Domain: specbench.org
Stories and comments across the archive that link to specbench.org.
Comments · 81
-
Re:What this mean...
Do you have any benchmarks to back that up? The ones I see from SPEC pretty much has Intel dominating in FP (compared to what few Itanium and POWER results there are).
-
Re:Well, from what I remember from the Keynotehttp://www.starlink.rl.ac.uk/star/docs/ssn23.htx/
n ode19.htmlWhile the results from this package should be in broad agreement with manufacturers' SPECmark ratings, they will provide a more realistic performance estimate for Starlink machines. SPECmark ratings tend to indicate the potential that it is possible to realise with a machine rather than the performance that will actually be returned when running `real' applications.
SPECMARK = Systems Performance Evaluation Cooperation Mark
http://www.specbench.org/ -
Re:These in-architecture tests are OK, but...
I'm interested too.
The closest I've found is SPEC CPU2000 results (warning, large page). Loads of various systems, put against a common benchmark, CPU2000. Should be details on the website about how it tests them.
There are a few links at this link to various CPU2000 results for specific tests - integer tests, floating point tests and throughput. -
Re:These in-architecture tests are OK, but...
I'm interested too.
The closest I've found is SPEC CPU2000 results (warning, large page). Loads of various systems, put against a common benchmark, CPU2000. Should be details on the website about how it tests them.
There are a few links at this link to various CPU2000 results for specific tests - integer tests, floating point tests and throughput. -
Re:What bothers me
I'm very interested in good OpenGL performance, since my Linux workstation is used for scientific graphics.
So are there any comparisons of rough benchmarks of such performance for recently-released videocards, along the lines of SPECviewperf?
-
Re:Maybe that's the answer...Unfortunately for apple fans, the number still stand on that site. Dell's own testing shows much better performance with the benchmark than Apple's does, and even makes the Dell machine win the benchmark. On the other hand, if they used GCC to compile the benchmark for the Dell machines, that might explain why they got such cruddy results. It's a widely accepted fact that GCC's code generation on CPUs with limited numbers of registers is pretty poor in terms of performance.
Of course, if you don't trust that website, how about ZDNet or even compare the numbers yourself. There's Veritest's Apple numbers, versus the offical published numbers from SPEC. There's also this site which goes into detail about the benchmark. They used -ffast-math on PPC, but not on x86, for instance. They explicitly turned off hyperthreading, which obviously hurt the Dell machine during the MP tests.
Then again, as the old saying goes, there's three types of lies. Lies, damn lies and benchmarks.
-
Re:ATX PowerPC
Actually, Alpha is being actively killed by HP as it would have wiped the floor with their new poster-child, Itanium. If I remember correctly, the more exact formulation was 'discontinued R&D, only one crippled architectural upgrade'.
As of SPARC, it looks to me like SPARC is kicking SPARC's ass - as in Fujitsu vs. Sun versions.
Finally, look at the difference between Itanium and x86/amd64-class CPUs. In theory EPIC is all fine and dandy, but ... why does it need 6MB of level 3 cache to show it? Makes one wonder how a comparable cache level would affect, say, a Xeon machine (given the obvious improvement that the extra cache brought to the P4EE CPUs). A similar story would hold for Power (see for instance this result, with 128MB off-chip L3 cache).
There might not even be a 'better way' to design a general-purpose CPU. Everybody has to optimize for something. Remember Intel bolting MMX then SSE/SSE2/SSE3 on x86 only because there was a heavy demand for it? And now, given the success of amd64, adding that as well to the Pentium4-class CPUs? x86 is not standing still. But that's the same for all the 'still alive' platforms. -
Re:There's always a trade-off
The fastest 64-bit AMD is rated 2.0ghz while the fastest AMD 32-bit is 2.2ghz.
First, that certainly doesn't mean the 64-bit chip is slower. AMD is not like Intel: they don't focus their design on stratospheric clock speeds, but on overall performance. The K8 core is faster cycle-per-cycle than the K7.Second, there is a 2.2GHz Opteron, as these SPECjbb2000 results show.
Third, even if there weren't, Moore's law tells us that a 10% clock speed difference amounts to, what, 11 weeks? So the K8 would be caught up by April. So who cares?
Using modern technology to build a 386 chip would result in one of the highest clock speeds ever but it would be practically useless.
This doesn't make any sense. If you use modern technology, it's not a 386, it's a Xeon. That's what a Xeon is.Maybe I just don't understand what you're saying.
-
Re:Not even platforms comparisonCare to back up the integer performance claims with a link?
I'm looking at spec.org and the SPECINT latest numbers between the closest comparable Opteron and Itanium2 systems are the 4 way results. 4 x 2000MHz Opteron get lower integer performance rates than 4 x 1500 MHz Itanics. Opteron 4way
Itanium2 4wayAlso explain to me why you think a webserver is
(a) necessarily CPU bound
(b) Is performant dependent on integer cpu performance.I would think a typical webserver is I/O bound. In the case of a "cpu bound" webserver likely your bottleneck is OS scheduling or other synch primitives in the OS. Disk subsystem and Memory will also dictate webserver performance.
-
Re:Not even platforms comparisonCare to back up the integer performance claims with a link?
