Intel's new microarchitecture: Merom, Conroe, Woodcrest
- Thread starter Gilbo
- Start date
). I have no doubt that AMD's dual-core prices are going to be significantly lower in several months.
CougTek
Hairy Aussie
It is hard to believe. How Intel could have concieved such an efficient processor after having developped utter P.O.S. like anythig on Netburst and the Itanium? AMD made such a convertion when they hired Dirk Meyer in the late nineties to develop the first Athlon core. Maybe Meyer has a younger brother and Intel decided to hire him.
I hope AMD will have had the time to make some money reserves in the last two years while they are eating huge market chunks from Intel, because if Conroe is THAT good and with Intel's capacity to sustain a price war, it could be a complete knockout. I don't wish to see AMD disappear. We certainly owe Conroe's development to AMD. It's because their Athlon 64 are so good that Intel was forced to flush the Netburst architecture down the drain and come up with a new one that makes more sense.
No Athlon 64 = no Conroe and Intel pushing the Itanium abobination down our throat.
I hope AMD will have had the time to make some money reserves in the last two years while they are eating huge market chunks from Intel, because if Conroe is THAT good and with Intel's capacity to sustain a price war, it could be a complete knockout. I don't wish to see AMD disappear. We certainly owe Conroe's development to AMD. It's because their Athlon 64 are so good that Intel was forced to flush the Netburst architecture down the drain and come up with a new one that makes more sense.
No Athlon 64 = no Conroe and Intel pushing the Itanium abobination down our throat.
Conroe is what competition does. Before, Intel had no real competition and then AMD produced the Athlon followed by the Athlon64. Suddenly, Intel was losing market share and the result, is the Conroe...
From what I can tell of the AMD published plans, they are still coasting because there is no Conroe killer upcoming. If Conroe lives up to its expectations, you can bet that AMD will accelerate its developement cycle and something good will be upcoming: It may take a couple of years, like with Intel but it will come...
From what I can tell of the AMD published plans, they are still coasting because there is no Conroe killer upcoming. If Conroe lives up to its expectations, you can bet that AMD will accelerate its developement cycle and something good will be upcoming: It may take a couple of years, like with Intel but it will come...
paugie
Storage is cool
Oh no! This means the intel boys in the forums I frequent will be swaggering back and forth and no one from the AMD camp will be able to stand up to them. Maybe not for the next 2 years.
Still, if this means cheap/fast AMD processors, I could probably let them (swagger, that is).
Still, if this means cheap/fast AMD processors, I could probably let them (swagger, that is).
Bozo
Storage? I am Storage!
Isn't Conroe based on the Petium M??
Bozo :mrgrn:
Bozo :mrgrn:
Gilbo
Storage is cool
Bozo,
Conroe takes many design lessons from the Pentium M, but it is an entirely different microarchitecture. It has a different pipeline, frontend, execution resources, and implements all sorts of new tricks that haven't shown up in any Pentium M derivative.
In general,
I think it is still too early to make any presumptions regarding AMD's competitiveness. While the system configurations seem pretty fair, I think it's important to remember that Intel configured those systems and that the entire bakeoff was a total PR stunt. There was not an ounce of independence or fair-mindedness in the people behind it. Setting aside hardware configuration (which does seem comparable) there are probably a hundred things Intel could have done to distort the results. Intel has used specially patched/tuned versions of games to show off processor performance before. I think it's always important to be skeptical of self-interested benchmarks. Remember, Intel has been almost as bad as Apple over the last couple years.
While I consider it likely, now, that Conroe is going to be clock for clock faster than the A64, I am 99% positive that when the independent benchmarks come out they'll tell a much less dramatic story. I wouldn't be surprised if a full 50% of the tested performance advantage "magically" disappears in independent tests. Of course, I could be wrong.
Regardless, I doubt AMD is in much trouble.
1. Their architecture is still dramatically superior in multi-socket server systems, which will give them solid revenue for at least another 2 years (Intel's own roadmaps declare that they are that far behind on interconnects). There's plenty of money for AMD here.
2. The cost of FB-DIMMs is apparently going to be quite high. (This applies only to Intel's server efforts of course.)
2. AMD's processor design teams have had 3 years to focus nearly their entire efforts and manpower patiently developing a succesor to the K8. I imagine that it only took a small fraction of AMD's design team to implement the improvements to the K8 core that have been done since it debuted. I don't think the rest have been sitting around gloating. In fact, I consider it very possible that AMD's next microarchitecture will be every bit as impressive as Intel's forthcoming one --if not more so considering all the free time Intel has given them. Of course, it's a couple years off, so this more of a longterm, positive forecast.
3. Intel chipsets have been ridiculously overclocker unfriendly lately. For many people with the know-how, AMD processors are going to massacre Conroe in terms of price/performance for long into the future unless the 975 is more like the 865 than the last couple years of chipsets would suggest --I would be shocked if this were the case. Present-day AMD CPUs and mobos simply overclock ridiculously far, ridiculously easily. Of course, this only matters to a small segment of the population, but it matters to me, and AMD doesn't have enough capacity to supply the entire population anyway.
