Official Conroe information thread

[shaihulud]

As awesome as Conroe is, I can't wait for Kentsfield. AMD better do something. Intel announces the Core architecture as well as a future roadmap of processors, and all AMD does is say "we are ramping up production and moving to AM2." AMD has nothing on the horizon, unless they are keeping it secret (which is likely).

AMD may not have anything that is comparable when we finally can compare the two perfectly in every bench. But they can always lower the cost of the processor making them more economically viable for many consumers. Core is supposed to be lower in cost, IIRC, to the consumer. This will mean that AMD will have to lower the cost, specially, if the Core is better performer.

Honestly, I do not like to think as you do. I rather see it from the many different perspectives that it can be. If a processor is truly bad and not competitive in the many aspects that can be. Then so be the bad label. However, look at the differences now with processors, which are similar in argument, and still yet many will purchase Intel.

Do not count AMD out so easily. This is competition and can stimulate change. This is nothing but a good thing for all.

[Sandog]

Originally Posted by shaihulud

AMD may not have anything that is comparable when we finally can compare the two perfectly in every bench. But they can always lower the cost of the processor making them more economically viable for many consumers. Core is supposed to be lower in cost, IIRC, to the consumer. This will mean that AMD will have to lower the cost, specially, if the Core is better performer.

I think AMD has been riding a big wave of prosperity compared to yesteryear which has led them to reinvest heavily in new fabs and tooling. These expenditures are partly based upon current income (which has been high for AMD as you are aware of their high pricing and increased sales) and part is relied upon predicted sales. Both of these basis's will be diminishing IMO. Since AMD's spending has risen I believe it will be hard to adjust to lower income just as you may have experienced in your own financial dealings before. Now Conroe shows up recently and it looks like the scariest thing AMD could have expected. Sales are going to drop drastically for AMD IMO. Sales in desktops and servers. Intel way outclasses AMD in support and marketing of their products and it won't be hard to gain shares back again when Intels chips outperform AMD's in speed and power. Add the fact that Intel is one die size ahead of AMD which almost cuts the cost in half of the chips and I see bad times for AMD, really. Mark my words, it will take AMD to come up with a whole new architecture (3-5yrs like Intel) in order to pass Intel again. Intel now has AMD's number and from now on Intel will always be looking over its shoulders. Not only is Conroe looking to be the perfect desktop chip, but Intel is not stopping the train at all. Neither this year or next year or the year after.

[Water_abx]

Originally Posted by shaihulud

AMD may not have anything that is comparable when we finally can compare the two perfectly in every bench. But they can always lower the cost of the processor making them more economically viable for many consumers. Core is supposed to be lower in cost, IIRC, to the consumer. This will mean that AMD will have to lower the cost, specially, if the Core is better performer.

Honestly, I do not like to think as you do. I rather see it from the many different perspectives that it can be. If a processor is truly bad and not competitive in the many aspects that can be. Then so be the bad label. However, look at the differences now with processors, which are similar in argument, and still yet many will purchase Intel.

Do not count AMD out so easily. This is competition and can stimulate change. This is nothing but a good thing for all.

The Conroe that was benchmarked against the FX-60 is going to retail for about $530. Sure, the price of the FX-60 will fall by the end of the year, but what about the FX-62? From what I have read, that is going to retail for about $1100-1200. Here's a hypothetical question. Say the FX-62 and the 2.6GHz Conroe are similar in performance. Would you spend over $1000 for the FX-62, or spend half that amount for Conroe? What if the FX-62 beat the Conroe by a small margin? Would you spend over $1000 just for a small increase in performance?

In this hypothetical situation, AMD would have to do a crapload of mass production to get the price of the FX-62 down so it could compete with Conroe.

[Intel_User]

The Conroe that was benchmarked against the FX-60 is going to retail for about $530. Sure, the price of the FX-60 will fall by the end of the year, but what about the FX-62?

Yes, but the FX-60 benchmarked is OVERCLOCKED to theoretical FX-62 speeds, while Conroe 2.66GHz on a platform which supposedly had immature I/O drivers(Conroe system loaded games/apps much slower), beat FX-60 @ 2.8GHz by 20-30%.

[Intel_User]

As far as anyone likes to think, 20-30% won't be overcome by putting DDR2-800 and 4MB L3 cache. The general feeling is that the number advantage is gonna go few % higher for Conroe for the production version(immature drivers).

[Intel_User]

Say the FX-62 and the 2.6GHz Conroe are similar in performance. Would you spend over $1000 for the FX-62, or spend half that amount for Conroe? What if the FX-62 beat the Conroe by a small margin? Would you spend over $1000 just for a small increase in performance?