I'm looking at spec.org and the SPECINT latest numbers between the closest comparable Opteron and Itanium2 systems are the 4 way results. 4 x 2000MHz Opteron get lower integer performance rates than 4 x 1500 MHz Itanics. Opteron 4way
Itanium2 4wayAlso explain to me why you think a webserver is
(a) necessarily CPU bound
(b) Is performant dependent on integer cpu performance.I would think a typical webserver is I/O bound. In the case of a "cpu bound" webserver likely your bottleneck is OS scheduling or other synch primitives in the OS. Disk subsystem and Memory will also dictate webserver performance.
-
Re:Here we go again:
Spec official benchmarks for the Opteron 142, Specfp_base2000 of 1029 vs Apples own published scores for the G5 2Ghz of 840.
-
Look at the actual numbers
If you look at the numbers Apple is claiming vs. the latest numbers posted at the SPEC site, then it would appear that the G5 is getting creamed by the Pentium 4 / Xeon.
First, single CPU performance. Apple claims 840 for SPECfp_base2000, and 800 for SPECint_base2000. A Dell Precision 360 with 3.2GHz P4 and DDR400 memory gives 1267 and 1242, respectively.
Next, dual-CPU. Apple claims 15.7 for SPECfp_rate_base2000 and 17.2 for SPECint_rate_base2000. A Dell Precision Workstation 650 with dual 3.06 GHz Xeons gives 18.0 and 25.6, respectively.
Of course, there are lies, damn lies, and benchmarks, but in this case I think it's fair to compare actual SPEC numbers with vendor claims.
And don't get me wrong, I think Apples are wonderful systems. I recommend them to many of my friends. But for raw CPU power, they lag the Intel powerhouse. -
Interesting numbers....http://www.specbench.org/cpu2000/results/rfp2000.
h tmlAnyone want to explain that?
Advanced Micro Devices: A4800 (1.4GHz Opterons, 4GB RAM ): 2 CPU: 21.0 base : 22.2 peak
Apple Computer: Dual PowerPC G5 2GHz: 2 CPU: 15.7 base: no peak
Dell: PowerEdge 2650 (3.06 GHz Xeon): 2CPU: 16.7 base:17.0 peak
Dell: Precision WorkStation 350 (3.06 GHz P4): 1 CPU: 12.7 base: 12.8 peakOr this: http://www.specbench.org/cpu2000/results/cfp2000.
h tml
Intel Corporation: D850MD motherboard (2.4 GHz, P4): 1 CPU: 806 base: 812 peak
Apple: G5 2GHz: 1 CPU: 840 base: no peak
HP: ProLiant DL380 G3(3.06GHz, Intel Xeon): 1 CPU: 1150 base: no peakI could go on, these just happen to be ones I selected. I abreviated some names (HP=Hewlett-Packard Company) and procs (P4) etc, but did not change the information, the numbers for apple were taken from http://www.apple.com/powermac/
-
Interesting numbers....http://www.specbench.org/cpu2000/results/rfp2000.
h tmlAnyone want to explain that?
Advanced Micro Devices: A4800 (1.4GHz Opterons, 4GB RAM ): 2 CPU: 21.0 base : 22.2 peak
Apple Computer: Dual PowerPC G5 2GHz: 2 CPU: 15.7 base: no peak
Dell: PowerEdge 2650 (3.06 GHz Xeon): 2CPU: 16.7 base:17.0 peak
Dell: Precision WorkStation 350 (3.06 GHz P4): 1 CPU: 12.7 base: 12.8 peakOr this: http://www.specbench.org/cpu2000/results/cfp2000.
h tml
Intel Corporation: D850MD motherboard (2.4 GHz, P4): 1 CPU: 806 base: 812 peak
Apple: G5 2GHz: 1 CPU: 840 base: no peak
HP: ProLiant DL380 G3(3.06GHz, Intel Xeon): 1 CPU: 1150 base: no peakI could go on, these just happen to be ones I selected. I abreviated some names (HP=Hewlett-Packard Company) and procs (P4) etc, but did not change the information, the numbers for apple were taken from http://www.apple.com/powermac/
-
Re:Lies, damn lies and benchmarksThe benchmarked configurations were Mac OS X on the G5, with GCC, and Linux on the PC gear, also with GCC.
But picking Linux as the test platform to run on is fair? I'd like to point out that 95% of all desktops run Windows. Also, what makes you think the scientific and technical users you refer to later on don't use Intel's C++/Fortran compilers for Linux when it matters? In my experience the speedup is quite dramatic, on the order of 30-40% faster - which is borne out also by Dell's posted SPEC benches: ~1200 for both int and float and 25(!) for rate (G5: 17). Ouch.
The only people who actually care about benchmarks are scientific and technical users. Everybody else cares about things like application availability, TCO, and ease of use.
Right. So why is Apple playing the performance flute so hard with the most unscientific comparisons I have ever seen right after air ionizers and snake oil?
-
Please look at the Dell results at the SPEC siteThis looks to be more of VeriTest again. If you look at the Third Quarter 2003 SPEC CPU2000 Results you will see that the Dells have CINT2000 scores in at least the 1100's. Apple gives the results for the Xeon and P4 as 836 and 889. While the Reg gives scores of 836 to the Pentium and 839 to the Xeon. The Reg article does not cite the source for the numbers in the article. I could not find anything I thought was new at Apple about this. Is this from the same statement produced by Apple with the VeriTest results from a while back? Maybe this was the first benchmark with the dual 2Ghz G5's?
