backtop


Print E-mail del.icio.us 147 comment(s) - last by Viditor.. on May 9 at 11:38 PM

AMD prepares its Phenom FX, Phenom X4 and Phenom X2 lineups for launch

AMD is preparing the launch of its next-generation K10-derived Stars-family single, dual and quad-core processors. The next-generation Stars-family splits into three different brand names – Phenom, Athlon and Sempron. Ringing in the flagship are three Agena FX-based AMD Phenom FX processors. AMD has yet to confirm clock speeds for the three models; however, the latest roadmap reveals ballparks for the processors.

The top-end AMD Phenom FX processor clocks in the 2.4-2.6 GHz speed range. Slotting below the top-end Phenom FX is a 2.2-2.4 GHz model. These two models occupy AMD’s upcoming Socket 1207+ and current Socket 1207 Quad FX platforms. AMD also has a Phenom FX for single-processor customers as well, clocked at 2.4-2.6 GHz.

AMD further differentiates its Phenom FX processors with different Hyper Transport 3.0 clock speeds. The flagship 2.4-2.6 GHz model features a 3.6 GHz HT 3.0 clock speed while the two 2.2-2.4 GHz models have a lower 3.2 GHz HT 3.0 clock. All three models share the same 4x512KB L2 cache and 2MB L3 cache configuration. AMD has yet to determine the TDP of its Phenom FX processors.

Catering towards high-end user are two Socket AM2+ AMD Phenom X4 processors. AMD remains undecided on its model numbers; however, clock speeds on the Agena-based Phenom X4 processors are set. The two AMD Phenom X4 processors clock in at 2.4 GHz and 2.2 GHz. These models share the same 4x512KB L2 cache and 2MB L3 cache configuration as the Phenom FX processors.

HT 3.0 speeds differ on the two models, the 2.4 GHz features a 3.6 GHz HT 3.0 speed while the 2.2 GHz model features a 3.2 GHz HT 3.0 speed. AMD rates the Phenom X4 processors with 89W TDPs. AMD plans to start taking orders for its Phenom FX and Phenom X4 processors in Q3’2007.

AMD’s dual-core Kuma processors will carry the Phenom X2 name and drop into Socket AM2+ and AM2 motherboards. There are three AMD Phenom X2 processors in the pipeline with 2.8 GHz, 2.6 GHz and 2.4 GHz clock speeds. HT 3.0 speeds vary on the AMD Phenom X2 processors. At the top with the AMD Phenom X2 2.8 GHz model. The HT 3.0 bus clocks in at 4.2 GHz – higher than the Phenom FX and Phenom X4 models. The two lower models have 3.8 GHz and 3.6 GHz HT 3.0 clock speeds, respectively.

All three models feature a 2x512KB L2 cache and 2MB L3 cache configuration. The higher end 2.8 GHz has an 89W TDP, similar to the Phenom X4 models, while the 2.6 GHz and 2.4 GHz receive a lower 65W TDP rating. AMD plans to take orders for its Phenom X2 processors beginning in Q4’2007.

If low power is a concern, AMD intends to offer three low power AMD Phenom X2 models. AMD rates these models with 45W TDPs. The low power Phenom X2 clocks in at 2.3 GHz, 2.1 GHz and 1.9 GHz with 3.4 GHz, 3.0 GHz and 2.8 GHz HT 3.0 speeds, respectively.

Although AMD plans to launch its new Phenom branding for high-end processors, the Athlon 64 X2 name lives on with Rana. There’s one Rana model in the pipeline clocked at 2.2 GHz. The Rana-based Athlon 64 X2 does away with L3 cache and only has 2x512KB of L2 cache, differentiating it from the Kuma-based AMD Phenom X2.

At the bottom of the Stars-family are the Spica Sempron models. Two Spica Sempron models occupy AMD’s value lineup. The new Spica Sempron models clock in at 2.4 GHz and 2.2 GHz with 512KB of L2 cache. As with the dual and quad-core products, the two Spica Sempron models have 3.6 GHz and 3.2 GHz HT 3.0 clocks, respectively. AMD rates the Spica Sempron models with 45W TDPs, as with the low power Phenom X2 models.

AMD plans to take orders for its low power Kuma Phenom X2, Rana Athlon 64 X2 and Spica Sempron models beginning in Q1’2008.


Comments     Threshold


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

People who pick sides crack me up.
By SavagePotato on 5/4/2007 11:26:51 AM , Rating: 3
I really do get a laugh at the kiddies that vehemently and brainlessly stand by "their" brand. Personally right now I own an E6400, which I got because it was fast, very overclockable, and just a really awesome chip. Fantastic upgrade from my Opteron 175 which was also a chip I got great usage out of.

What I have a hard time with is the notion that some people can be so short sighted as to cheer for one company to be beat out by "their" company. If you think for a minute without a competitor Intel won't be severely hiking prices think again.

It's just like the NVIDIA ATI war with people cheering the expected failure of the X2900. Have you looked at the costs of the 8800 series cards lately? Without competition you will be paying 2 grand for high end cpu's and 1k plus for video cards in very very short order.

