actually shipping today. However, if price/performance is the criteria the Celeron-M 370 is clearly the best processor today. I think he merely predicts that the high end will follow today's reality on the low end.
 

endyen

Splendid
true that Dothan could only clock to 2.8Ghz using standard air cooling
In your dreams. The dothan is hitting a brick wall @ 2.58 basicly. That is not a marketable speed, due to extra voltages, and non standard fsb. Dont expect it to come to market above 2.4. That is not a good enough number to compete with current A64s.
If they do add stages to conroe, expect the IPC to drop. Sure it may catch the A64s speedwise, but at lower perf/mhz. The old mhz race again.
I really dont get you. You remind me of a guy who talked his sister into buying a prescott. I asked if he understood what a bad move that was. He said well yea, but Intel will get better. You to seem ready to sacrifice your sister, in the hope that some day, your great god Intel will get better.
 

shawnlizzle =]

Distinguished
Feb 2, 2005
1,061
0
19,280
one more thing i would like to point out. you can't just compair a 2.5ghz dothan w/ a 2.4/6ghz k8. the speed(in ghz) is scaled between cpu to cpu. notice how much the dothan was overclocked just to compete w/ the k8s. scale the k8s the same clock speed up and they again dominate

<font color=red>gforce mx100/200 @ 230/440 =]</font color=red>
 

ltcommander_data

Distinguished
Dec 16, 2004
997
0
18,980
I'm just doing a comparison of the efficiency of the architectures. In such a case, I'm seeing what they can do at the same clock speeds. Thus, a 2.56GHz Dothan competes very similarly to a 2.6GHz AMD FX-55. Obviously its weaker in media encoding, but thats a weakness Yonah is to address.

http://techreport.com/etc/2005q3/idf/index.x?pg=2

Tech Report tells the enhancements to SSE/SSE2/SSE3 and FPU enhancements for media encoding.

If you don't like me OCing Dothan to 2.56GHz to compare to a 2.6GHz FX-55, you could take another example. In the anandtech report they didn't include a 533MHz FSB 2.0GHz Dothan. But even if you look at the 400MHz FSB 2.0GHz Dothan its results are slightly slower but close to the performance of AMD's 2.0GHz 3200+. Regardless of what speed you want to look at, Dothan compares similarly to K8 at the same clock speeds. Of course, Dothan is doing it with half the K8's execution units and without an integrated memory controller.