Conroe takes many design lessons from the Pentium M, but it is an entirely different microarchitecture. It has a different pipeline, frontend, execution resources, and implements all sorts of new tricks that haven't shown up in any Pentium M derivative.
In general,
I think it is still too early to make any presumptions regarding AMD's competitiveness. While the system configurations seem pretty fair, I think it's important to remember that Intel configured those systems and that the entire bakeoff was a total PR stunt. There was not an ounce of independence or fair-mindedness in the people behind it. Setting aside hardware configuration (which does seem comparable) there are probably a hundred things Intel could have done to distort the results. Intel has used specially patched/tuned versions of games to show off processor performance before. I think it's always important to be skeptical of self-interested benchmarks. Remember, Intel has been almost as bad as Apple over the last couple years.
While I consider it likely, now, that Conroe is going to be clock for clock faster than the A64, I am 99% positive that when the independent benchmarks come out they'll tell a much less dramatic story. I wouldn't be surprised if a full 50% of the tested performance advantage "magically" disappears in independent tests. Of course, I could be wrong.
Regardless, I doubt AMD is in much trouble.
1. Their architecture is still dramatically superior in multi-socket server systems, which will give them solid revenue for at least another 2 years (Intel's own roadmaps declare that they are that far behind on interconnects). There's plenty of money for AMD here.
2. The cost of FB-DIMMs is apparently going to be quite high. (This applies only to Intel's server efforts of course.)
2. AMD's processor design teams have had 3 years to focus nearly their entire efforts and manpower patiently developing a succesor to the K8. I imagine that it only took a small fraction of AMD's design team to implement the improvements to the K8 core that have been done since it debuted. I don't think the rest have been sitting around gloating. In fact, I consider it very possible that AMD's next microarchitecture will be every bit as impressive as Intel's forthcoming one --if not more so considering all the free time Intel has given them. Of course, it's a couple years off, so this more of a longterm, positive forecast.
3. Intel chipsets have been ridiculously overclocker unfriendly lately. For many people with the know-how, AMD processors are going to massacre Conroe in terms of price/performance for long into the future unless the 975 is more like the 865 than the last couple years of chipsets would suggest --I would be shocked if this were the case. Present-day AMD CPUs and mobos simply overclock ridiculously far, ridiculously easily. Of course, this only matters to a small segment of the population, but it matters to me, and AMD doesn't have enough capacity to supply the entire population anyway.
sechs
Storage? I am Storage!
But can I cook eggs on it?
Stereodude
Not really a
Looks good.
K9/K10 K??
I will second the comments on these Conroe v overclocked FX-60.
It appears that the BIOS on the motherboard for the FX-60 used by Intel did not actaully support the FX-60, so a few (or more) of the control registers might not have been set correctly. When the original (preview) reviews of the Athlon 500/550 surfaced, things did not look so hot, and this was specifically a MSSR/control register issue. I'd wait for independent benchmarks.
I do believe that Conroe is very good and definitely a player. The Empire will be back.
What has AMD been doing. AFAIK, AMD have scrapped/pushed back two alternative designs for the K8 successor. K8L is the next big thing. From what I can gather, it has 2MB/core cache (4MB total) double the FP units of K8, plus memory controller enhancements. It can be had as dual core or quad core. Quad core/DDR2 chips may appear this year—but I would only expect them as more as a marketing thing.
A note on cache. Intel is going the unified L2 cache route (1 x 4MB vs 2 x 2MB on a dual core chip). Unified is reputed to give you 10% more performance versus split. The downside is cache thrashing. Very few benchmarks show cache thrashing, though servers are more prone to it. Most servers with unified L2 (HT XEON's) ship with HT tuned off by default or benchmarks show the servers as performing better with HT disabled. As the toolsets for building applications get better, this problem will lessen, but be very careful of the benchmarks used to proclaim x over y. The benchmark might prevail, but will the system in the real world? Split caches are the safer alternative.
The real successor to K8 won't be seen until mid '08 at the earliest.
I will second the comments on these Conroe v overclocked FX-60.
It appears that the BIOS on the motherboard for the FX-60 used by Intel did not actaully support the FX-60, so a few (or more) of the control registers might not have been set correctly. When the original (preview) reviews of the Athlon 500/550 surfaced, things did not look so hot, and this was specifically a MSSR/control register issue. I'd wait for independent benchmarks.
I do believe that Conroe is very good and definitely a player. The Empire will be back.
What has AMD been doing. AFAIK, AMD have scrapped/pushed back two alternative designs for the K8 successor. K8L is the next big thing. From what I can gather, it has 2MB/core cache (4MB total) double the FP units of K8, plus memory controller enhancements. It can be had as dual core or quad core. Quad core/DDR2 chips may appear this year—but I would only expect them as more as a marketing thing.
A note on cache. Intel is going the unified L2 cache route (1 x 4MB vs 2 x 2MB on a dual core chip). Unified is reputed to give you 10% more performance versus split. The downside is cache thrashing. Very few benchmarks show cache thrashing, though servers are more prone to it. Most servers with unified L2 (HT XEON's) ship with HT tuned off by default or benchmarks show the servers as performing better with HT disabled. As the toolsets for building applications get better, this problem will lessen, but be very careful of the benchmarks used to proclaim x over y. The benchmark might prevail, but will the system in the real world? Split caches are the safer alternative.
The real successor to K8 won't be seen until mid '08 at the earliest.
Sol
Storage is cool
Sadly all the next generation CPU designs do seem to have missed the P4s intergrated sstove-top advantage...
It's worth remembering when looking at these benchmarks that Shania Twain would not be greatly impressed... (O.K so it wasn't really...)
Tannin
Storage? I am Storage!
Mark's daughter?
AnandTech goofs on F.E.A.R. and other issues
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=4
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716&p=4
Bozo
Storage? I am Storage!
Re-tested
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716
But Intel still ahead, but not by as much.
Bozo :mrgrn:
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2716
But Intel still ahead, but not by as much.
Bozo :mrgrn:
CougTek
Hairy Aussie
22% on average, I call that a huge lead for a similarly clocked processor. AMD will need more than just DDR2 to catch up. I doubt the new memory controller will boost them for more than perhaps 5%. The new Crossfire from ATI (RD580) is barely faster than the older RD480, so don't count on that to narrow the gap.
AMD better hopes that Intel will have troubles with their yield for the 4MB L2 cache Conroe, but I doubt that. Intel has a lot of experience with large cache architectures and Conroe, which is a low frequency processor, shouldn't be that complicated to manufacture for Santa Clara's smurfs.
AMD better hopes that Intel will have troubles with their yield for the 4MB L2 cache Conroe, but I doubt that. Intel has a lot of experience with large cache architectures and Conroe, which is a low frequency processor, shouldn't be that complicated to manufacture for Santa Clara's smurfs.
CougTek
Hairy Aussie
I think so.
Stereodude
Not really a
That's what I'm thinking.LunarMist said:
*as I post from a 3.06gHz P4 with HT built during Q4 of 2002*
Gilbo
Storage is cool
Of course, it all depends on prices. Even when Intel has had the performance lead, AMD has always been the better buy at any given price point. If Intel's performance advantage is indeed this significant then their chips will definitely at least be solid contenders on my price/performance list. The only Intel processor I've even considered buying in the last three or four years was the 2.4C and only for a moment --until the Barton 2500+ was discovered.
Personally, I'll be waiting till about this time next year to upgrade in all probability, or later. Nothing I have seems remotely slow at any tasks that require my active attention. RAM prices the last couple years have actually been the big thing that has driven my upgrade habits. I usually invest in a new system when I can inexpensively stock it full of much more RAM than my old system. I've always found that RAM has been the single biggest factor in the responsiveness of a computer I'm using. I'll probably upgrade again when 2GB sticks near the GB/dollar ratio of 1GB sticks. That'll probably be more like 18 months from now I suppose.
I certainly think the next 18 months are going to be interesting, and I have a feeling that however things turn out they're going t o be great for us. I couldn't care less who makes my CPU as long as it's inexpensive and genuinely more useful than my old one.
Personally, I'll be waiting till about this time next year to upgrade in all probability, or later. Nothing I have seems remotely slow at any tasks that require my active attention. RAM prices the last couple years have actually been the big thing that has driven my upgrade habits. I usually invest in a new system when I can inexpensively stock it full of much more RAM than my old system. I've always found that RAM has been the single biggest factor in the responsiveness of a computer I'm using. I'll probably upgrade again when 2GB sticks near the GB/dollar ratio of 1GB sticks. That'll probably be more like 18 months from now I suppose.
I certainly think the next 18 months are going to be interesting, and I have a feeling that however things turn out they're going t o be great for us. I couldn't care less who makes my CPU as long as it's inexpensive and genuinely more useful than my old one.
sechs
Storage? I am Storage!
Unless AMD suddenly gets stupid, they will always hold on to that price/performance edge. AMD's continuing problem is chipsets.
Bozo
Storage? I am Storage!
Bozo
Storage? I am Storage!
Anybody know what socket the Conroe CPU uses?
Bozo :mrgrn:
Bozo :mrgrn:
Gilbo
Storage is cool
Current LGA775 boards are supposed to support Conroe with only a BIOS update.
I think it goes without saying of course that "support available with a BIOS update" is one of the most abused phrases in personal computer history.
I think it goes without saying of course that "support available with a BIOS update" is one of the most abused phrases in personal computer history.