Think of that while you are making your haw haw I hope they go bankrupt comments.




RE: People who pick sides crack me up.
By Shadowmaster625 on 5/4/07, Rating: -1
RE: People who pick sides crack me up.
By zsouthboy on 5/4/2007 2:35:17 PM , Rating: 5
Why is that?

I have a 6300 render box at home, next to my 165 server, and the 6300 blows it out of the water doing ... well, rendering, as it's supposed to. On order of 20-30 % I would assume his situation is similar.


By bjacobson on 5/8/2007 10:27:30 AM , Rating: 2
Well...there really aren't that many of these hardcore fanboys, as evidenced by AMD's latest quarter earnings.


By SavagePotato on 5/4/2007 4:46:24 PM , Rating: 1
Eh? Think again. For one I made a mistype there, I have an e6600 2.4ghz. However a 6400 will smoke an Opteron 175 as well.

In addition to that, my Opteron was good for about a 200 mhz overclock, just didn't get the pick of the litter. My 6600 I slapped with a 1 ghz overclock without even trying hard

For that matter even at stock the 6400 is damn near as good as an FX 62. and the 6600 at stock will crush an FX 62.

Oh my Thermalright ultra-120 extreme is on It's way too, so well see if we cant turn that modest little 3.4ghz into something a little more respectable like 3.6 or 3.8.

So I guess "lol" right back at you kiddie.


RE: People who pick sides crack me up.
By dnosferatu on 5/5/2007 1:09:45 AM , Rating: 2
...then why your "brainy" failed to compare Opt175 to E6400?
you can wait for K10, K11, or K-Whatever, but you can't deny the fact, those numbers are real, in case you don't have those numbers in your "brain", go google kid!


RE: People who pick sides crack me up.
By just4U on 5/5/2007 1:50:50 AM , Rating: 4
While I don't have my amd fanboi tee shirt on I can definitely say in all honesty that I would like to see them have continued success. I think it would be pretty good over all if both companies had close to the same hold on this market instead of the lopsided affair it's always been.


By dnosferatu on 5/5/2007 2:09:15 AM , Rating: 2
I agree, we need the competition along with the fanboy, not to mention that fanboy's posts are enjoyable to read.

My main reason is the price, boy i love price-cut everytime it happens, LOL.
Been with AMD since the day Athlon XP 1800+, only because i care for price/performance so much, not the performance crown-go digging dollar.

No point to talk the numbers we don't know yet but those kid already "i'm sure this, i'm sure that..", LOL
everything can happen..


RE: People who pick sides crack me up.
By Pythias on 5/9/2007 2:14:12 AM , Rating: 2
Then amd needs to build quite a few more fabs. Even if they gained majority market share, they couldn't supply it.


By Viditor on 5/9/2007 11:38:16 PM , Rating: 2
quote:
Then amd needs to build quite a few more fabs. Even if they gained majority market share, they couldn't supply it

Not true actually...
Many people look at the total number of Fabs that Intel has rather than the number of their Fabs that can actually produce modern CPUs...that number is 3 BTW.
AMD (when the Fab 38 conversion is complete in Q1 08) will have easily enough Fab capacity to deliver more than 60% marketshare.


RE: People who pick sides crack me up.
By crystal clear on 5/5/2007 2:26:39 AM , Rating: 1
quote:
Without competition you will be paying 2 grand for high end cpu's and 1k plus for video cards in very very short order


1)Yes competition is essential for higher quality products & lower prices.

2)For us the buyers we need an a market, where the market share of the 2 companies should be-(Intel)70% (Amd)30%.
In such conditions the buyer always stands to gain to enable you get higher quality products & lower prices.

3)EXCESS of competition & LACK of competition is BAD for us-we the buyers, the ULTIMATE DECIDERS of market share for these companies should maintain the above RATIO !(70 to 30).

4)Higher quality products & lower prices is an incentive or motivator for buyers/users/consumers to UPGRADE.

5)The lack of incentives to upgrade, means people simply dont upgrade-thats all ! they prefer to make do with what they have & wait for prices to drop.

6) Example for the above is VISTA !
People are just not motivated to upgrade & prefer to make do with XP & prefer to wait.They simple DONT BUY !

But let there be Apples O.S. "X" (tiger/leopard) up for sale
without the Apple hardware-Then watch !
If Apple sell its O.S. like M.S.-then see what happens-
YOU will see real WAR-Intel/Amd style.

We will get higher quality & lower prices !

7)Just remember-People are not that stupid-If the prices are HIGH they simply DONT BUY.So Intel can charge in your scenario "2G" & Nvidia "1K"- result-NO BUYERS !
I am sure there are millions out there who say this-
"I can live without a 8800 series card-Whats the hurry"
So Nvidia can charge what it wants-NO BUYERS !

Summary-WE the buyers decide the FATE of Intel & Nvidia !
They need US more than WE DO !


RE: People who pick sides crack me up.
By just4U on 5/5/2007 2:57:45 AM , Rating: 2
Actually crystal, I'd like to see it even more even say 55-45 either way with a constant struggle for that 10% That way we'd see less of the doom and gloom posts along with the bankruptcy scares.. provided they both were making money as you know even the more profitable companies can find ways to lose it all sometimes.