In any case there was much consternation in the past about the VeriTest benchmarks becuase they did not use the same compilers that Dell used. Also VeriTest used things like an optimized malloc library on the G5's and faster memory with semi-secret memory timing tweaks in OF. If you want to take these benchmarks with a grain of salt, you should compare the DELL numbers from the SPEC site to those of the G5 from Veritest.
-
Re:BenchmarkingYeah, SPEC scores contain real world benchmarks, and assuming that the individual scores on the tests are pretty evenly distributed, the SPEc score would be a pretty good method of determining performance. But sometimes, a processor/chipset combo simply excels at one or two particular components of the benchmark. In that case, real world performance can be vastly different.
- Fugitsu Primepower 900 Score:1228/1004
- IBM eServer p630Score 1198/1097
- SGI Altix 3000 Score 1090/1082
The Fugitsu excels at art (image recognition using neural networks, C), the IBM at galgel(computaional fluid dynamics, F90), the SGi at swim (shallow water modeling, F77) and applu (partial differential equations, F77) All three have roughly similar aggregate scores. Yet the strengths of each patform lie in different areas.
(fwiw, the Powermac g5 showed strengths in mesa (3d image library, C), galgel and equake(seismic wave propagation simulation, C). The Intel machines, at least in Veritest's protocols, also performed strongly on those three tests.) - Fugitsu Primepower 900 Score:1228/1004
-
Re:BenchmarkingYeah, SPEC scores contain real world benchmarks, and assuming that the individual scores on the tests are pretty evenly distributed, the SPEc score would be a pretty good method of determining performance. But sometimes, a processor/chipset combo simply excels at one or two particular components of the benchmark. In that case, real world performance can be vastly different.
- Fugitsu Primepower 900 Score:1228/1004
- IBM eServer p630Score 1198/1097
- SGI Altix 3000 Score 1090/1082
The Fugitsu excels at art (image recognition using neural networks, C), the IBM at galgel(computaional fluid dynamics, F90), the SGi at swim (shallow water modeling, F77) and applu (partial differential equations, F77) All three have roughly similar aggregate scores. Yet the strengths of each patform lie in different areas.
(fwiw, the Powermac g5 showed strengths in mesa (3d image library, C), galgel and equake(seismic wave propagation simulation, C). The Intel machines, at least in Veritest's protocols, also performed strongly on those three tests.) - Fugitsu Primepower 900 Score:1228/1004
-
Re:BenchmarkingYeah, SPEC scores contain real world benchmarks, and assuming that the individual scores on the tests are pretty evenly distributed, the SPEc score would be a pretty good method of determining performance. But sometimes, a processor/chipset combo simply excels at one or two particular components of the benchmark. In that case, real world performance can be vastly different.
- Fugitsu Primepower 900 Score:1228/1004
- IBM eServer p630Score 1198/1097
- SGI Altix 3000 Score 1090/1082
The Fugitsu excels at art (image recognition using neural networks, C), the IBM at galgel(computaional fluid dynamics, F90), the SGi at swim (shallow water modeling, F77) and applu (partial differential equations, F77) All three have roughly similar aggregate scores. Yet the strengths of each patform lie in different areas.
(fwiw, the Powermac g5 showed strengths in mesa (3d image library, C), galgel and equake(seismic wave propagation simulation, C). The Intel machines, at least in Veritest's protocols, also performed strongly on those three tests.) - Fugitsu Primepower 900 Score:1228/1004
-
Re:It's important to know...The SPEC marks use only one CPU for SPEC_int and SPEC_fp.
Why doesn't Apple publish their marks on specbench.org? Why don't people look at
this:
http://www.specbench.org/cpu2000/results/res2003q2 /cpu2000-20030421-02108.html
or
this:
http://www.specbench.org/cpu2000/results/res2003 q2 /cpu2000-20030421-02109.html
The compiler?
Compiler: Intel C/C++ 7.0 build 20021212Z and
Intel Fortran 7.0 build 20021212Z,
Compaq Visual Fortran Compiler Version 6.6
Update B, Microsoft Visual Studio .NET (libraries)7.0.9466,
MicroQuill Smartheap Library 6.0
A few samples of the tweaks used to get peak?
C +FD0 -O3 -QxW -Qipo ;
C++ +FD0 -Qipo -GX -GR ;
Fortran +FD0 -O3 -QaxK -Qipo -Oi-
Apple should be forced to do full disclosure and publish results. I think SPEC should forbid the quoting of unpublished SPEC marks.
I think people need to better understand these benchmarks before commenting on them.
I personally consider the peak scored for INT and FP because the OS, the compilers, the compiler flags and libraries used are generally "real" and good for multipurpose general use -not the case in a certain Apple benchmark where a library was used that would be useless on a production system quote from an Extreme Tech article" Installed a high performance, single threaded malloc library. This library implementation is geared for speed rather than memory efficiency and is single-threaded which makes it unsuitable for many uses. Special provisions are made for very small allocations (less than 4 bytes). This library is accessed through use of the -lstmalloc flag during program".
What I find the most interesting thing to come of this whole mess is the fact than the Opteron produced some scores which challenge the Itanium 2 on Intel's own compiler. -
Re:It's important to know...The SPEC marks use only one CPU for SPEC_int and SPEC_fp.
