Jump to content
Forum upgrade is live! Read more... ×
mark84

Atomic Superpi Scores V5

Recommended Posts

Carrying on from the great work done by morgoth and Chipped HERE and The_Hawk HERE.

Important Notes!! - Be sure to read
- One entry per person per CPU category
- Only submit correct iterations. 1M and 32M for SuperPi, 8M and 256M for MaxxPi and 1B for y-cruncher.
- Include a CPU-Z window in your screenshot showing CPU specs.
- Posts MUST include a screenshot

SuperPi
- SuperPi screenshot with window showing calculation complete (with check sum visible) and CPU-Z open with CPU stats visible
- No longer Accepting PI versions that do NOT have millisecond output
- You must use the SuperPI Mod 1.5 from HERE
- Entries should be able to be validated (ie at http://www.techpowerup.com/superpi/)
- Note: For AMD Piledriver based CPUs, use this x87 'fixer' to improve your score. Parts affected: AMD Barracuda (Zambesi, Vishera), AMD Comal (Trinity, Richland), AMD Virgo (Trinity, Richland)

 

y-cruncher

- y-cruncher is constantly evolving, be sure to grab the latest version from HERE before benching

- We use the 1 billion decimal place multi-threaded benchmark test. Use options 0 > 1 > 6, when you run y-cruncher.

- Used result is the "Total Compute Time". Spot check must read "Good through 1,000,000,000". Both of these lines must be visible in y-cruncher along with CPU-Z in your submitted screenshot.

 

(Due to MaxxPi not being updated, and the fact that it's workload wasn't fully multi-threaded, I've decided MaxxPi is no longer worth benchmarking with. Please use y-cruncher instead for multi-threaded Pi benchmarking)

MaxxPi - Multi
- MaxxPI² - PreView - Multi can be downloaded from HERE (as a backup source try HERE)
- MaxxPi screenshot (can be made using 'Screenshot > Save to file' menu) showing score in K/sec, checksum and system info. CPU-Z inclusion in screenshot optional.
- MaxxPi screenshot with window showing calculation complete (with check sum and KB/s visible) and CPU-Z open with CPU specs visible. CPU-Z inclusion required as the CPU-Z part of MaxxPi code may only get updated each time MaxxPi is, which isn't often enough. Can cause incorrect info to appear in MaxxPi.
- Don't use batch mode. Using inbuilt screencap feature blocks check sum view and your fastest score might not have the correct check sum showing if manually screen capping.
- CRC must read as valid in the screenshot, proving the CPU didn't mess up.
- Please note that the 256M option is only available on 64-bit OS's.



_______________________________________________________________________________

SuperPi
Top 10 - 1 Million Decimal Places

Top 10 - 32 Million Decimal Places

 

y-cruncher

Top 10 - 1B Decimal Places

  1. 79.151 mark84
  2. 94.021s mark84
  3. 120.064 sora3

 

MaxxPi - Multi
Top 10 - 8 Million Decimal Places

Top 10 - 256 Million Decimal Places



_______________________________________________________________________________


SuperPi
___________
1 Million Decimal Places

AMD

AMD Zen based

  1. 00:10:035 mark84
  2. 00:10:900 sora3

AMD Excavator based

 

AMD Steamroller based

  1. 00:17:645 sora3

AMD Piledriver based

AMD Bulldozer based

AMD K10(.5) based

AMD K8 based

AMD K7 based


Intel

 

Intel Skylake based

 

Intel Broadwell based

  1. 00:17:969 Dasa

Intel Haswell based

Intel Ivy Bridge based

Intel Sandy Bridge based

Intel Nehalem based

Intel Core based

Intel Netburst based

Intel P6 based

Intel Atom based

VIA

Multiple Processors

Non-Windows OS


_______________________________________________________________________________


32 Million Decimal Places


AMD

AMD Bulldozer based

AMD K10(.5) based


AMD K8 based

AMD K7 based


Intel

Intel Skylake based

  1. 06:40:339 Dasa

Intel Broadwell based

  1. 14:46:997 Dasa

Intel Haswell based

Intel Ivy Bridge based

Intel Sandy Bridge based

Intel Nehalem based

Intel Core based

Intel Netburst based

Intel P6 based

Intel Atom

 

