Opteron 150 VS 3.6GHz Xeon Narcona review up!
- Thread starter Sonic587
- Start date
its' very strange.
1. why does he suspect the nocona will do better in multi processor systems? Dont intel cihps do worse?
2. HE SAYS the nocona will be more expensive than the opteron so they are not direct competitors, but doesn't that make it worse for intel?
1. why does he suspect the nocona will do better in multi processor systems? Dont intel cihps do worse?
2. HE SAYS the nocona will be more expensive than the opteron so they are not direct competitors, but doesn't that make it worse for intel?
Markfw
Moderator Emeritus, Elite Member
- May 16, 2002
- 27,110
- 16,021
- 136
It would seem to me that this time they commented a lot on why this was still not completly a fair shootout for various reasons, but that they could however say that the Opteron still was faster (much faster on database operations where they are primarily used), and cheaper than the new Intel CPU which isn't even available yet. I also am dying to see 2 way and 4 way benchmarks. And did anyone else read the tomshardware article on the Opteron farm that was used for StarwarsII, and will be used for Starwars III ?
I wonder if the Opteron system will really start flexing its muscle in the 4way benchmarks. Should be interesting.
Yeah, a much better review, but too much speculation in the conclusions. I don't see how multiprocessor system would improve the Xeon performance. I think I have seen a review where dual Xeons still lose against dual opterons.
The reasons:
-The Opteron architecture solves motherboard bottleneck problems with Hypertransport and the built-in memory controller. The Opteron architecture easily scales to eight CPUs.
-Traditionally, communication between the hard drives, AGP/PCI cards, memory and CPU have been slowed by having to pass through various controller and Bridge Chips on the motherboard. With Hypertransport, AMD has allowed components to talk to each directly.
-The on-CPU memory controller lowers latencies because it gives direct access to RAM. The memory controller also clocks in at CPU speed. Multi-processor systems have traditionally been hampered by memory latencies - after all, you have to feed all those processors information. This is where the competition can't keep up.
These two processor are really direct competitors, since intel don't have anything better to face the opterons. AMD is also going for APPLE. If you visit the Apple website, and look at the G5 CPU comparisons, you'll see that opteron is missing. Now you can tell who the processor giant is.
http://www.apple.com/powermac/performance/
The reasons:
-The Opteron architecture solves motherboard bottleneck problems with Hypertransport and the built-in memory controller. The Opteron architecture easily scales to eight CPUs.
-Traditionally, communication between the hard drives, AGP/PCI cards, memory and CPU have been slowed by having to pass through various controller and Bridge Chips on the motherboard. With Hypertransport, AMD has allowed components to talk to each directly.
-The on-CPU memory controller lowers latencies because it gives direct access to RAM. The memory controller also clocks in at CPU speed. Multi-processor systems have traditionally been hampered by memory latencies - after all, you have to feed all those processors information. This is where the competition can't keep up.
These two processor are really direct competitors, since intel don't have anything better to face the opterons. AMD is also going for APPLE. If you visit the Apple website, and look at the G5 CPU comparisons, you'll see that opteron is missing. Now you can tell who the processor giant is.
http://www.apple.com/powermac/performance/
i read some news that intel will be having a dual fsb on xeon next year
do you guys it would help them?....
but this review is much better
it explains everything from the beginning
do you guys it would help them?....
but this review is much better
it explains everything from the beginning
how do u have a dual fsb? the only thing close to it i can think of is amd's hypertransport that has a 2 way bus. 800Mhz each direction.
Each cpu would have a separate bus between itself and the MCH, like the dual-G5.
http://images.apple.com/...architecturediagram06232003.jpg
Ahh... but if like the pic you listed, it would still only have 1 path to the memory. Something AMD would still lead over in high end boards.
Dont take this as a bash against the review, however I have ALWAYS heard that the GCC compiler favors AMD CPUs, and I remember back on Acehardware a post saying that there isnt even a decent scheduler for the P4 Arch in GCC.
I remember reading a review on Ace's showing that while MySQL ran faster on AMD, MS SQL ran much faster on the Xeons.
Most of my day is in a windows server world, and the benches I have done on this side of things tend to give the Xeons an edge. It would be nice to see the performance differences with these to chips on other OSs, and more importantly other compilers.
I am curious in the best possible performance. Just as when Apple benched their G5s vs the Xeons, you could run the same test with the Intel compiler and have tripple the performance in some areas where the GCC would just drag its feet on the Intel arch.
I guess my main beef with these reviews is that 80% of the benches most people will never use in real life applications. I would like to see benches of say Xeon vs Opteron using enterprise SQL servers (Oracle and MS SQL), Perhapse Exchange benches. Maybe use IIS in some benches instead of Apache, and some IIS vs Apache (I havent seen one of those in years). I would bet alot of this is not wanting to review with beta or RC software, but I think it would be far more interesting. Also, maybe some notes on the stabilty of the systems, a tally at the end of what crashed, when it crashed, etc.
I remember reading a review on Ace's showing that while MySQL ran faster on AMD, MS SQL ran much faster on the Xeons.
Most of my day is in a windows server world, and the benches I have done on this side of things tend to give the Xeons an edge. It would be nice to see the performance differences with these to chips on other OSs, and more importantly other compilers.
I am curious in the best possible performance. Just as when Apple benched their G5s vs the Xeons, you could run the same test with the Intel compiler and have tripple the performance in some areas where the GCC would just drag its feet on the Intel arch.