Why doesn't Apple publish their marks on specbench.org? Why don't people look at
this:
http://www.specbench.org/cpu2000/results/res2003q2 /cpu2000-20030421-02108.html
or
this:
http://www.specbench.org/cpu2000/results/res2003 q2 /cpu2000-20030421-02109.html
The compiler?
Compiler: Intel C/C++ 7.0 build 20021212Z and
Intel Fortran 7.0 build 20021212Z,
Compaq Visual Fortran Compiler Version 6.6
Update B, Microsoft Visual Studio .NET (libraries)7.0.9466,
MicroQuill Smartheap Library 6.0
A few samples of the tweaks used to get peak?
C +FD0 -O3 -QxW -Qipo ;
C++ +FD0 -Qipo -GX -GR ;
Fortran +FD0 -O3 -QaxK -Qipo -Oi-
Apple should be forced to do full disclosure and publish results. I think SPEC should forbid the quoting of unpublished SPEC marks.
I think people need to better understand these benchmarks before commenting on them.
I personally consider the peak scored for INT and FP because the OS, the compilers, the compiler flags and libraries used are generally "real" and good for multipurpose general use -not the case in a certain Apple benchmark where a library was used that would be useless on a production system quote from an Extreme Tech article" Installed a high performance, single threaded malloc library. This library implementation is geared for speed rather than memory efficiency and is single-threaded which makes it unsuitable for many uses. Special provisions are made for very small allocations (less than 4 bytes). This library is accessed through use of the -lstmalloc flag during program".
What I find the most interesting thing to come of this whole mess is the fact than the Opteron produced some scores which challenge the Itanium 2 on Intel's own compiler. -
Editor Moron: The base SPEC marks: tsarkon reports
The base spec mark is very conservatively compiled, the peak is more of the all out see if you can "cheat" with the compiler. The official published results, something Apple hasn't done, very clearly discloses peak tweaks. And Opteron uses the FUCKING INTEL COMPILER to get its maximum SPEC mark. He got the fucking numbers from the published results at www.spec.org / www.specbench.org
. Why don't you try reading the tests on specbench.org and ask yourself this. Why the fuck has Apple NEVER published results there? Huh? Maybe you should look at this for a little clarity, asshole. Or this, a blinding SPEC_fp score.
Why are you stupid Mac zealots unable to accept that the G5 is a fucking farce and a lie, just as everything Apple has done in the way of performance since the days when the machines were truly more interesting the PC. Talk about riding your apron string on your dollar, zealots.
Why not talk about POWER 4+ or the upcoming POWER 5, accept that a piece of shit OS like OS X will never run on hardware of that power level, and move on with life. The best CPU for the buck and the best 4, 2 and 1 way systems will be Opteron for some time to come. You guys are like Intel pissing on the Opteron. Big marketing machines discrediting superior technology by pointing out corner cases and lying about benchmarks to look better.
FreeBSD or die - fo shizzle the nizzle biznatch fat sexless live in basement of parents jobless zealots. -
Editor Moron: The base SPEC marks: tsarkon reports
The base spec mark is very conservatively compiled, the peak is more of the all out see if you can "cheat" with the compiler. The official published results, something Apple hasn't done, very clearly discloses peak tweaks. And Opteron uses the FUCKING INTEL COMPILER to get its maximum SPEC mark. He got the fucking numbers from the published results at www.spec.org / www.specbench.org
. Why don't you try reading the tests on specbench.org and ask yourself this. Why the fuck has Apple NEVER published results there? Huh? Maybe you should look at this for a little clarity, asshole. Or this, a blinding SPEC_fp score.
Why are you stupid Mac zealots unable to accept that the G5 is a fucking farce and a lie, just as everything Apple has done in the way of performance since the days when the machines were truly more interesting the PC. Talk about riding your apron string on your dollar, zealots.
Why not talk about POWER 4+ or the upcoming POWER 5, accept that a piece of shit OS like OS X will never run on hardware of that power level, and move on with life. The best CPU for the buck and the best 4, 2 and 1 way systems will be Opteron for some time to come. You guys are like Intel pissing on the Opteron. Big marketing machines discrediting superior technology by pointing out corner cases and lying about benchmarks to look better.
FreeBSD or die - fo shizzle the nizzle biznatch fat sexless live in basement of parents jobless zealots. -
Editor Moron: The base SPEC marks: tsarkon reports
The base spec mark is very conservatively compiled, the peak is more of the all out see if you can "cheat" with the compiler. The official published results, something Apple hasn't done, very clearly discloses peak tweaks. And Opteron uses the FUCKING INTEL COMPILER to get its maximum SPEC mark. He got the fucking numbers from the published results at www.spec.org / www.specbench.org
. Why don't you try reading the tests on specbench.org and ask yourself this. Why the fuck has Apple NEVER published results there? Huh? Maybe you should look at this for a little clarity, asshole. Or this, a blinding SPEC_fp score.
Why are you stupid Mac zealots unable to accept that the G5 is a fucking farce and a lie, just as everything Apple has done in the way of performance since the days when the machines were truly more interesting the PC. Talk about riding your apron string on your dollar, zealots.
Why not talk about POWER 4+ or the upcoming POWER 5, accept that a piece of shit OS like OS X will never run on hardware of that power level, and move on with life. The best CPU for the buck and the best 4, 2 and 1 way systems will be Opteron for some time to come. You guys are like Intel pissing on the Opteron. Big marketing machines discrediting superior technology by pointing out corner cases and lying about benchmarks to look better.
