AMD dominates Supercomputers, used in more systems!!!

[thunderman]

Found this very interesting article...it appears AMD completely frags Intel where supercomputers are concerned. Intel just cannot compete. Look at the benchmarks quoted...AMD frags!

Article:
http://www.mysolutioninfo.com/news-display.aspx?Code=12854&t=AMD%20Dominates%20Top%2010%20Supercomputer%20Sites%20and%20Provides%20the%20Muscle%20for%20World

Article Quotes:

Bangalore, Karnataka, IND, 2008-11-18 (MySolutionInfo - IndiaPRwire.com) AMD (NYSE: AMD) today announced that its Quad-Core AMD Opteron™ processors are powering “Jaguar,” the first ever wholly x86-based supercomputer to achieve the petaflop performance milestone. AMD Opteron processors now help drive seven of the Top 10 supercomputer systems in the world

Today’s TOP500 Supercomputing Sites list reaffirms AMD’s leadership and the tremendous performance capability that HPC customers have enjoyed for years,” said Patrick Patla, vice president and general manager, Server Workstation Division, AMD. “Seven of the ten most powerful computing systems in the world now leverage the balanced platform of our Direct Connect Architecture. Our commitment to the HPC community only gets stronger with the recent launch of our 45nm Quad-Core AMD Opteron processor which has already set new standards in HPC performance.”

Just look at the Speed!!......Benchmarks below

Recently, AMD announced widespread availability of its 45nm Quad-Core AMD Opteron processors, codenamed “Shanghai.” These latest Quad-Core AMD Opteron processors already hold multiple HPC workload-related performance records including:
· 34 percent higher floating point throughput performance than the competition on SPECfp®_rate2006 for two-socket servers¹
· 51 percent more memory bandwidth performance than the competition on STREAM benchmarks for two-socket servers²
· Approximately 65 percent faster job runs than the competition on FLUENT 12 beta (ANSYS) for two-socket servers³
· More than 30 percent better performance than the competition on LSDYNA MPP971s for two-socket servers³.
Additionally, readers of HPCWire this week voted “Shanghai” as one of the Top 5 New HPC Products or Technologies to Watch.

I must say this is impressive! I can't describe my happiness....Intel has much work to do.

I can't wait for the release of 45nm AMD quads!

AMD4Life!!

 

[yomamafor1]

EDIT: A single Nehalem is faster than dual Shanghai in SPEC_int rate, and just 28% slower in SPEC_fp rate.

AMD's server days are numbered.

 

[toosober]

2 Words. Fan boy.

 

[Guest]

Yeah because they have an integrated memory controller... as soon as core 7 is released AMDs supercomputer power will come to a halt as the new power will be core i7

 

[piesquared]

Unless they continue to have cache coherency validation problems with their new nehalem architecture in 2p and up.

 

[yomamafor1]

cache coherency validation?

 

[piesquared]

Sorry, no link. There's a rumor going around that they are having trouble with coherency validation. It may turn out to be false, but either way it shouldn't mean anything to desktop users.

 

[yomamafor1]

Well you said "they continued to have cache coherency validation problems", so when was the last time Intel was affected with cache coherence issues? I've never heard of it ever since P4 w/HT.

 

[piesquared]

Well, if they are having validation problems now, then they are continuing to have problems, no?

 

[spathotan]

Who cares. None of us are using supercomputers.

 

[piesquared]

Here's something a bit more relevant, although still related to servers.

AMD's effervescent Leslie Sobon chats AMD Shanghai

 

[kyeana]

hey thunderman is back! Now how is it that i just knew who was going to be the starter of this thread...

Now... where is enigma?

 

[yomamafor1]

Well, if they are having validation problems now, then they are continuing to have problems, no?

Well I have not heard of any problems with cache coherency. It was a hot topic when Conroe was launched, and many suspected Conroe may suffer from cache coherency problem due to the size of the cache. However that turns out to be false.

So, I think what you're referring to is cache coherency problems induced by inclusive L3 cache and HyperThreading. As early test shows, this is simply not the case. With HT enabled, Nehalem was able to outperform dual Shanghai chips in integer performance, and was only 28% slower in floating point.

 

[eklipz330]

this thread needs a little concrum

 

[emp]

thunderman4Life!!! Down evil common sense!!!

Come on guys! Join the thunderman fan club.

 

[roofus]

This is the nail in the coffin for Intel. Think about how many of us build supercomputers! Intel is completely in the dark about this as their i7's are being grabbed up like they are free and not seeing the bigger threat of losing out on supercomputer market. R.I.P. Intel.

