How many GFLOPs for i5-3570k?

[galego]

I got a doc from Intel giving 112 GFLOPs for i7-3770k (not turbo), but could not find a similar doc for the i5. Someone knows some source?

If none is available could we estimate that the i5 will be behind 100 GFLOPs taking in consideration the slower clock and the lack of HT?

 

[Puppies04]

Depends which bench you use to measure it, some benchmarks actually return lower numbers with HT than without.

 

[dave_the_nerd]

Depends on the benchmark. (Some don't like hyperthreading, actually.)

I'd be happy to bench mine when I get home, but afaik, they should be close, and you can probably find Intel Burn Test GFLOP results with a bit of googling, given how many people like the i5-3570k.

Except for any effect from hyperthreading, performance should be a direct reflection of clock speed and not much else. (Otherwise the chips are the same, right? Same cache, architecture, etc.)

 

[TuxDave]

I got a doc from Intel giving 112 GFLOPs for i7-3770k (not turbo), but could not find a similar doc for the i5. Someone knows some source?

If none is available could we estimate that the i5 will be behind 100 GFLOPs taking in consideration the slower clock and the lack of HT?

Are you looking for theoretical double precision gflops? Doing some backward calculations, I'm wondering if you get 112 from the math below or from some benchmark (which could be lower than what I have listed).

i7 3770K = 3.5GHz * 4 DP Vectors * 2 uops/cycle * 4 cores = 112 gflops

You can change the frequency and # of cores for your i5.

 

[Charles Kozierok]

http://download.intel.com/support/processors/corei5/sb/core_i5-3500_d.pdf

 

[Phynaz]

If none is available could we estimate that the i5 will be behind 100 GFLOPs taking in consideration the slower clock and the lack of HT?

HT wouldn't affect the theoretical GFLOPS number, that is fixed based upon architecture.

What it affects is your ability to use those GFLOPS.

 

[pantsaregood]

112 GFLOPs seems a little high, unless it is entirely theoretical. My i5-2500K gets 120 GFLOPs at 4.4 GHz.

 

[TuxDave]

112 GFLOPs seems a little high, unless it is entirely theoretical. My i5-2500K gets 120 GFLOPs at 4.4 GHz.

I'm pretty sure we're all quoting theoretical numbers and not actual. At least the 112 number matches the calculation I showed above.

 

[galego]

Yes I was asking about theoretical (peak) value, not about the measured using some concrete benchmark. Charles gave just the link to what I was searching. The value given by Intel is 109 GFLOP.

Thanks to everyone.

 

[Exophase]

Also note that peak SP FLOPs is double that.

 

[Charles Kozierok]

And was exactly what TuxDave said, by the way, so you can find the answers for any other CPUs in the same line the same way.

 

[ShintaiDK]

I got a doc from Intel giving 112 GFLOPs for i7-3770k (not turbo), but could not find a similar doc for the i5. Someone knows some source?

If none is available could we estimate that the i5 will be behind 100 GFLOPs taking in consideration the slower clock and the lack of HT?

A stock 3570 with 16GB 1600Mhz memory gets around 102GFlops in Linpack.