FreeBSD or die - fo shizzle the nizzle biznatch fat sexless live in basement of parents jobless zealots. -
Re:Separate compiler from hardware?Apple are using their compilation and runs of the SPEC benchmark s/w in a way that (they claim) only tests the processor(s), but use the results to justify their claim of 'fastest personal computer' (ie 'system'), NOT, 'fastest processor', so I don't see it as a valid test.
From the SpecCPU 2000 Faq
Q4: What components do CINT2000 and CFP2000 measure?
A4: Being compute-intensive benchmarks, they measure performance of the computer's processor, memory architecture and compiler. It is important to remember the contribution of the latter two components -- performance is more than just the processor.
If Apple had turned out record setting Spec scores, it would be perfectly justified in assuming the mantle of "fastest personal computer." Right now, I think they have to settle for the title of "fastest personal computer using GCC", which is less than catchy. -
Re:More marketing games.
See SPEC's disclaimer.
SPEC doesn't do testing themselves. Testing is performed by a lab that licenses the SPEC benchmarks. They just decide what code lives in the benchmarks and coalesce the results. -
Amazing how people trust some no-named net personHe made so many errors in his "debunking" yet so many people took it for gospel.
Like, the switch -mfpmath=sse when used in a P4 *does* use SSE2, but this guy thought just cause the switch flag says sse that it must be SSE only.
Then someone else (can't find the post, on usenet, under the mac advocacy group) pointed out that Dell's SPEC tests also disabled hyperthreading.
Then, based on this person's web page who no one even knows who he is, they start drawing conclusions that if Apple faked these (based on his flawed analysis), that they also must have faked those Adobe, Mathmatica, and other demos -- despite the execs for those companies being on stage also confirming the results.
Gotta love the net...
As for me, I don't know what to believe. I'm just going to patiently wait until some reputable sites spend a lot of time and do an in depth analysis and their own benchmarks, like Tom's Hardware for example. Then I may start drawing my own conclusions.
As for me, all I want is to be able to encode mpeg video at something greater than real time. Show me *that* benchmark please!
-
Speed
-
Speed
-
Re:Yay!Since we're throwing out SPEC numbers, I thought I'd add some numbers I turned up on the SPEC WWW site.
These are actual measured values recorded at SPEC. I tried to pick the fastest of the listings for a given processor. Note that the Itanium I & II integer preformance problems are plainly evident as is the reason SGI is going to Itanium anyway. You can also see the benefits of the 800 MHz bus for the Pentium 4 and the reason why people question the 3200+ designation on the 2.2 GHz athlon.Hardware Clockspeed(GHz) SPECint2k SPECfp2k
Athlon XP 3200+ 2.20 1044 873
Opteron 144 1.80 1170 1122
P4(533 MHz bus) 3.06 1089 1053
P4(800 MHz bus) 3.00 1164 1213
Itanium 1 0.80 314 645
Itanium 2 1.00 807 1356
UltraSparc IIIcu 1.20 642 1074
MIPS R14000 0.60 483 499
IBM Power4 1.70 1077 1598
We don't have measured SPEC numbers for the PPC970, but we can compare the Power 4 from which it is derived. To match the integer performance of the Opteron@1.8GHz or the Pentium4@3.00GHz would take roughly a 1.8 GHz Power4. If we then compare the FP performance of these same processors, the Power4 would win by 40% over the P4 and 50% over the Opteron. And the Power 4 doesn't have a vector unit.
The open question, which will probably be answered by IBM on Monday, is how well the PPC970 matches up with its sibling. Remember the main differences between the PPC970 and the Power4 are a smaller L2 cache, Altivec, and a very different main memory architecture, so the real question is how well the HyperTransport bus feeds the processor. Given the low memory bandwidth/flop of the Power4s we have at work, it's not impossible that the PPC970 will be as fast or faster. This would affect the FP performance more than the int performance and the rumors are that the IBM estimates quoted by Ace's are conservative at least for FP. But we'll have to wait and see (hopefully not long).
Certainly once these systems hit market, unlike today, Apple will have hardware which is competitive for key markets like video and image editing, without appealing to Altivec.
-
Re:What about BSD?
I think you are comparing apples with oranges. SPECweb99_SSL is a benchmark that shows limit on a number of simultaneous connections for web server with SSL. Terry Lambert's tests are much simplier.
-
Sun is going down
_ALL_ Sun servers are very stable, but slow. SPARC speed is poor, take a look at SPEC CPU2000 Results. The memory bandwidthis _very_ low. In Linpack-top500 you won't see SUN in the 100 first places.
The Fujitsu SPARC64 V is better chip and 100% compatible with SUN solaris/SPARC. And better servers with 128 CPUs !!!!
LiNUX is a better alternative below 8 CPUs: Migrating Oracle9i - Based Sun Servers to Dell Servers Running Linux and Migrating Oracle9i - Based Sun Servers to Dell Servers Running Linux, Part 2. LiNUX+x86/ia64 , and soon AMD x86-64, is cheaper and faster than Solaris/SPARC
DEC/Compaq/HP have the best chip (Alpha EV7) and the best UNIX servers (ES47,ES80,GS1280) in RISC arch. It's a pity that Alpha is going to die to put intel ia64 instead.
And if you need NUMA machine, SGI Altix is for you.
Why do you need to buy a SUN server? -
Re:Won't fix Sun's biggest problem
_ALL_ Sun servers are very stable, but slow. SPARC speed is poor, take a look at SPEC CPU2000 Results. The memory bandwidth is _very_ low. In Linpack-top500 you won't see SUN in the 100 first places.
The Fujitsu SPARC64 V is better chip and 100% compatible with SUN solaris/SPARC. And better servers with 128 CPUs !!!!
LiNUX is a better alternative below 8 CPUs: Migrating Oracle9i - Based Sun Servers to Dell Servers Running Linux and Migrating Oracle9i - Based Sun Servers to Dell Servers Running Linux, Part 2. LiNUX+x86/ia64 , and soon AMD x86-64, is cheaper and faster than Solaris/SPARC
DEC/Compaq/HP have the best chip(Alpha EV7) and the best UNIX servers (ES47,ES80,GS1280) in RISC arch. It's a pity that Alpha is going to die to put intel ia64 instead.
And if you need NUMA machine, SGI Altix is for you.
Why do you need to buy a SUN server?
- because my programs _only_ run with solaris/sparc -
SGI should be put out of its miseryNot intending to start a Holy War, I realize the 64 CPU monsters have their place but their workstations are just ignorant (this is coming from a previous SGI only owner)...
"These systems were around $40,000 when first released. Each R12000 400MHz has a SpecFP2000 of around 350-360 and so it's approximately equal to an Athlon 1.2GHz. The caveat is that the SpecFP2000 benchmark is actually made up of a bunch of other, smaller, tests. For computational fluid dynamics or neural network image recognition, the 400MHz SGI CPU is 2.5 to 5 times faster than the Athlon!"
WOW! 2.5 times faster than a 1.2Ghz Athlon!? Man, you'd almost need a $168 2.4 Ghz Athlon to keep up! I wish they made them!
P.S. The 3.06 Ghz P4 is just under 1000 on the SpecFP benchmark.
-
Macs are slow and expensive for cpu intense ...
processes but fast in terms of the actual work you get done.
There is no need to take an either or attitude to PC vs Mac. They play well together...
I do a certain amount of video editting. Doing the actual editting is faster on the Mac with Final Cut Pro vs Premier / Pinnacle / Ulead on the PC ( I've tried them all in an effort to stay with only one platform.)
Now doing the actual rendering on the Mac is slow....... Check out
mac vs HT PC
The above article far understates the true advantage of Intel hardware. To truly match the
Apple architecture with Intel hardware. Go to
Dell and configure a dual 2.8 Ghz Xeon Workstation with Hyperthreading on each chip. Now you have 4 processors working for you , two real and two virtual per CPU. For $3200 you get two 2.8 Ghz Xeons with 1 gig of RAM, 2 80 gig harddrives (separating the system disk from your video disk is important) and a 4x dvd+rw drive.
Still think the G4 dual processor Mac is great? Why not use the industry standard to measure your chip of choice. In the supercomputing world how fast your machine runs is more than just bragging rights, it's job security. For that reason the SPEC benchmarks were created to get standardized validated results on any hardware. Mac OS X has a SPEC suite see
Mac SPEC
Now that you've looked at that go to SPEC and look at the CPU benchmarks. Note the scaling factor for the Xeon 2.8 Ghz with two processors
SPEC CPU
Excerpt of CPU INT multiprocessor
Chip Result
2.8 GHz Xeon 10.2
2 CPU 2.8 GHz Xeon 18.0
The SPEC is designed to show good scaling with parallelism (multiple CPU) and here shows a 1.8 scaling factor.
So your Dell machine with HT will have a greater than 2 scaling factor for highly parallel processes.
NB: I only have a 2.4 Ghz P4 as my rendering machine and it's still faster than the Powermac by enough to make me stick with the mixed network approach.
So work on the Mac -> DV over gigabit ethernet to a multiprocessor Intel dual processor machine that renders AND burns the DVD, VCD, SVCD faster.
We all know Macs are better at what matters during the creative process. Let the Intel hardware bear the drudgery ;-) -
Re:RDRAM vs. DDR
>On the other hand, interleaved RDRAM has the same >peak theoretical bandwidth of interleaved DDR >SDRAM one quarter its clock rate because RDRAM >chips have one quarter the bus width of SDRAM >chips. 800 MHz RDRAM would be the same speed as >200MHz SDRAM assuming both (or neither) are >interleaved. Many if not most modern SDRAM >controllers support memory interleaving, >including my old abit board which isn't even new >enough to run an athlon XP.
Theoretical performance doesn't hold up in practical applications. That is why there is SPEC. Besides, RDRAM is designed so that you can easily make it a multi-channel architecture:
Quote: "By moving the core logic to the CPU and thus incorporating the Rambus Memory Controller as a part of the CPU itself, much of the current latency problems plaguing the technology will disappear. Both Sun's upcoming MAJC and the Playstation 2 are examples of embedded solutions with ondie RMCs. Another example is Compaq's upcoming EV7 (Alpha 21364), which also uses 8 channels to support massive bandwidth requirements and to keep latencies down (instead of accessing large volumes of DRDRAM in serial from a single channel, which would increase latencies)."
DDR SDRAM is more complicated to design in a multi-channel layout because of timing and motherboard design complexity issues. Besides, DDR SDRAM can't sustain its peak bandwidth even close to as well as RDRAM. Bursting isn't sustained bandwidth.(Remember, to transfer a byte, you have to transfer a QWORD. To transfer the next byte, you have to wait an entire clock cycle). Therefore, it only really allows 1/4 the peak bandwidth in this case.