 

[enigma067]

http://www.enumae.com/images/Nehalem-965-900.png

EDIT: A single Nehalem is faster than dual Shanghai in SPEC_int rate, and just 28% slower in SPEC_fp rate.

AMD's server days are numbered.

http://www.youtube.com/watch?v=Ivx0XYMCZJw

http://www.youtube.com/watch?v=Xi4yxKNWehU

http://www.youtube.com/watch?v=_l2e0mf3CcA

http://www.youtube.com/watch?v=XbC16xJpzkA

Intel says, "AMD's beating our asses!"

 

[wild9]

'IBM Roadrunner Remains World's Fastest Supercomputer'

http://www.tomshardware.co.uk/IBM-Roadrunner-Top500-Supercomputer,news-29670.html

According to the recently released list of the world’s Top500 supercomputers, the title of world’s fastest supercomputer goes to IBM’s Roadrunner, located at the Los Alamos National Laboratory.

The Roadrunner was also the world’s first supercomputer to use a hybrid design, currently combining 12,960 IBM PowerXCell 8i CPUs and 6,912 AMD Opteron dual-core processors.

 

[roofus]

so there's the catch. 12,960 IBM CPU's.

 

[NMDante]

Seems like T-man's article just skipped a chunk of information about the Top 500 Supercomputer list.

Here are some other highlights from the November 2008 list:

* A total of 379 systems (75.8 percent) are now using Intel processors. This is virtually unchanged from six months ago (375 systems, 75 percent). Intel continues to provide the processors for the largest share of TOP500 systems.
* The IBM Power processors and the AMD Opteron family are almost tied as second most common processor family with 60 and 59 systems each (12 percent and 11.8 percent). Both had only minor changes from six months ago.
* Quad-core processor based systems have taken over the TOP500 quite rapidly. Already 336 systems are using them. 153 systems are using dual-core processors, and only four systems still use single core processors. Already seven systems use IBMs advanced Sony PlayStation 3 processor with nine cores.
* HP took over the lead in systems with 209 systems (41.8 percent) over IBM with 188 systems (37.6 percent). IBM had 210 systems (42.0 percent) six months ago, compared to HP with 183 systems (36.6 percent). But IBM still stays ahead by overall installed performance.

What is interesting is that they list 7 IBM systems using Sony's PS3s. I would never have guessed that they would be in the Top 500 Supercomputer list.

Top 500 SuperComputers

 

[Silverion77]

Nana nana nana nana

THUNDERMAN!!!!!!!!

 

[BeakerUK]

http://www.enumae.com/images/Nehalem-965-900.png

EDIT: A single Nehalem is faster than dual Shanghai in SPEC_int rate, and just 28% slower in SPEC_fp rate.

AMD's server days are numbered.

I would be more interested to see what the i7-920 does against the 2384. See what the similarly clocked machines do in comparison. You're comparing the 3.2Ghz i7 against a 2.7Ghz AMD. Also what are the power draws in the respective CPUs? the 965 as I understand it has a TDP of 130W, and the 2384 has a reported 75W power draw.

 

[yomamafor1]

The test also compares 1 Nehalem vs. 2 Shanghai processors.

TDP does not necessarily represents real world power usage, as it is only a number system designed for OEM as a reference for cooling solution. Since Nehalem's server parts haven't been launched, its difficult to compare with Shanghai on power consumption.

 

[BeakerUK]

the test compared 8 processing threads to 8 processing threads. Date over the interconnects is going to be slower than on-die anyway, also the ECC-DDR in the opty is going to hit performance at anything up to 10%. You're comparing Apples to Oranges by comparing an i7 desktop chip to a server board Opty setup.

 

[turpit]

Here's something a bit more relevant, although still related to servers.

AMD's effervescent Leslie Sobon chats AMD Shanghai

LMAO, Pie, man I may be wrong, but that looks the same techno incompetant cow (ok, she did lose a few pounds) that was pushing Phenom over youtube.


LMAO I just finished watching. ".... what we call 'switching'....." :lol: :lol: :lol: The only thing missing was the Dr Evil finger quotation marks. Shes still a clueless marketing cow quoting a promo sheet prepared by the real brains at AMD, and interspersing her own propaganda.


Man....AMD really needs to hang up the advertising and to get back to what it was good at....making a good product and letting the performance sell itself. The attempts at marketing just hurts them more than it helps.