AMD handily beats the Intel configuration: AMD 160 SPECfp vs Intel's 113.
I have a strong interest in engineering calculations which almost invariably use either finite-difference time-domain (FDTD) or finite-elements algorithms, and here the AMD system (at substantially lower clock rate) is more than twice as fast as the Intel-based system.
P.S. Newegg sells the Xeon X7350 CPU at $2500+ a pop.
I am not partial to either AMD or Intel, I have used many systems in the past from both. I am in the industry as many of you are and I get to see both vendors confidential roadmaps. From what I can see I doubt that AMD will produce ANY CPU in the near future and get it to market with any serious quantities that will be competitive with Intel.
I do agree that Intel needs to have a competitor but I don’t think it will be in the area of performance from AMD in the near future.
When Intel puts the memory controller on the processor later next year the gap will get even worse for AMD.
Only thing I see here is the same thing AMD posted on their site. They put numbers from cores they didn't even sell. Either way the whole thing is hard to understand without seeing it side by side.
Plus for regular people and not enterprises it would be best to see it in the real world benchmarks under Windows. I prefer Intel but AMD might have something here. But I doubt it will last as long as going to Nehalem.
Well first off, just because AMD doesnt sell the 2.5 Ghz doesnt mean they dont have one almost ready for launch. Also if you look at it, the amd system is at a disadvantage so that fact that it could keep up is pretty cool. 10K sata vs 7200rpm sata, 64 gigs ram vs 32gigs almost 3.0Ghz vs 2.5Ghz.
Also most compilers are already pretty well optimised for the intel processor, so even if it was sponsered by AMD they put their system at a disadvantage.
This shows that AMD can be competitive and that good things should be in our future. Cant wait till the quad cores come to us normal people.
Which system with half the ram and a slower HDD and clocked lower kicked the **** out of the overheated Zeon POS which draws heaps more power due to the northbridge and FB's??
That my friend would be the Barcy.
They are just as rare as hens teeth at present.
Hopefully the volume with increase and the Intel fanbois can eat some humble pie ... either in power consumption ... or power per watt ... or performance (providing the new quad scales well and can be produced at higher frequencies).
The end result of this will see Nehalem produced as a reality much quicker ... thats really good for everyone.
Intel will invariably produce a better product we can overclock.
AMD will keep them on their toes.
Make sure you tell the noobs to buy AMD so they can stay in business.
That way we can chose what to buy ... and not be limited to one chiptatorship.
SPEC is abbreviation for Standard Performance Evaluation Corporation, which compiles benchmarks for corporations to run on their servers. Since we're only dealing with processor here, I'll only get to CPU2006.
CPU2006 has four different tests, SPECfp, SPECint, SPECfp_rate, and SPECint_rate.
SPECfp and SPECint are benchmarks that test a CPU's IPC in both fp and int. In other words, those tests test the core of a processor, rather than the machine itself. This is a good indicator of single threaded performance.
Since these tests are only single threaded, the result is pretty much independent of memory (size / bandwidth), hard disk configuration, and so on.
On the other hand, SPECfp_rate and SPECint_rate are both "throughput" tests, as they test the machine as a whole. As a result, its a good indicator of the machine's performance. As you can imagine, these benchmarks stresses the processor (s), and RAMs. As for the base and peak scores, peak scores are obtained when compilers are used, and base scores are obtained when non-compilers are used.
This is the result OP posted (base / peak):
Intel Xeon 2.93 106 / 113
AMD Barcelona 2.5 149 / 160
WOWsers.... looks like Barcelona is the performance king!! OMG Intel is going banckrupt!!!
Now we look at, the pure processor performance
This is the result of AMD Barcelona 1.9Ghz vs. Intel Xeon 2.0Ghz.
SPECint = AMD 1.9GHz 9.97/11.3
Intel 2.0GHz 14.2/15.6
SPECfp = AMD 1.9GHz - 10.7/11.2 - Note 2 socket system
Intel 2GHz - 14.5/16.9 - Note 2 socket system
Oops... now the performance looks less than stellar. What is happening here?
It looks like on a core vs. core basis, Barcelona actually falls short of same clocked Xeon. Since the hardware configuration of those machine don't play a significant role in the benchmark, we can only assume that core for core, Barcelona performs lower than Xeon. In other words, Barcelona has lower instruction per clock (IPC) than Core 2.
Now, since SPEC_rate scores are based on the throughput of the machine, so having a higher score in SPEC_rate means Barcelona scales a lot better than Xeon, which is not surprising. AMD designed Barcelona as a native quad core, because by having a native quad core, the scalability will be a lot better than the MCM approach Intel used.
This is the SPEC_rate score for integer performance
SPECint_rate : AMD 2.0Ghz 72.8 / 83.2
Intel 2.0Ghz 67.7 / 72.6
As you can see, since int favors more to Intel, so the gap between performance is closed. Still, the excellent scalability of Barcelona is undeniable.
By scaling to quad cores from two cores, Barcelona gained 68% of the performance, while Xeon only gained 30%.
Now, since we have no idea when Barcelona 2.5Ghz is coming out (this November according to Hector), and the availability of current 2.0Ghz Barcelona is scant (you still can't order them from server vendors), we can only speculate the performance of 2.5Ghz Barcelona and beyond. It looks like if AMD can ramp up the clockspeed, Barcelona is very competitive in server segments.
So is Barcelona a good chip? Definitely. Will it save AMD? Not sure. Does it pwn Intel's entire offering? Not necessarily.
As for OP, I would strongly recommend you to go for Barcelona, as it will benefit you a lot more than Xeon, given its price and performance in the floating point segment. However, if you want better performance than 2.0Ghz, I'm afraid you're in for the wait.
We can also speculate Phenom's performance based on SPEC non rate scores, as desktop applications are generally single to dual threaded. If you take a look at the chart from Anandtech, you'll see that Barcelona only gained 59% by going from single to dual cores, while Core 2 gained a whooping 73%. This led me to believe that Phenom will not likely to be competitive against current Intel's offering. The battle for AMD will only get worse as they need to face the higher clocked, yet more efficient Penryn in November.