A very old article from Ace's Hardware says: "The Rambus channel sends out the data twice as fast as the SDRAM, but the SDRAM can send out the first 8 byte without waiting, while Rambus has to transfer 16 bytes. As Rambus can send 2 bytes every cycle, it takes 4 cycles of 2.5 ns to transfer 16 bytes or 10 ns."
Of course, this is regarding 800Mhz RDRAM, and 1066Mhz RDRAM is currently out.
Also, not many applications only send 16 bytes at a time. For random access bursting applications, like servers, this is common. On the other hand, for 3d games, you want a sustained bandwidth to send all that data to the graphics card.
The aforementioned site speaks of examples where DDR SDRAM would be better because of its lower latency. In those cases, though, RDRAM will still be adequate.
Also, I'm curious to see any mentions of RDRAM versus DDR performance on newer chipsets.
> The benchmark you cite shows better performance > for DDR SDRAM on intel's solutions than on the
> athlon system, which leads me to believe that
> it is possibly cpu-dependent.
Since memory bandwidth depends on FSB, Intel systems should have an advantage.
>The belief that RDRAM is a superior technology to >SDRAM is at best a matter of opinion and at worst >an absurd myth.
Didn't see you try to refute the fact that unused cores shut down (great for power management).
> RDRAM makes great sense when you're not using >very much of it (down in the ~16 MB range...
Comparison: 1GB system.(Picking anything more would severely skew things in RDRAM's favor).
PC3700 DDR vs PC1066 RDRAM
source of prices: pricewatch.com
Generally RDRAM comes with two channels, while DDR generally comes with one.
PC1066 RDRAM theoretical peak bandwidth (2 channels)= 1066*2*2 = 4.3 GB/s
DDR3700 DDR theoretical peak bandwidth (1 channel)= 466 * 8 = 2.98 GB/s
Price:
DDR (512MB modules * 2) = $145 * 2 = $290
RIMM (256MB modules * 4) = $80 * 4 = $320
Yes, the motherboards for RDRAM based sets are more expensive, but the memory appears to be bearly more expensive in this case for twice the theoretical bandwidth. The RDRAM 512MB is more expensive than DDR 512MB, but since you usually find RDRAM in higher-channel configurations, this isn't an issue. Comparing prices is difficult since you have to pair RDRAM up.
-
Re:It's hellaciously fast
sorry, nosferatu-man, but those marks are misinterp'd
As Rob Young pointed out in his post:
"Not sure what you cobbled together but threads are your CPU counts. All the EV7 results are for 16 CPUs"-RY
Tester Name System Name CPUs Threads Base Peak
Compaq Computer Corp AlphaServer GS320 Model 32 64/731 16 16 5073 --
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 4 6027 6824
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 8 10349 11929
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 16 17420 20066
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 4 5482 6324
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 8 10040 11547
Hewlett-Packard Comp AlphaServer GS1280 Model M16 16 16 17724 20637
Info from here. -
SpecOMP (link)
Info on SpecOMP, just in case anyone's interested. Also, here's a snippet from the FAQ:
Q3: What components does SPEC OMP measure?
A3: Since the benchmarks are designed to reflect applications requiring compute-intensive parallel processing, they measure performance of the computer's processors, memory architecture, operating system, and compiler. It is important to remember the contribution of the latter three components.
'jfb -
Re:SPEC CPU2000
The 3.06 GHz Hyper-Threaded Pentium 4 performs even better than the Xeon I listed.
Dell Precision WorkStation 350 (Pentium 4)
3067 GHz
SPECint2000 = 1130
SPECint_base2000 = 1085
SPECfp2000 = 1103
SPECfp_base2000 = 1092
CINT2000
CFP2000 -
Re:SPEC CPU2000
The 3.06 GHz Hyper-Threaded Pentium 4 performs even better than the Xeon I listed.
Dell Precision WorkStation 350 (Pentium 4)
3067 GHz
SPECint2000 = 1130
SPECint_base2000 = 1085
SPECfp2000 = 1103
SPECfp_base2000 = 1092
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
SPEC CPU2000
CINT2000 and CFP2000 results from the Standard Performance Evaluation Corporation website.
AlphaServer ES45
1250 MHz
SPECint2000 = 928
SPECint_base2000 = 845
SPECfp2000 = 1365
SPECfp_base2000 = 1019
CINT2000
CFP2000
HP server rx5670 (Itanium 2)
1000 MHz
SPECint2000 = N/A
SPECint_base2000 = 807
SPECfp2000 = 1431
SPECfp_base2000 = 1431
CINT2000
CFP2000
Dell Precision WorkStation 530 (Xeon)
2800 MHz
SPECint2000 = 957
SPECint_base2000 = 921
SPECfp2000 = 887
SPECfp_base2000 = 878
CINT2000
CFP2000 -
Re:Behind the times.I thought enough material had finally invaded the net for people to realize Mhz means nothing... I guess I was wrong.
You mean Apple ads?
Seriously, what 'material' are you talking about? I know about SPEC, according to which the currently fastest CPU is the Itanium 2 1000 MHz, followed closely by the PIV 3.06 GHz. From that I would deduce that even if you've got a relatively slow CPU (in terms of computations per clock cycle), if you manage to run it at very high frequencies, you'll still have one of the fastest CPUs out there.
