Intel demonstrated two quad-core processors Tuesday, ‘Clovertown’ for servers and ‘Kentsfield’ for PCs, directing attention toward the future during a more troubled present. Pat Gelsinger, a senior vice president in Intel’s Digital Enterprise Group, demonstrated both processors in a speech at the company’s Intel Developer Forum here. Both chips are built using Intel’s 65-nanometer manufacturing process and will ship in the first quarter of 2007, Intel representatives said.
Or maybe they will take us by surprise and release an quad core cpu way before we know it, just like they did with 1 GHz cpu.
I can´t help it, but i wish AMD will beat Intel again, becuase i really like the performance of the Opterons and AMD64:s.
BUT, intel still holds the crown in portable pc segment.
They aren’t shipping until 2007. About the same time-frame as AMD’s, last I read. This is pure PR, similar to paper launches.
How good it is we will see. I also hope AMD will beat Intel again, because doing so more than once in a row will win them business, and heighten the competition (without the Opterons and A64s, do you think we’d have Merom on the way, at this point?), like no amount of advertising ever could.
*smacks self*
…replace that with Conroe. Code names are like TLAs…
Intel does hold the crown in the mobile segment but not for long especially with the Yamato platform.
Intel does have the Conroe and Merom to be reckoned with. I dont care which side wins as long as there is a setup that is faster. Though I like AMD since they are the underdog so I am contradicting my own statement.
<bold>Intel does hold the crown in the mobile segment but not for long especially with the Yamato platform.</bold>
Has Intel something like the AMD Geode NX 1750@14W processor that is a x86? Like the Geode NX 1500@6W processor?
“Has Intel something like the AMD Geode NX 1750@14W processor that is a x86? Like the Geode NX 1500@6W processor?”
Yes, since Banias times you can software underclock P-M to the minimum clock, the usually have a TDP of 4-8 W, enough low to be fanless cooled.
Even @ 600 – 800 – 1000 MHz those processors have awesome horsepower if compared to other low-tdp processors as VIA, Transmeta or AMD Geode.
Edited 2006-03-08 08:43
I’m developing very cpu intensive statistical applications (read data, do stuff) and I constantly test different systems performance (Intel, AMD, Itanium, Power, Sparc, etc.) and I have to say that AMD Opteron cpus are best.
I was very very surprised when 64-bit Intel Xeon 3.6GHz lost cpu intensive benchmark to AMD Opteron 275 (which is 2.2GHz) under my tests.
Of course I have to admit that my applications are non-parallel…but still…
May I ask why you were surprised when the Xeon lost? Is this the 1066 FSB Xeon you are talking about?
I’m developing very cpu intensive statistical applications (read data, do stuff) and I constantly test different systems performance (Intel, AMD, Itanium, Power, Sparc, etc.) and I have to say that AMD Opteron cpus are best.
Your test results do not say about general chips performance but rather about architecture of your particular application. Either your application is highly memory intensive (Which Power cpu you have tested, Power4 or Power5? The latter has on-die memory controller) or there is some non-optimised algorightms used.
My applications are only cpu intensive (does not require lot’s of mem) and I have tested both power4 and power5.
And I’m sure that I don’t have too much optimized tests for each platform Just basic optimizations (GCC -O2) used.
There is some very cool stuff in the new “Core” architecture:
1) Three (likely pseudo-symmetric) integer pipelines. This is a good number for heavy integer code — most code can’t take good advantage of 4, but can take advantage of more than 2.
2) Micro-ops and macro-ops fusion. In addition to fusing LOAD-OPs after decode, Core will now do things like fuse TEST-JUMP before decode. This effectively raises both the decode width and the issue width of the processor.
3) Indirect branch prediction and loop prediction. The former is going to be good for object-oriented code. I did some tests on a P4 once, and found that if you have < 100 targets for a method dispatch, its much faster to go through a decision tree to find the target then to do a simple indirect jump indexed by class ID. I’d really like to see if Core’s indirect branch predictor changes that.