I guess my main beef with these reviews is that 80% of the benches most people will never use in real life applications. I would like to see benches of say Xeon vs Opteron using enterprise SQL servers (Oracle and MS SQL), Perhapse Exchange benches. Maybe use IIS in some benches instead of Apache, and some IIS vs Apache (I havent seen one of those in years). I would bet alot of this is not wanting to review with beta or RC software, but I think it would be far more interesting. Also, maybe some notes on the stabilty of the systems, a tally at the end of what crashed, when it crashed, etc.
Markfw
Moderator Emeritus, Elite Member
- May 16, 2002
- 27,110
- 16,021
- 136
I love my Opterons, but you have many very valid points, and I would love to see some enterprise level sql benchmarks ! The only ones I saw were years ago at Anandtech of all places, and the Opterons wiped the floor with the Xeon.
BUT, this is the next generation Xeon, and almost the same generation Opteron (PC3200 memory support now is all I know of for improvements), so things could have changed. Anand, care to do a replay with new hardware ?
we'll, based on the benchmarks, the opteron at 130nm does very well against intel's new 90nm xeon. this just shows that when the 90nm opterons come out, they will perform much better compared to the xeons.
Check the frontpage news section here. AMD just started shipping the 90nm chips. the first ones are for laptops. Also dual core chips are still on target for Q2 2005 for AMD
90nm + core enhancements + SSE3 + more speed for Opteron = BIG TROUBLE FOR XEON!
Now, if Nforce4 would hurry up and get released...
Now, if Nforce4 would hurry up and get released...
Dug, I just skimmed the article but didnt see what OS and which SQL server Anand uses (And of course what was used to compile the apps (if *nix )
DrMrLordX
Lifer
- Apr 27, 2000
- 22,733
- 12,718
- 136
Um, not quite.
The 3.6e is a desktop prescott without EMT64
The 3.6f is a desktop prescott with EMT64
The processor that is benchmarked vs the Opteron 150 on AT is a 3.6 ghz Xeon, which is just a 3.6f that requires registered memory and may be run in multi-proc configurations(workstation/server cpu).
The 3.6e is a desktop prescott without EMT64
The 3.6f is a desktop prescott with EMT64
The processor that is benchmarked vs the Opteron 150 on AT is a 3.6 ghz Xeon, which is just a 3.6f that requires registered memory and may be run in multi-proc configurations(workstation/server cpu).
Follow The God That Failed : Intel
90 nm improvments for A64/opteron.
AMD has begun using a 90-nm SOI manufacturing process at its Fab30 in Dresden, Germany ( as you all know)
- 90-nm AMD64 is in active pilot mode in Fab 30 and prototype parts are running in systems.
- 90-nm process would come in with all the advanced features acquired during previous generations; including copper interconnect, a Black Diamond low-k technology, and as an SOI process with the base wafers provided by Soitec SA.
- The move to 90-nm has reduced the die size to 114 square millimeters, about 40 percent of the die area compared with the established 130-nm process, with reference to an Opteron prototype
- The use of SOI at the 130-nm process node had reduced processor power consumption to the 45-W to 55-W mark and that the shrink to 90-nm would produce an additional benefit
- Increasing the speed of the embedded memory controller and reducing the latency in the HyperTransport technology
- 90-nm process started with a nine-layer interconnect back-end to match the 130-nm process to allow easy transition of established products, but that for new or re-laid products the process could be extended to 11 layers.
- AMD expects commercial shipment of products made using the 90-nm process to begin in the third quarter.
- 300-mm wafer processing now expected early in 2006.
- Strained-silicon would be introduced by AMD at some point in the future.
VR Zone
90 nm improvments for A64/opteron.
AMD has begun using a 90-nm SOI manufacturing process at its Fab30 in Dresden, Germany ( as you all know)
- 90-nm AMD64 is in active pilot mode in Fab 30 and prototype parts are running in systems.
- 90-nm process would come in with all the advanced features acquired during previous generations; including copper interconnect, a Black Diamond low-k technology, and as an SOI process with the base wafers provided by Soitec SA.
- The move to 90-nm has reduced the die size to 114 square millimeters, about 40 percent of the die area compared with the established 130-nm process, with reference to an Opteron prototype
- The use of SOI at the 130-nm process node had reduced processor power consumption to the 45-W to 55-W mark and that the shrink to 90-nm would produce an additional benefit
- Increasing the speed of the embedded memory controller and reducing the latency in the HyperTransport technology
- 90-nm process started with a nine-layer interconnect back-end to match the 130-nm process to allow easy transition of established products, but that for new or re-laid products the process could be extended to 11 layers.
- AMD expects commercial shipment of products made using the 90-nm process to begin in the third quarter.
- 300-mm wafer processing now expected early in 2006.
- Strained-silicon would be introduced by AMD at some point in the future.
VR Zone
btw concerning 90 nm, AMD Athlon 64 3700+ & 4000+ (Winchester) is the 90nm successor to the Newcastle Athlon 64 core, expected to be released in September / October. @ 2.6Ghz (4000+)
I think opteron/ A64 will reach 3.2Ghz on a 90 nm, big trouble for Xeon/Presccot ( remember A64/Opteron can reach 2.6Ghz on a 130 nm process, the FX 55 will be made on a 130 process). The 0.13 micron process has one last speed bump left in it but i spose AMD just want to cut costs by going over to 90 nm quickly, plus the yeilds @ 2.6Ghz might be say 50%.
I think opteron/ A64 will reach 3.2Ghz on a 90 nm, big trouble for Xeon/Presccot ( remember A64/Opteron can reach 2.6Ghz on a 130 nm process, the FX 55 will be made on a 130 process). The 0.13 micron process has one last speed bump left in it but i spose AMD just want to cut costs by going over to 90 nm quickly, plus the yeilds @ 2.6Ghz might be say 50%.
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter