SuperPi A Useful Benchmark?

[Faikius]

We all know it's a decent stability test but is SuperPi a meaningful benchmark?

I guess what I'm asking is "Is it a realistic gauge of processor performance?"

I'm getting 32s for the 1M test with the setup in my sig. It look like I'm beating out quite a few other systems but does it really mean anything?

 

[themusgrat]

Only of raw performance. I don't think that it takes AMD's mem controller in to account, so OCed Intels actually have a chance. It does not reflect real world performance. But I guess if the PC is a physics etc. related workstation, it might reflect that kind of performance.

 

[liebremx]

It's meaningful taking into account the type of workload SuperPI generates on the system: mainly stresses the CPU (FPU), branch prediction probably and the cache. Beyond that it doesn't say much about the rest.
I'd say is as realistic as what you are interested in measuring. If you want an idea of overall performance of your system then it's not a meaninful benchmark. If you're interested in FPU performance alone, it is.

 

[Green Man]

Originally posted by: Faikius
We all know it's a decent stability test but is SuperPi a meaningful benchmark?

I guess what I'm asking is "Is it a realistic gauge of processor performance?"

I'm getting 32s for the 1M test with the setup in my sig. It look like I'm beating out quite a few other systems but does it really mean anything?

Superpi is worse than useless because it gives an impression that can be very far from real world applications. e.g. If someone bought a Dothan based system because of superpi 1m, they would be extremely disappointed.

 

[dexvx]

Originally posted by: Green Man
Superpi is worse than useless because it gives an impression that can be very far from real world applications. e.g. If someone bought a Dothan based system because of superpi 1m, they would be extremely disappointed.

Disappointed in what? Dothan is extremely competitive in a lot of real world applications against Pentium-4's and Athlon64's.

And your definition of "real world" applications is far fetched. Real world applications are very different from each other. SuperPi is representative of some real world applications and not representative of others.

 

[dguy6789]

Originally posted by: dexvx

Originally posted by: Green Man
Superpi is worse than useless because it gives an impression that can be very far from real world applications. e.g. If someone bought a Dothan based system because of superpi 1m, they would be extremely disappointed.

Disappointed in what? Dothan is extremely competitive in a lot of real world applications against Pentium-4's and Athlon64's.

And your definition of "real world" applications is far fetched. Real world applications are very different from each other. SuperPi is representative of some real world applications and not representative of others.

:roll:


You clearly have not looked at some real world benches.

 

[furballi]

It's a very useful number crunching application that excludes the video processor. It can detect small changes in CPU and RAM settings. I use it if I want to see a quick feedback on these changes. Make sure you get the SP mod that calculates out to three decimal places.

Use it to estimate the performance between PCs, minus the video processor. I would not use Super Pi as a stability test. Super Pi will get you into the ball park, but you should use Prime95 and Memtest for fine-tuning. A high-end overclocked AMD system will break the 30 sec barrier.

Here are a few datapoints for the AMD-based CPUs:

-Athlon 64 3000 (Winchester) overclocked to 2.35GHz...35 sec
-Sempron 3100 (130nm) overclocked to 2.39GHz...39 sec

An Opti 144 or 146 should be able to hit 32 sec or lower (+2.6GHz) at a fraction of the cost of your Intel. A hard-core gamer should stick with the AMD Athlon 64 or Opti platform. I don't see a huge performance gain with dual-core to warrant the additional cost. All others should do well with the overclocked Sempron at $90 (CPU and motherboard combo at Fry's).

I've just completed building a Sempron system. The system can hit 36.5 sec if I max out on Vcore and use decent overclocking RAM.

 

[Faikius]

Originally posted by: furballi
An Opti 144 or 146 should be able to hit 32 sec or lower (+2.6GHz) at a fraction of the cost of your Intel.

I would have to disagree with part of that statement. While I can't speak to AMD performance numbers, the Opti 146's are going for $232 vs. the D920 at $261.

$29 difference

 

[furballi]

The 146 (939) was selling for $165 earlier this month. Monarch had the 144 on sale this week for $130 with the purchase of a MB. If you have an okay MB, and a good 144 chip, then you should be able to hit 2600MHz. The only difference between the 144 and 146 is the 9x vs 10x CPU multiplier. I have yet to see a 144 that cannot hit 2400MHz (33.4 sec Super Pi). At this speed, the true limitation is the video card, not the CPU.

Have seen one Athlon 64 Venice 3200 running at 2670MHz...30.2 sec Super Pi.

 

[dexvx]

Originally posted by: dguy6789

Originally posted by: dexvx

Originally posted by: Green Man
Superpi is worse than useless because it gives an impression that can be very far from real world applications. e.g. If someone bought a Dothan based system because of superpi 1m, they would be extremely disappointed.

Disappointed in what? Dothan is extremely competitive in a lot of real world applications against Pentium-4's and Athlon64's.

And your definition of "real world" applications is far fetched. Real world applications are very different from each other. SuperPi is representative of some real world applications and not representative of others.

:roll:

You clearly have not looked at some real world benches.

Clearly

http://www.x86-secret.com/articles/divers/ct479/ct479-4.htm
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=12

 

[dguy6789]

You missed the point. Game performance is just about the ONLY place Dothan excels.

 

[dexvx]

Originally posted by: dguy6789
You missed the point. Game performance is just about the ONLY place Dothan excels.

Oh really?

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=10

Seems to do pretty decent in Spec, aka workstation performance.

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2382&p=11

And it seems to do middle of the road in Media Encoding.

And btw, since when is Game Performance a non-Real World benchmark?

 

[dguy6789]

The point of the discussion is this: If someone bought a Dothan based system based solely on the performance it gets in Super Pi compared to other processors, they would be very disappointed because it does not represent how it really performs compared to other processors. For the performance a Dothan gives out, you could get something as fast for less money, or faster for the same money. This is why someone would be dissapointed if they bought a Dothan system after looking only at Super Pi as a benchmark.

Perhaps I was a bit misleading by only directly stating benchmarks, when in reality I was speaking of performance per dollar.

 

[dexvx]

Originally posted by: dguy6789
The point of the discussion is this: If someone bought a Dothan based system based solely on the performance it gets in Super Pi compared to other processors, they would be very disappointed because it does not represent how it really performs compared to other processors. For the performance a Dothan gives out, you could get something as fast for less money, or faster for the same money. This is why someone would be dissapointed if they bought a Dothan system after looking only at Super Pi as a benchmark.

Perhaps I was a bit misleading by only directly stating benchmarks, when in reality I was speaking of performance per dollar.

Performance per dollar in what?

Notebooks - As a complete package, most people find Dothan notebooks to be as cheap as or cheaper than Pentium-4/Athlon64/Turion based notebooks of a comparable configuration. In fact, if you go on the Hot Deals Forum, you can consistently find Inspiron books for very competitive prices.

Desktop - It's not meant for the desktop. It requires a specialized motherboard or an Adapter, courtesy of Asus. However, it does come with one plus side. Since it's considered a notebook component, it can be had for a heavily discounted price (especially used), courtesy of eBay. I got a new PCI-Ex16 Asus P4GPL-X ($70) + new Adapter ($50) + new Dothan 1.6Ghz ($70, clocking at 2.7Ghz now with the dinky stock cooler) for a mere $190 total.

 

[dguy6789]

Originally posted by: dexvx

Originally posted by: dguy6789
The point of the discussion is this: If someone bought a Dothan based system based solely on the performance it gets in Super Pi compared to other processors, they would be very disappointed because it does not represent how it really performs compared to other processors. For the performance a Dothan gives out, you could get something as fast for less money, or faster for the same money. This is why someone would be dissapointed if they bought a Dothan system after looking only at Super Pi as a benchmark.

Perhaps I was a bit misleading by only directly stating benchmarks, when in reality I was speaking of performance per dollar.

Performance per dollar in what?

Notebooks - As a complete package, most people find Dothan notebooks to be as cheap as or cheaper than Pentium-4/Athlon64/Turion based notebooks of a comparable configuration. In fact, if you go on the Hot Deals Forum, you can consistently find Inspiron books for very competitive prices.

Desktop - It's not meant for the desktop. It requires a specialized motherboard or an Adapter, courtesy of Asus. However, it does come with one plus side. Since it's considered a notebook component, it can be had for a heavily discounted price (especially used), courtesy of eBay. I got a new PCI-Ex16 Asus P4GPL-X ($70) + new Adapter ($50) + new Dothan 1.6Ghz ($70, clocking at 2.7Ghz now with the dinky stock cooler) for a mere $190 total.

I believe we were speaking of desktop. I would not buy anything other than a Dothan for a laptop as of now. I find them great for laptops as they provide good performance while conserving power. But as far as pure performance goes in the desktop arena, there are better options

 

[Green Man]

My point wasn't to denigrate the Dothan processor. I simply mean that Dothan is very competitive in some applications, with a few shortcomings. In superpi 1m, however, it is absolute king. If I base my expectations on superpi alone, I will be sadly disappointed to find out the the system is merely decent.
Make no mistake, superpi is a synthetic benchmark.

 

[Megatomic]

I hate to ask an OT question in here, but I'd hate to create a thread to ask this one question (I did a rudimentary search btw):

Is it possible to successfully run two instances of SuperPi 32M on a dual core chip? I've tried and I get an error (same one always) on one core (just one and it's either one, it's random) every time I've tried to do this. I can do 2 x Prime95 so I'm a little confused by my results.

 

[nyker96]

I think superpi is probaly not too useful as real world app test. Since it's not stressing to RAM, only computation: just FPU I think id they using the normal algs for Pi like generating functons. Probaly not much use.

 

[zanejohnson]

my 930d @ 3.6GHz does it 2x in 34s

 

[zephyrprime]

Superpi is pretty useless in my opinion. Why on earth we use something like that to test systems boggles my mind. It's not even very modern program is it? Even if one wanted to do a specific processor only benchmark, using something like cpumark makes a lot more sense.

 

[orangat]

Originally posted by: zephyrprime
Superpi is pretty useless in my opinion. Why on earth we use something like that to test systems boggles my mind. It's not even very modern program is it? Even if one wanted to do a specific processor only benchmark, using something like cpumark makes a lot more sense.

I don't see how it is useless. CPUmark is a composite synthetic benchmark while Superpi is a real world benchmark that actually does something.

 

[the cobbler]

Originally posted by: liebremx
It's meaningful taking into account the type of workload SuperPI generates on the system: mainly stresses the CPU (FPU), branch prediction probably and the cache. Beyond that it doesn't say much about the rest.
I'd say is as realistic as what you are interested in measuring. If you want an idea of overall performance of your system then it's not a meaninful benchmark. If you're interested in FPU performance alone, it is.

actually SuperPi times/stability are largely dependent on RAM. moreso than CPU, which is why SP32M is the "standard" RAM stability test.