That's almost a definite NO. AM2 platform with DDR2-800 is supposed to bring 3-5% and that's the final production version. Say the AM2 FX's add 4MB L3. That still won't do it. Look here: http://www.anandtech.com/showdoc.aspx?i=1965&p=4

As little as 4-5% to much as 16%. That's talking about the Pentium 4 generations which were memory bandwidth sensitive. FX's are gonna get much less gain. The advantage is still gonna come to be 10-20% from today's 20-30%. AMD will have a heck of a time producing a 90nm chip with 4MB L3 slapped on IF they do that, and still not enough performance.

2.66GHz Conroe is July/August. AM2 is July 6th. That's awfully close.

[Sandog]

Like I pointed out in the previous post (didn't want to link to another forum board) a guy has a 2.4GHz Conroe running with only 2x256MB of RAM and stock 7800GTX and is getting over 10K in 3DMark05 and 21s in Super Pi benchmark. What will happen to this chip when it is clocked at 3-3.3GHx, has 667FSB, and is running with 2GB of fast DDR2? The AM2 release is going to look embarassing if it is anywhere near Conroes release. Also Clovertown is coming at the end of the year which has 2xWoodcrest dual cores with 4MB of cache. Pretty much every chip Intel produces after Conroe is going to be awesome. By the way, the current C1 Preslers are very nice chips IMO and it's a shame they won't be getting a longer life. They are miles ahead of the original 5xx series Prescotts.

[Intel_User]

On the subject matter of HT. I have a feeling that HT will not come to the processor. Even though it may work I do not think that given the logical layout and the parallelism of code it would be feasible. The governing fact is that 4 core SMP processors are around the corner. Although, HT is completely different in aspect to SMP I do feel that it will not create the advantage gain necessary. Instruction dependency with HT causes too much thrashing with resources. Core balances itself well and if HT is to make a come back it will be a different iteration that complements this architecture.

I think the main factor is that Pentium 4's were lacking IPC's to begin with and has a greater benefit with Hyperthreading/SMT. Considering how Pentium M has greatly improved IPC over PIII Tualatins, and how Core has greatly improved IPC over the latest iteration of the architecture, the Core Duo, HT may not benefit unless a new architecture is made for HT.

More Info about Conroe, specifically a 2.4GHz version on a bugged G965 chipset.

http://vic.expreview.com/index.php

Synthetic scores look AMAZING.

[shaihulud]

See the issue for many consumers is to see from a business perspective. If AMD's time was up it would have been long ago. Intel had superior products during K5, K6 and even parts of the Athlon era. Yet AMD did not disappear. If AMD was to fold it would have been during this period, and honestly, you do not want them to.

Almost definite is not definite. The engineering sample that is used for the speculation is not a good baseline. But what must ultimately has to be considered is the fact that performance is not the metric that is used for all purchases. This can be seen with the current processors now. If anything Water you are proving my point with your FX-62 argument, which means that you have also miscomprehended my statement. The only thing that I disagree with all of you is the fact that you all state nothing but gloom for AMD. Without doubt this will change the opinions and desires, but AMD will not be so easily incapacitated.

I agree that 4MB of L3 for the Athlon64 is not going to be as beneficial for its architecture is not as robust with cache, unless there is a completely new design to it. The returns on DDR2 should not be as high too. But these two changes will add up and can decrease the performance gap even if minimal. I also do not think that the L3 will include many if any non-enterprise markets. This is all speculation. Not until the chips are on the floor and the web sites have herald the new wearer of the crown can we come up with probable situations.

Where I know that the Athlon64 will hurt the most is SSE performance. This means many applications using these instructions will change the tide. Games will obviously benefit and the marketing decisions will adjust to Intel's favor . Now you know why I was so happy about the SSE units and ability of the processor to feed them. The Athlon64 per clock cannot decode, dispatch, execute SSE like the Core.

Second area that AMD has a weakness is supporting vendor system core logics. There is nothing that can compare to Intel's chipsets. VIA is horrible and should never be considered in anything but Chinese computers that stay in China. nVidia has made some better chipsets but are still not anywhere near the comparable mark of Intel's, and recently have many issues. SIS is not going to be considered much for it is not as robust, and ATI's introductions are still lacking in many aspects.

Soon we will all see how it unfolds on multiple fronts.

[Water_abx]

Originally Posted by Sandog

Like I pointed out in the previous post (didn't want to link to another forum board) a guy has a 2.4GHz Conroe running with only 2x256MB of RAM and stock 7800GTX and is getting over 10K in 3DMark05 and 21s in Super Pi benchmark. What will happen to this chip when it is clocked at 3-3.3GHx, has 667FSB, and is running with 2GB of fast DDR2? The AM2 release is going to look embarassing if it is anywhere near Conroes release. Also Clovertown is coming at the end of the year which has 2xWoodcrest dual cores with 4MB of cache. Pretty much every chip Intel produces after Conroe is going to be awesome. By the way, the current C1 Preslers are very nice chips IMO and it's a shame they won't be getting a longer life. They are miles ahead of the original 5xx series Prescotts.