RE: People who pick sides crack me up.
By crystal clear on 5/5/2007 3:45:34 AM , Rating: 2
YES & NO

"Yes" indeed but unfortunately AMD is comapritively weaker financially to take on Intel to reach that ratio !
They need IBM(as a partner) to take on Intel to reach that levels-
It requires R&D+Finance+Marketing+Support etc that all put together as package.

"No" that struggle for that 10% is bad for STABILITY as MOBO manufacturers need to adjust their product offerings-they have to have time.
Too many new CPUs coming out into the market in short span is fine on paper,but in reality its upsetting everybodies(OEM/ODM) product cyles & schedules.Just count how many Cpus Intel released in short span (Core2duo)till today.

In short a period of stability is needed for all-from Buyers to Vendors to OEMs to ODMs & the whole industry itself.Thats the reason I quoted-

quote:
EXCESS of competition & LACK of competition is BAD


Fine tuning that ratio is the key !


RE: People who pick sides crack me up.
By KernD on 5/5/2007 10:00:08 PM , Rating: 2
About even market share wouldn't cause more architectural change than we have now, I don't think Intel would get them out faster than one a year, they wouldn't have the money to make it any faster than they are now. 45-55% split would obviously be ideal, they could have a similar R&D budget, just look at the graphic side of things, they have a balanced market and it works fine, the OEM aren't complaining.


RE: People who pick sides crack me up.
By crystal clear on 5/6/2007 12:47:59 AM , Rating: 2
You have a valid point there,but some very interesting news is on its way that could change everything.

ASSET LIGHT ! AMD !

quote:
asset light is code word for the company moving toward more of a "fabless" model of doing business in which computer chip companies outsource their manufacturing to factories called foundries, many of which are located in Asia.


This model frees up a lot of resources for both the companies(AMD/Intel),enabling them divert these resources to more essential/crucial areas namely R&D.
Resources I mean Money/Manpower/Time that will bring about as you say-

quote:
"more architectural change" & "would get them out faster than one a year"


The race for more & more "CORES" has already started-
today its 2 & 4 cores-up it goes to 8 then to 16 & more.
Even though, there isnt sufficient SOFTWARE around to exploit the full potential of these cores & what each cores does in the design.

So in a such scenario as you say-
quote:
45-55% split would obviously be ideal


RE: People who pick sides crack me up.
By Viditor on 5/6/2007 5:35:48 PM , Rating: 3
quote:
Asset Light...This model frees up a lot of resources for both the companies(AMD/Intel),enabling them divert these resources to more essential/crucial areas namely R&D

Unfortunately, Asset Light doesn't work for companies like AMD and Intel. Both companies are far ahead of the independant Fabs in Asia as far as technology goes, and that Fab technology is a major part of their R&D budget because it has to be. If it weren't for their Fab tech, we wouldn't have things like HKMG, SOI, APM, Copper interconnects, strained silicon, etc...


RE: People who pick sides crack me up.
By crystal clear on 5/7/2007 4:11:27 AM , Rating: 1
I had made a comment recently on Asset Light & gave the source from where I picked this information-see below

http://timesunion.com/AspStories/story.asp?storyID...

Details are vague & I asked the Ed(Kristopher Kubicki) to check on this as he has his ways to check up.

Yes in a way you are right-but who knows what AMD has in mind.
They are expected to release details on this in June.


By Viditor on 5/7/2007 9:12:10 AM , Rating: 2
The link doesn't work for me...


By KristopherKubicki (blog) on 5/7/2007 7:03:45 PM , Rating: 2
Link no worky. Email it to me direct.


By crystal clear on 5/8/2007 12:37:14 AM , Rating: 1
The new strategy, which Ruiz calls "asset light," is still being developed by the Sunnyvale, Calif.-based company. Caudell is part of the management team that is working on asset light, and the company has said it will give Wall Street analysts a better idea about how that strategy will work in July.

http://timesunion.com/AspStories/story.asp?storyID...

I hope now its ok.


RE: People who pick sides crack me up.
By crystal clear on 5/8/2007 12:54:03 AM , Rating: 1
Whats you E-mail by the way-helps- could send you a lot of stuff(news)-off the web communications.


RE: People who pick sides crack me up.
By Viditor on 5/6/2007 5:28:35 PM , Rating: 3
quote:
Higher quality products & lower prices is an incentive or motivator for buyers/users/consumers to UPGRADE

But consumer upgrades based on affordability are only a very small portion of the market. The larger ones are
1. Business upgrades based on need (to remain competitive)
2. New systems for both consumer and business

While I agree that even these would be affected negatively by large increases in cost, if there was no competitor then the profits from a price increase would easily surpass the resulting drop in overall market size...and remember that those who waited are still guaranteed customers at some point.


RE: People who pick sides crack me up.
By crystal clear on 5/7/2007 4:47:39 AM , Rating: 1
I made a comment on diffferent article- (on IBM)

Client-Server Computing Is Dead -IBM's view of business computing's future.

