I.nfraR.ed - DDR SDRAM @ 273MHz - 630.5 marks MaxxMem

[Mr.Scott]

Current version is actually 2.01.

[Strunkenbold]

IMO db should be cleared with those old versions. If I remember correctly, there was some discussion about incorrect/cheated results.

[I.nfraR.ed]

Most of the guys with old scores have absolutely no fault.

I've updated my score and with old version I get 790 vs 710 with same settings and timings.

[Strunkenbold]

Thats true, but this is same situation like we had with wprime. Old versions had a major performance advantage. You cant beat those scores even with better clocks. Dont know what would be a fair solution but as maxxmem doesnt receive any points anyway I would say let them disappear...

[Luebke]

maybe put the old results in an own section "maxxmem - old" and close it?

[TASOS]

A moderator should do a split thread , so we can discuss freely about maxmem.

[TaPaKaH]

I don't think that the score split/blocking is worth the effort unless we can somehow guarantee that current/future versions will be consistent.

[I.nfraR.ed]

I don't think it's necessary. Leave it this way. I'm fine with it as long as we all use the latest version for the competition. I don't even run this bench and I wouldn't if it wasn't the competition .

[Stelaras]

Can you please run with v2.01 (like kotori's run) at these clocks again ? I'm curious to check something .

 

Thanks man and congratulations for your score !

[I.nfraR.ed]

I have ran it and the score is about the same. I think his score is bugged

What are you getting with similar clocks?

 

Bypass max is wrong here, but everything else is tight enough

 

 

Edit: btw, expert and venus should be more efficient than ultra-d/sli-d, but I don't think the difference in read and write could be that huge.

Edited April 11, 2014 by I.nfraR.ed

[Stelaras]

Honestly mate , i havent made any runs at 260Mhz+ , except some very quick testing at 240-250 just to check the OS and the differences between the versions of this benchmark.

 

I have both expert and ultra-d , so i'll do my best to test them .

 

As for my opinion , kotori's run doesn't seem bugged to me ... but the pattern seems to match with my test runs(done with the expert board at lower clocks as i wrote before)

[I.nfraR.ed]

Hmm, ok. So I will try with Expert, although my board can't clock BH5 very high for some reason.

I get sometimes similar read and write, but I thought it is bugged