 

 

y-cruncher
___________
Multi-threaded 1 billion decimals

 

AMD

 

AMD Zen

  1. 94.021s mark84
  2. 120.064 sora3

 

Intel

 

Intel Broadwell

  1. 79.151 mark84

 

MaxxPi - Multi
___________
8 Million Decimal Places

AMD

AMD Bulldozer based

AMD K10(.5) based

AMD K8 based
AMD K7 based

Intel

 

Intel Haswell based

  1. 1680.1 K/s mark84

Intel Sandy Bridge based

Intel Nehalem based

Intel Core based

Intel Netburst based
Intel P6 based
Intel Atom


256 Million Decimal Places

AMD

AMD Bulldozer based

AMD K10(.5) based

AMD K8 based
AMD K7 based

Intel

 

Intel Haswell based

  1. 974.4 K/s mark84

Intel Sandy Bridge based

Intel Nehalem based

Intel Core based
Intel Netburst based
Intel P6 based
Intel Atom

Edited by mark84

Share this post


Link to post
Share on other sites

It's simply impossible to get that sort of score that Zebra got, and he is not using the one that we are all suppose to.

 

I'll see what I can pull outta my ass for this thread. ;-)

What gear was Zebra using?

 

 

I find that score mind boggling :S

 

 

Could it have been botched somehow?

Share this post


Link to post
Share on other sites

What gear was Zebra using?

 

 

I find that score mind boggling :S

 

 

Could it have been botched somehow?

 

I'm pretty sure Zebby was using a multithreaded version of SuperPi on Server Grade machinery.

 

So yeah, you're screwed

Share this post


Link to post
Share on other sites

zebra from the looks of it was using an itanium processor which none of us would ever have a chance of buying

 

also did a mod delete one of my posts in this thread? i swear i agreed with FB about the version thing.

Share this post


Link to post
Share on other sites

@ Medicine Boy

 

Yeah, zebra was actually using Itanium 2 hardware which is very very very fast when it comes to doing something like this. And yes, he's probably using a supercomputer when that happened :D

Share this post


Link to post
Share on other sites

zeb's result wasn't using SuperPi Mod 1.5 and therefore doesn't apply to the rules of this little results collective.

 

While it's neat to see what a multi-threaded Pi calculation can be done in, it's the single threaded solution that is the prize here.

Share this post


Link to post
Share on other sites

zeb's result wasn't using SuperPi Mod 1.5 and therefore doesn't apply to the rules of this little results collective.

 

While it's neat to see what a multi-threaded Pi calculation can be done in, it's the single threaded solution that is the prize here.

Fair point. I'll remove him from first place.

Share this post


Link to post
Share on other sites

Okay...

 

260*10 is totally unstable but 255*10 is probably the best.

 

1M, Phenom X4

http://img.photobucket.com/albums/v671/sor...s/SuperPI_2.jpg

 

EDIT: The problem with the current Phenoms as compared to the Athlons is the weak Lv2 cache. The previous great overclockers had 1MB of Lv2 cache which is probably why the Phenoms suck so much...

Share this post


Link to post
Share on other sites

Thanks for that one sora3, wondered how those went. Seems like clock for clock they're about the same as the A64's :\

 

c0nc0n - Just wondering what multi/fsb you're using for that one? I managed to get a similar clock speed on my quad but yours is a smidge faster. Also what's your ram timings?

Share this post


Link to post
Share on other sites

1:1 strap

 

470FSB x 9.5 at 1.43 volts after droop

Timings 5-5-5-12

 

My board sucks, anything above 450FSB refuses to run at 4-4-4-12, not even 4-5-5-12 :(

Share this post


Link to post
Share on other sites

Thanks for that one sora3, wondered how those went. Seems like clock for clock they're about the same as the A64's :\

Well, sort of. The Agena core is just a K8 but with added Lv3 cache and some optimisations. Though I do feel in doing that, the Lv2 cache is too slow given the latencies for it. The Lv3 cache does next to nothing here as the Lv1 and Lv2 caches are more important here, which is why the Core 2s annihilate it. If I get the updated 9550, then it might be a bit faster given the fixes and optimisations in it. But I'll have to wait for the Deneb then I'll see if that makes any difference.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now

×