"The PC client-server model has run its course," proclaimed Palmisano, speaking in St. Louis at PartnerWorld, IBM's annual gathering of software developers and technology resellers.

As an alternate to the classic IT setup in which workers use applications stored locally on PCs while expensive servers are reduced to the role of traffic cop, Palmisano said IBM wants "to offer a new architecture for data centers

Source of information-

http://www.informationweek.com/news/showArticle.jh...

IBM road plan is worth reading about.

quote:
and remember that those who waited are still guaranteed customers at some point.


I wish I could quote you the link-I dont recall where I read it,but it says a consumer survey recently taken shows on an average the consumer spends around USD 1200 per annum
on high tech goods.
Spread this on cell phone/cameras/mp3/computer components etc.Depends on what & how much.

So money doesnt come out of the wallet so easily.


By crystal clear on 5/7/2007 8:04:18 AM , Rating: 1
A study conducted by the Consumer Electronics Association (CEA) has determined that the average American household spends approximately $1,200 annually on electronics products. The study, which also provides insight into the popularity of various kinds of consumer electronics products, was devised and conducted by CEA Market Research in February.

http://arstechnica.com/news.ars/post/20070430-cea-...


RE: People who pick sides crack me up.
By Viditor on 5/7/2007 9:14:06 AM , Rating: 2
quote:
IBM road plan is worth reading about

I can't get this link to work either...is it just me?


By crystal clear on 5/8/2007 12:42:36 AM , Rating: 1
Hey something is not OK here-strange

I clicked on the link & I got the page -no problem !


RE: People who pick sides crack me up.
By Viditor on 5/8/2007 8:19:58 PM , Rating: 2
Thanks Crystal...both of those links are now working for me.


RE: People who pick sides crack me up.
By Some1ne on 5/7/2007 4:04:56 PM , Rating: 2
Hello, cache? Get an E6600 next time, and then you won't have to try as hard to justify your purchase of a lesser chip with not enough cache to random people on the Internet.


By crystal clear on 5/4/2007 7:35:01 AM , Rating: 6
In an interview Wednesday with InformationWeek, Randy Allen, corporate vice president of AMD's server and workstation division, said performance and ease of adoption by computer makers were behind AMD's confidence in Barcelona.

Over the last couple of years, AMD has built the launch pad for Barcelona with the success of its current dual-core Opteron, which until the end of last year boosted AMD's market share at Intel's expense. Opteron today is offered by all the major x86 server makers, such as Sun Microsystems, Hewlett-Packard, IBM and now Dell.

In moving OEMs from dual-core to quad-core machines, AMD is using the same no-design-change strategy as its move from single-core to dual-core several years ago. Manufacturers only need to change the BIOS on new machines. Barcelona will fit into the same socket as the older chips. "There's a very low resistance path to adoption," Allen said.

Of course, better performance without increasing power consumption will be key to corporations, particularly one of AMD's key sectors: financial services. AMD believes it has that base covered; claiming Barcelona will deliver a 70% performance boost over its duo-core Opteron, while consuming the same amount of power.

AMD plans to release a full set of third-party benchmarks for Barcelona vs. Intel's Xeon quad-core -- codenamed Clovertown -- when Barcelona launches. In the meantime, AMD is tossing some numbers that it says back the argument that Barcelona will be better.

http://www.informationweek.com/news/showArticle.jh...




By ObscureCaucasian on 5/4/2007 9:49:35 AM , Rating: 5
I think I may be the only one here who thinks that K10 will become a competitive chip. Everyone's so quick to take sides on who they think is better when we have absolutely no data.


By Targon on 5/4/2007 1:55:01 PM , Rating: 2
I think there are a good number of people who feel that K10 will be competitive, but a huge number feel that AMD needs something MUCH better than Core 2 Duo in order to regain market share and recover. They don't know how well the K10 based processors will perform clock for clock, let alone in overall cost/performance, so they have been claiming that Intel will be able to hold a large lead.

So, let them doubt, since we only have another few months before real benchmarks are available.


By crystal clear on 5/5/2007 1:09:34 AM , Rating: 1
quote:
In moving OEMs from dual-core to quad-core machines, AMD is using the same no-design-change strategy as its move from single-core to dual-core several years ago. Manufacturers only need to change the BIOS on new machines. Barcelona will fit into the same socket as the older chips. "There's a very low resistance path to adoption," Allen said.


Just think on this-

Manufacturers only need to change the BIOS on new machines. Barcelona will fit into the same socket as the older chips.

Same SOCKET but a different BIOS.

So the vital factor is "BIOS"-it better be very good.


By raven3x7 on 5/6/2007 12:31:29 PM , Rating: 3
What the hell are you talking about? It needs a new bios so it can recognize the CPU. just like when 939 X2s came out. Or any newer CPU for that matter. You could update most mobos to have support. And those that didnt were not limited by design but rather by the (un)willingness of their manufacturer to update the bios


By crystal clear on 5/6/2007 4:48:20 PM , Rating: 1
quote:
Same SOCKET but a different BIOS.


Different or new Bios is the same thing.

Whatever the terminology,the whole point is "Quality".