Overall, it looks like Conroe is going to be a very well-rounded processor. There are still issues (the lack of an integrated-memory controller is going to hurt server tasks, it’ll probably still have 1/2 the L1 cache of a K8), but those aren’t dealbreakers. The preliminary SPEC scores are great — SPECint is up ~30% per clock over the Opteron, and SPECfp is about equal. Intel’s 65nm process is really going to help them — if they ship the top-end parts at 3GHz+, they’ll have the IPC and clockspeed lead. If AMD’s 65nm process was as good as their overachieving 90nm has been, AMD could still be competitive, as the K8 could likely scale to 3.5 GHz+ at 65nm, but it looks like we won’t see 65nm from AMD for awhile.
Edited 2006-03-08 01:07
As I wrote in my blog, http://uadmin.blogspot.com/2006/03/quad-xeon-but-still-missing-impo…
Intel is missing the most important part to the equasion, what is the point of adding cores, if you can’t supply enough data to keep them busy, even AMD’s low end dual core chips beat Intels best dual core offering. I honestly expect AMD’s highend dual core chips to beat Intel’s quad core ones. Of course AMD has allready hinted that it has quad core CPU’s near production ready, so once again Intel will find it self a year or more behind AMD.
I forecast lots more missed quarterly sales quotas before Intel gets a clue and fixes its root problem, an on chip memory controller
I honestly expect AMD’s highend dual core chips to beat Intel’s quad core ones.
Sorry to break your expectations, but…
http://www.anandtech.com/tradeshows/showdoc.aspx?i=2713
2.66 Ghz Conroe is upto 20-40% faster than 2.8Ghz FX-60.
I forecast lots more missed quarterly sales quotas before Intel gets a clue and fixes its root problem, an on chip memory controller
Intel CSI interconnect bus will appear in 2007. As for now, well, the 1333MHz FSB is not so bad.
one little problem… the FX-60 isn’t AMD’s highend chip, its the highend of the middle of the road, opteron is AMD’s high end line, not the FX-60 that the article would have you beleve. The opteron 2xx and 8xx series have even more memory bandwidth, so don’t let the pro-intel sites fool you.
I Didn’t see any world record benchmarks posted from Intel perhaps they are hiding something?
Explain for the reader the computationally-significant differences between the FX-60, Opteron 185, Opteron 285, and Opteron 885 in single-processor configurations. Also help the reader by pointing out where he can purchase a CrossFire Opteron 885 board.
Explain for the reader the computationally-significant differences between the FX-60, Opteron 185, Opteron 285, and Opteron 885 in single-processor configurations.
http://www.devx.com/amd/Article/27340
“AMD’s means of connecting the chips to other system components, such as I/O devices or other processors is called HyperTransport technology. It is based on the concept of links (that is, pathways) that are 16 bits wide and capable of moving 3.2Gbytes/second in each direction.[per HT link]”
“All Opteron models come with three HyperTransport links (vs. a single link for Athlon 64). On single-processor Opterons, the links are connect only to the I/O devices[Upto 9.6GB/s] IO bandwidth to devices]. On dual-processing Opterons, one link connects to the other processor, and two links connect to I/O devices.
As you can see above the difference from an athlon to opteron is that the opteron has at least twice the IO bandwidth of the athlon. Is 3.2GB enouth for two pci-e x16 slots, and 4x SATA controllers plus a gigabit nic? Its not just bandwidth but latency that is effected This increased bandwidth and lowered latency will help most benchmarks including games and any IO benchmarks.
Explain for the reader the computationally-significant difference that would alter the results of the benchmarking you’ve ignored completely. Or if you can’t, then find a benchmark demonstrating the difference between a FX-60, Opteron 185, Opteron 285, and Opteron 885 in a CrossFire configuration. Hint: this is a rigged exercise. Perhaps you’ll uncover my devious machinations and realize why your response was silly.
Conroe isn’t Woodcrest, and games aren’t really what one uses a server processor for. Conroe isn’t Clovertown either, which is the real target of his criticism. There are some comments about the Hexus and AnandTech game benchmarks that I could make, but I’d rather wait.
… as long i can put two of those in my (future) motherboard, im happy!
[homer]Mmmmm 8 core linux workstation[/homer]
Anyway I guess that Intel Kentsfield is what we can espekt the Intel PowerMac’s is gonna use to lure apples hardcore ppc fans to intel. My guess is a 2007 Q1/Q2 Intel PowerMac release!