667MHz FSB? I thought the Conroes would have a 1066MHz FSB? Or is that only for the "Extreme Edition Conroes?"

Don't forget, Kentsfield is right around the corner. As far what to look forward to after Kentsfield, I am not sure. I am more interested in Bloomfield (which is a 4 core, single-die processor) than Yorkfield (which, I believe, is an 8 core, multi-die processor).

[Intel_User]

The only thing that I disagree with all of you is the fact that you all state nothing but gloom for AMD.

Definitely not. At least for me. However, I hate how people think AMD is invincible. Some can't seem to realize that they can make mistakes. True, they had almost flawless execution since A64. But they can screw up. They did that during Athlon days, the were really good at it, and around 0.13 micron time, they started messing up. Of course people didn't believe it or want to believe it.

The thing is though, over time I found rumors are very accurate. Inquirer is a more credible site than most other sites, or what most people think of. Architecturally, there's nothing that AM2 brings to the table, except for DDR2. It has been known from the beginning of A64's that memory bandwidth is very insignificant for A64's.

People also talk about incompetency in Intel. However, I think it only applies to their desktop design. They have multiple design teams, and they have competent ones. One is their mobile division. They always had a superior product on their mobile division than their desktop versions.

Remember the screwups that Intel had for their desktop CPUs. They had one practically EVERY time after Pentium II. I don't really remember a time when there wasn't a delay.

One good competent team is the one behind Pentium M. Original Pentium M Banias were only 8 days late according to AT. For Dothan, it was delayed, but originally it was supposed to be 1.8GHz, and there were talks surfacing that it runs hot. The final revision came at 2.0GHz, and lower power consumption than Banias. Yonah came with no delay as far as I remember.

Look into Pentium 4 architecture, and you can't believe how some designs could have actually came through when they were so crap.

667MHz FSB? I thought the Conroes would have a 1066MHz FSB? Or is that only for the "Extreme Edition Conroes?"

Sandog is probably referring to base clock, and boy is he wrong. Mainstream Conroe's are 1066MHz FSB, and Extreme Editions are 1333MHz FSB. I guess he got confused with the fact its QDR, not DDR.

Without doubt this will change the opinions and desires, but AMD will not be so easily incapacitated.

Again, the problem I see is that people think AMD is "invincible". The fact is though, Intel is so big, its almost like having multiple companies in one. People can believe all they want, but follow the history of Intel's Desktop CPUs and Laptop CPUs, and you'll see the difference. The difference is greater when looking at starting from Pentium M generation.

Almost definite is not definite. The engineering sample that is used for the speculation is not a good baseline.

Sure, but is people thinking Conroe can't improve in the final version while AM2 can?? It's a wrong assumption. One of the sites was reporting that Conroe system had poor load up time because the I/O driver was not mature. While for AM2 reports are increases will be less than 5%.


Hehe, we don't seem to be good at being in topic here . It's supposed to be about Conroe information only.

[Sandog]

Originally Posted by Intel_User

As far as anyone likes to think, 20-30% won't be overcome by putting DDR2-800 and 4MB L3 cache. The general feeling is that the number advantage is gonna go few % higher for Conroe for the production version(immature drivers).

Not to mention how hot AMD is going to get (FX-60 is already up to 110W).

Inte_User: I accidently edited your original post instead of quoting it. Sorry, nothing changed.

[Sandog]

Nothing is going to happen to AMD that is ever going to ruin the company. Heck I believe I heard AMD celebrated their 45 anniversary recently. Everybody knows AMD has been great for the business. They brought prices down and generated new technology on both sides of the fence and all this while under monopolistic Intel. The point is there are Intel and AMD fans and as an Intel fan I didn't like the way AMD was constantly embarrasing Intel in public. My take on the AMD64 chips all along is AMD never had the talent or weight to try bold new architecture like the Prescott. AMD kept their copied old x86 technology and modified it while Intel basically shot itself in the foot with bold new steps. Now instead of Intel going ahead in unchartered waters trying to develop a new totally different architecture, their strategy is to gun for AMD's chips with low power and fast as possible throughput. Therefore we are seeing the Pentium3/M technology over again albeit much improved and with more features like VT.

[Sandog]

In my previous post about the Conroe EE I was refering to the CPU bus as running at 1333 FSB yes, I shouldn't have said 667MHz. I didn't know it was a given yet that the Conroe EE would be running at that speed or what the clock frequency will be. I knew the regular Conroes would have a 1066 FSB though.

[Sandog]

Article on Conroe pricing.

http://www.dailytech.com/article.aspx?newsid=1619