By Viditor on 5/6/2007 5:38:44 PM , Rating: 2
quote:
Whatever the terminology,the whole point is "Quality"

I must admit that I'm confused by this as well...could you maybe give an example of bioses that were "lower quality"?


By Verran on 5/4/2007 2:17:47 PM , Rating: 1
The problem is that being "competitive" isn't good enough anymore. By the time K10 hits the desktop, AMD will be at least a year late, if not more. In the processor market, being "competitive" with a chip that's a full year older is a joke.

We will have waited over a year past C2D for K10. People are going to need to see a year's worth of progress past C2D, or K10 will be a let down.

When you figure Intel's 45nm should hit this year, K10 has some big shoes to fill.


By AMDJunkie on 5/4/2007 11:24:30 AM , Rating: 3
I'm not going to argue the timing of EOLing Socket 939, but the early swap to Socket AM2 makes a lot more sense now since you can just drop in a dual-core Barecelona with just a BIOS flash...


By Targon on 5/4/2007 1:57:54 PM , Rating: 5
The reason for the move to AM2 in the first place was because Intel was pushing DDR2, and OEMs didn't want to support two different types of memory. AMD would continue to support Socket 939 if and only if there was enough demand for systems based around the older DDR memory, and at this point, that demand just isn't there anymore.

There will be a big complaint about the move to DDR3 memory as well, since a new socket will be needed there as well, but the move will only happen if there is a HUGE performance advantage, or if Intel pushes hard enough to move to DDR3.


By leexgx on 5/5/2007 11:53:16 PM , Rating: 2
i see alot of users complaneing about amd Droping support for 939 socket for am2 but it needed to be done for DDR2 support, when thay moved to am2 system builders OEM computer comps was alot happyer to use AMD as it ment thay did not have to have DDr2 and DDR1 for ordering as it costs more to do it that way now thay can go Mass ordering DDR2 only
thay only need to pick the motherboad and the CPU and heat sink as dif in the build amd or intel

i just Slided to an 5600+ X2 now, sold my 3800 X2 to some one so the upgrade was only realy £60 to me just for the performace boost

when this New cpu comes out i should be able to keep the same motherboard and Drop and x4 in (if all goes well in the performace of it)


This is a bit Strange
By subhajit on 5/4/2007 5:04:46 AM , Rating: 2
Why a quad core processor should have lower HT Speed than a dual core one?




RE: This is a bit Strange
By billeman on 5/4/2007 5:10:58 AM , Rating: 2
It appears to be in direct relation to the processor's clock speed, not the number of cores


RE: This is a bit Strange
By subhajit on 5/4/2007 5:20:53 AM , Rating: 2
That's right. But shouldn't 4 cores need more data transfer rate than 2 cores?


RE: This is a bit Strange
By TodX on 5/4/2007 6:09:23 AM , Rating: 2
Yep, thats why socket 1207 has 3 HT links.


RE: This is a bit Strange
By AntDX316 on 5/4/2007 1:23:03 PM , Rating: 1
thats some insane bus speed

tho my proessor running at 1600 to 1800 fsb made absolutly no difference in super pi time and sometimes made it worse

but its set to 1800

1600 x9 = 3.6
1800 x8 = 3.6


RE: This is a bit Strange
By PlasmaBomb on 5/4/2007 6:10:51 AM , Rating: 2
I would put it down to thermal limitations with the quad cores, I guess we will find out when they are released.

It could lead to nice overclocks if the HT link is capable of running at 4.2GHz and is only clocked at 3.6GHz


RE: This is a bit Strange
By Zurtex on 5/4/2007 6:17:21 AM , Rating: 2
One would think, but it shouldn't provide a problem on benchmarks given the amount of data transfer HT 3.0 has is insane, the current HT 2.0 that K8 uses has much much more bandwidth than any of Intel's CPUs.


RE: This is a bit Strange
By PlasmaBomb on 5/4/2007 6:41:48 AM , Rating: 2
I was just thinking if the boards can run really high HT speeds then there is less need to fiddle. Certainly from the S939 days dropping the HT down to let you overclock made little difference.


RE: This is a bit Strange
By leexgx on 5/5/2007 11:56:51 PM , Rating: 2
HT speeds do not get used that much untill 3x (600Mhz/1200Mhz) is used, thats why lowering it does not do any thing to benchmarks

HT in K8 has plenty of bandwith in it K10 HT 3.0 going to be insane be intresting if it even needs it


On a positive note ...
By MartinT on 5/4/2007 6:19:17 AM , Rating: 3
Seems like AMD managed to wring another 100 MHz out of their <100W SKUs. Good for them.




RE: On a positive note ...
By PlasmaBomb on 5/4/2007 6:44:34 AM , Rating: 2
89 W seems good for a quad core, do we have any idea when they will be released yet?


RE: On a positive note ...
By GlassHouse69 on 5/4/07, Rating: -1
RE: On a positive note ...
By kamel5547 on 5/4/2007 12:08:02 PM , Rating: 3
Seeing as this is a core re-design I don't think a MHZ comparison makes any sense... I mean just look at P4 -> Core2.

Unless I missed something?


RE: On a positive note ...
By mkruer on 5/4/2007 1:38:49 PM , Rating: 1
You are missing something; The P4 is an entirely different branch of the design line. The Core 2 lineage is P3 > PM > Core > Core 2

AMD K10 lineage is K7 > K8 > K10

The K10 has more design changes then the Core to Core 2 yet the Core 2 is considered a major design change, go figure.


RE: On a positive note ...
By MartinT on 5/4/2007 5:14:55 PM , Rating: 2
I was actually referring to earlier roadmaps that had shown Barcelona to top out at 2.3 GHz / 95W, needing a 120W thermal envelope for 2.4 GHz and up.

Seems like AMD has bested their estimates in that regard, at least.


RE: On a positive note ...
By coldpower27 on 5/6/2007 2:22:37 PM , Rating: 2
Yeah, a little though this is talking about Agena, and not Barcelona, Server being 1 grade behind is right where they should be more or less for AMD at least.


RE: On a positive note ...
By raven3x7 on 5/6/2007 4:53:01 PM , Rating: 3
Huh? AMD's priority have been server chips ever since the first K8. It does not make sense for AMD to release a server chip with an increased TDP rating compared to standard desktop chips. The Barcelona TDP you are citing was released quite some time ago and was probably only simulated. Quite a few things could have changed. I guess only time will tell. Barcelona should be close now, maybe late June.


RE: On a positive note ...
By coldpower27 on 5/6/2007 10:32:51 PM , Rating: 2
No, AMD has had these Socket F TDP ratings for awhile now, and the 2 bottom ones are higher TDP at 68W and 95W compare to the desktop ones of 65W and 89W.

AMD's priority is server chips, but higher TDP doesn't mean unstable chips.

I am just saying 2.3GHZ for 95W TDP is normal because stability is more important then speed on the server side of things.

2.4GHZ for 89W on the desktop is probably achievable where things aren't quite as mission critical.

Barcelona will arrive, in Q3 2007 and will use Opteron TDP's not desktop numbers.


HyperTransport
By DesertCat on 5/4/2007 10:13:04 AM , Rating: 2
All of this talk about the various speeds of HyperTransport 3.0 really has me wondering if the AM2 upgrade path is viable with only HT 1.0. I know they are supposed to be socket compatible with AM2+ systems, but are these older AM2 boards going to severly cripple the performance of Stars processors? I was really hoping that HT 3.0 was just going to be room to grow for the future and not be super critical. If dropping a Kuma into an AM2 board throttles it back to the speed of a Pentium III, it really doesn't matter to me if it can fit in the socket or not.

All along I figured HT 1.0 might be too limiting for quad core, but I was hoping it might be adequate (as in no real performance penalty) for the Phenom dual cores. I'm not sure that's very hopeful after reading this piece. Or maybe I'm more concerned than I should be?




RE: HyperTransport
By Hoser McMoose on 5/4/2007 1:22:44 PM , Rating: 5
quote:
but are these older AM2 boards going to severly cripple the performance of Stars processors?

I highly doubt it. The only time this should have any effect at all is if the current Hypertransport bandwidth (4.0GB/s in either direction) is insufficient for your I/O. The one time that it's likely to have a noticeable (easily 10%) effect will be on system boards with integrated video when playing 3D games. Such a setup sends a LOT of data between memory and the motherboard chipset and right now Hypertransport is a bottleneck (4.0GB/s of bandwidth vs. up to 12.8GB/s from main memory). There may also be a TINY difference in really high-end 3D video setups with add-in cards, particularly if you're talking about Crossfire or SLI setups. However considering that a single PCIe 16x video card is limited to 4.0GB/s in each direction as well, HT 1.0 isn't much of a limitation here, we're probably looking at less than a 5% difference in performance.

The real advantages for HT 3.0 will only be realized in server and I/O intensive workstation applications in the near future. If you've got a 16+ 15Krpm SCSI disks in multiple RAID arrays and dual 10-gigabit Ethernet adapters, then HT 3.0 will be a godsend. For desktops though there just isn't that much I/O and Hypertransport 1.0 already provided a LOT of bandwidth.


RE: HyperTransport
By DesertCat on 5/4/2007 11:40:47 PM , Rating: 2
That's exactly the kind of technical information I was hoping for. Thank you very much!


RE: HyperTransport
By leexgx on 5/5/2007 11:59:19 PM , Rating: 2
i agree good post


Well I'm excited
By Zurtex on 5/4/2007 6:29:09 AM , Rating: 5
Well this has perked my mood up about AMD CPUs, I was dreading the horrible situation of them only being able to come out with 2.0 - 2.3GHz CPUs and not being able to improve on that for sometime. I realise that in FP it's supposed to out perform Intel's current CPUs by 50+%, which would make AMD's 2.3GHz equivalent to Intel's 3.45GHz, but life isn't just about FP performance and Intel aren't far off that and they'll be getting a 5-10% boost clock for clock off Penryn and Penryn will be getting a higher clock speed.