-
Re:RISC
Ok, when you get to the Real World, let us know.
Switching architectures is not that trivial. You seem to think that every company has the source code available for every piece of software they run. That isn't true. You seem to think that programs can easily be compiled between programs if written in C/C++ - also untrue. You think that the bug fixes for compiling between platforms are "small peculiarities" -- well, they may be small, but that doesn't make them easy. In fact, it makes it fucking hard because the differences are so buried in libraries, case-specific, and undocumented that it's a nightmare to find them. Yes, I've done this kind of thing. It's godawful.
Changing architecture is difficult. This is not a closed vendor market - anyone can put together an x86 box and you have at least 3 different CPU vendors to chose from, 3 - 5 motherboard chipsets, and a virtually infinite variety of other hardware. If Dell computer suddenly decides to move to a PPC architecture what's going to happen? They're going to lose all their customers and fast. Because the very limited benefits of a different architecture do not make up for the costs of going to one.
Yes, I said limited benefits. Yeah, when I was in college taking CompE, EE, and CS courses on CPU and system design I also found the x86 ISA to be the most demonic thing this side of Hell. Well, I'm older and wiser now and while x86 isn't perfect, it's not that bad either. It's price/performance ratio is utterly insane and getting better yearly. Contrary to the RISC architecture doom and gloomers, x86 didn't die under it's own backwards compatibility problems. It's actually grown far more than anyone expected and is now eating those same manufacturers for lunch.
You know, back in the early 90s when RISC was first starting to make noise the jibe was that Intel's x86 architecture was so bad because it couldn't ramp up in clock speeds. Intel was sitting at 66 MHz for their fastest chip while MIPS, Sparc, etc. were doing 300 MHz. Of course, now Intel has the MHz crown, with demonstrations exceeding 4 GHz, and the RISC crowd is saying that MHz isn't everything and they do more work/cycle than Intel (which is true, but the point remains).
All that said, go look at the SPEC CInt2000 and FP2000 results. Would you care to state what system has the highest integer performance? And whose chip has the highest floating point?
Oh, and let's not forget that I can buy roughly 50 server-class x86 systems for the price of one mid-level Sun/IBM/HP/etc. server.
Note - server performance isn't all about CPU, but since the OP wanted to make that argument, I just thought I'd point out how wrong he is. There is still quite a bit of need for high end servers with improved bus and memory architectures, but don't even try to argue that the CPU is more powerful. It isn't. -
Re:RISC
Ok, when you get to the Real World, let us know.
Switching architectures is not that trivial. You seem to think that every company has the source code available for every piece of software they run. That isn't true. You seem to think that programs can easily be compiled between programs if written in C/C++ - also untrue. You think that the bug fixes for compiling between platforms are "small peculiarities" -- well, they may be small, but that doesn't make them easy. In fact, it makes it fucking hard because the differences are so buried in libraries, case-specific, and undocumented that it's a nightmare to find them. Yes, I've done this kind of thing. It's godawful.
Changing architecture is difficult. This is not a closed vendor market - anyone can put together an x86 box and you have at least 3 different CPU vendors to chose from, 3 - 5 motherboard chipsets, and a virtually infinite variety of other hardware. If Dell computer suddenly decides to move to a PPC architecture what's going to happen? They're going to lose all their customers and fast. Because the very limited benefits of a different architecture do not make up for the costs of going to one.
Yes, I said limited benefits. Yeah, when I was in college taking CompE, EE, and CS courses on CPU and system design I also found the x86 ISA to be the most demonic thing this side of Hell. Well, I'm older and wiser now and while x86 isn't perfect, it's not that bad either. It's price/performance ratio is utterly insane and getting better yearly. Contrary to the RISC architecture doom and gloomers, x86 didn't die under it's own backwards compatibility problems. It's actually grown far more than anyone expected and is now eating those same manufacturers for lunch.
You know, back in the early 90s when RISC was first starting to make noise the jibe was that Intel's x86 architecture was so bad because it couldn't ramp up in clock speeds. Intel was sitting at 66 MHz for their fastest chip while MIPS, Sparc, etc. were doing 300 MHz. Of course, now Intel has the MHz crown, with demonstrations exceeding 4 GHz, and the RISC crowd is saying that MHz isn't everything and they do more work/cycle than Intel (which is true, but the point remains).
All that said, go look at the SPEC CInt2000 and FP2000 results. Would you care to state what system has the highest integer performance? And whose chip has the highest floating point?
Oh, and let's not forget that I can buy roughly 50 server-class x86 systems for the price of one mid-level Sun/IBM/HP/etc. server.
Note - server performance isn't all about CPU, but since the OP wanted to make that argument, I just thought I'd point out how wrong he is. There is still quite a bit of need for high end servers with improved bus and memory architectures, but don't even try to argue that the CPU is more powerful. It isn't. -
Re:Is it me....
not true. check out some spec scores... "real" workstations (where "real" often just means "real expensive") only become truly faster once you're spending *big* money. a cluster of PCs these days will outperform anything, $ for $, on any CPU-intensive task, and IMHO, are The Way To Go.
empirically, we have low-end (dual) IBM p620's and p660's, and for our (CPU-intensive) applications, they are slower than most of my teams' desktops, whilst managing to be over an order of magnitude more expensive per unit.
matt