Seeing 2.6GHz on launch for the Socket 1207+ and 2.8 and 2.6GHz CPUs in the pipeline for AM2+ socket is rather pleasing. If AMD can make a continous effort to keep refining on this CPU at manufacture level and working on minor architectural improvements (1 - 2 year boosts) and work on their next generation tech at the same time, AMD might yet have a bright future *fingers crossed*

Intel has shown it works best well it feels threatened by competition, so I hope AMD provides really tough competition in the later half of this year.




RE: Well I'm excited
By Mitch101 on 5/4/2007 10:48:49 AM , Rating: 5
Ive heard similar bechmarks. A 2.5ghz K10 is supposed to outrun a 3.0ghz quad. That not the fudzilla link someone else posted as there are about 3 sources now on the web confirming this but dont go into details.

The cool part is K10 will improve dramatically in all applications while the Penryn will only improve on those with SSE4 instructions otherwise Penryn is just a slight improvement over the existing conroe core.

It comes down to IPC and efficiency and the K10 is really an excellent design. If I recall 70% of it has been redesigned from the existing K8 thats a pretty huge design change so its not a mere K8 redesign like the Penryn is a speed bump of the Conroe on 45nm.

Either road one chooses lets just hope AMD makes enough cash to return to profit and be around for a while.


Order availability vs. availability
By defter on 5/4/2007 8:27:23 AM , Rating: 2
I have one question:
Does "order availability" mean that distributors can order those chips from AMD? What does that mean for actual availability on the shelves? Can you tell approximately how much later (a week, month or quarter) actual availability will occur?




RE: Order availability vs. availability
By flatout on 5/4/2007 8:37:48 AM , Rating: 2
RE: Order availability vs. availability

It simply means they can be ordered nothing else .. delivery of actual product can be like six months away.


By Viditor on 5/4/2007 12:59:37 PM , Rating: 3
Order availability means availability to the channel (OEMs are pre-contracted). Ironically, this means that "shipping" happens before "order availability"...


So AMD did it again
By Khyron320 on 5/5/2007 2:26:10 AM , Rating: 2
Maybe i am missunderstood but what im reading is that if i want a quad core AMD i need yet another new socket. AMD is so frustraighting with all the socket changing i never understood why they didnt plan ahead more.




RE: So AMD did it again
By Viditor on 5/5/2007 3:52:30 AM , Rating: 2
quote:
Maybe i am missunderstood but what im reading is that if i want a quad core AMD i need yet another new socket

No...the AM2+ chips will work just fine in current AM2 mobos. The "+" means that if you want access to HT3 you will need a mobo that supports it, but other than that it will be just about the same.


It doesn't matter
By Domicinator on 5/4/2007 11:52:36 PM , Rating: 1
In the computer hardware world, being a fanboy makes no sense. It doesn't matter which one is faster, which one will be faster, or which one used to be faster. If you're blowing a bunch of money on a processor, it only makes sense to buy the fastest one you can afford. Who cares which company made it? If it's the fastest one you could afford, then you came out on top, whether it's Intel or AMD.




RE: It doesn't matter
By Zurtex on 5/5/2007 7:46:40 AM , Rating: 2
Because the people with lots of money buy the fastest, they also shape the market, whether they be server administrators that have to continuously buy 1000s of high end computers, or whether they be games enthusiasts. They make a big impact on the rest of the market as to what company makes quality goods.

AMD needs these CPUs to be faster than Intel's, both the high end to just be plain faster and the lower end to be better on price / performance. If this happens AMD can seriously compete again, if we have 2 strong competing companies we have a better market. We all know what happens when Intel feels secure that it has a more or less monopoly on the market.


Amd
By jlanders646 on 5/4/2007 9:59:08 AM , Rating: 2
Amd has always been my personal favorite processor, even now I would rather own an amd processor than an Intel one.
Hopefully this will pan out for them.




Wildcard
By Slaimus on 5/4/2007 6:08:46 PM , Rating: 2
While the increased execution units and bandwith is much needed, the real wildcard here is the L3 cache. Seeing how the onchip memory controller is what made the K8 so fast in games, I think the L3 cache will be the thing to watch.




Good to hear
By athfbum on 5/4/2007 7:41:24 PM , Rating: 2
I am glad AMD has finally unveiled K10. Hopefully it will be the C2D punisher it has been hyped up to be. K10 will definitely be in my Q3 2007 build :]

I know Intel fanboys are going to mock my above statements, but remember this: Without AMD, Intel would be charging people as much as they wanted for CPUs. Nobody wants a $500 E6320 now do we? :P




HEY NOW
By Phenick on 5/4/07, Rating: -1
Yawn!
By drunkenmastermind on 5/4/07, Rating: -1
RE: Yawn!
By bubbacub616 on 5/4/07, Rating: -1
RE: Yawn!
By Griswold on 5/4/2007 5:52:38 AM , Rating: 5
Use your brains. AMD is not in the position to ruin their own business before they can actually deliver that product by telling everyone how much better it will be than their current offerings. Intel did that last year and paid for it with horrible quarter figures until they actually launched C2D.


RE: Yawn!
By defter on 5/4/07, Rating: 0
RE: Yawn!
By flatout on 5/4/07, Rating: -1
RE: Yawn!
By Viditor on 5/4/2007 1:04:02 PM , Rating: 4
quote:
AMD has told that "K10" ramp will be relatively slow

No they didn't...they said that PROFIT from K10 would be slow. They don't expect K10 to make a big dent in the bottom line till early 08...
quote:
Horrible means >$1B profits per quarter?

Yes...if you view that as a return on investment.


RE: Yawn!
By Phynaz on 5/4/2007 1:54:32 PM , Rating: 3
Wrong. They said ramp (sales) would be low.

From Hectors own mouth:
"CRN: You mentioned the importance of Barcelona. What will be its impact on AMD this year?

RUIZ: This is an incredibly important product transition. We don't expect the ramp [this year] to be dramatic because it's a new core, new micro architecture and platform."

http://www.crn.com/sections/breakingnews/dailyarch...


RE: Yawn!
By Viditor on 5/4/2007 2:05:34 PM , Rating: 3
quote:
Wrong. They said ramp (sales) would be low

He said the ramp of Barcelona would be slow and that it would follow the same pattern as Opteron...
Barcelona is a server chip, but it's not the only K10 chip. Server chips take 6 months for validation, so they are ALWAYS slower. The same thing happened to Woodcrest.


RE: Yawn!
By Phynaz on 5/4/2007 3:03:29 PM , Rating: 2
So then you are expecting large shipments of desktop K10 chips this year?

How about a number? What percentage of desktop chips shipped will be K10?


RE: Yawn!
By Le Québécois on 5/4/2007 4:32:02 PM , Rating: 3
I doubt Viditor have number on this. But you didn't get the point of his comment. He was stating that it was Barcelona, not Agena, that would begin with slow sales numbers because of the way the servers market works. He was simply explaining what was the meaning behind Hector Ruiz quote.


RE: Yawn!
By defter on 5/4/2007 5:04:12 PM , Rating: 2
You seem to forget K8 ramp:
Opteron: April 2003
Athlon64: September 2003
Mid-range Athlon64s: early 2004

Use some common sense, server CPUs have significantly higher ASPs than desktop CPUs. It's logical that AMD will sell majority of initial K10 shipments for server market. Which means that if server ramp is relatively slow, desktop ramp will not be fast either.

Or do you think that AMD wants to sell 95% of K10 production to desktop marker for $200-300 if they can sell significantly higher portion to server market at $500-$4000?


RE: Yawn!
By Viditor on 5/5/2007 12:34:55 AM , Rating: 3
quote:
It's logical that AMD will sell majority of initial K10 shipments for server market. Which means that if server ramp is relatively slow, desktop ramp will not be fast either

One has nothing to do with the other...
The reason server ramps (for all companies) are slow is that servers are far more "mission critical" than desktops. OEMs and companies looking to purchase large quantities will test them (called validation) in a system first for at least 6 months before taking the large order.

The fact that the server ramp will be slow no matter what has no bearing on the desktop ramp though...over half of AMD's chips currently in production are already 65nm, and there is no reason why AMD would have made all of those K8s. I would bet that AMD has already begun a buildup of Agena and Kuma for Q3 release...


RE: Yawn!
By defter on 5/4/2007 5:08:35 PM , Rating: 2
quote:
No they didn't...they said that PROFIT from K10 would be slow. They don't expect K10 to make a big dent in the bottom line till early 08...


LOL, how can the volume be high if the revenues/profit will be low?? Do you really think that K10 will have lower ASPs than K8 line? K8 line will have very low ASPs in H2 2007...

quote:
Yes...if you view that as a return on investment.


You mean Intel's investment? Intel has very high margins, which means very high return on investment.


RE: Yawn!
By Viditor on 5/5/2007 1:29:31 AM , Rating: 3
quote:
how can the volume be high if the revenues/profit will be low

Mainly from the price war...but the majority of profits for AMD come from the server lines, and they will be limited by the time it takes for server validation.

quote:
Intel has very high margins, which means very high return on investment

Intel's margins have been dropping like a stone over the last year or 2...
Q1 07 is down 10% from Q1 06 and 15% from Q1 05...


RE: Yawn!
By ijakings on 5/6/2007 1:29:33 PM , Rating: 4
I felt the need to create an account over here just to reply.

Put down your fanboy stick. He said the ramp on barcelona would be slow. Barcelona is meerly the Server part of the K10 lineup. Server Ramps are always slower due to validation as was posted above.

Also Its quite easy for Profits to be low when Ramp is high. I Wonder if you actually know what ramp is?

In any case profits measured on a part will only come after the R+D and other pre release expenses on that part has been taken out.

In future when posting you need to

a) Stop being an Intel Fanboy
b) Research what your about to post so you dont do what you have done here and make a complete idiot out of yourself
c) If you cant understand something after you have researched it dont post it anyway.

Also.... Fuad? Is that you? If its not the amount of BS that spews from you is in an equal amount to his. You should go into Business.


RE: Yawn!
By jazkat on 5/4/07, Rating: 0
RE: Yawn!
By