The Official PC building thread - 4th Edition

Very unlikely unfortunately. I'm still hopeful, because even though I've stuck with Intel for offering faster (disregarding absolute value) CPUs for over 6 years, competition is always a good thing - look at how stagnant things have become on the Intel side of late, without AMD pushing from behind. Right now the second successor to my CPU costs 20% more than mine did, and is around 40% faster out of the box, maybe 35% faster than my overclocked CPU when also overclocked. With the cost of a new board too, it ain't worth it!

 

Myself it's more for selfish reasons LOL! I would really like to see CPU's making huge gains. I'd like to see BD-25 encodes averaging under 1hr for HIGH quality encodes

 

Depends on the film length. I haven't encoded a film lately, but low-bitrate 1080p (e.g. youtube) will encode on my i5 at around 1.7-2.0x realtime, so 40-50 minutes for an 85 minute film, and around 70-80 minutes for a 140 minute film.
Up the bitrate and that performance will come down a bit, so for a long 1080p feature at high bitrate, I'd guess at around 100-120 minutes on my current CPU.
What I'd really like to see is a program that can actually use OpenCL to encode video, but so far none seem to actually use the GPU to encode at all, even when accelerated...

 

Using BD-Rebuilder, I can set for the "Faster-Better" quality and average 3hrs or less on most discs, but a lot of discs have a crap ton of extra features, which I also want to keep I rarely every do a movie only encode. Soon I'll be using MultiAvchd to consolidate whole cartoon seasons (480P) to single BD-25. If I'm right, quality will be unaffected.

The faster setting supposedly cannot be compared to high quality setting. But since I plan to use a projector for my movies one day, I prefer my backups were as good as they can be.

 

Current record for Super Pi is 18.002 seconds Have had it in the high 17s at 4GHz but it was never quite stable. Would IBT under normal stress but wouldn't pass under high stress. Sadly Super Pi has been proven to be quite Intel biased, regardless of actual differences in processing power.

Omega, I collect cartoons as well

 

Super Pi just another fraudulent program and single thread tool only. At least use Hyper PI or Prime95 but again not the best.

 

Estuansis,

That can't be right! This is for 1M, right? I was in the mid 13's (13.880) with my Intel E6750, and I think I saw Sam in the mid 10's with one he posted, a week or so ago. Some Gaming Rig, I think he said.

Russ

 

It was yesterday, and yes, my i5 750 @ 4.116Ghz

Jeff: Are you running the same version of SuperPi we are?

 

Ummmm on a Phenom II X4 at 3.8GHz 18.002 or essentially 18 seconds flat is pretty much par for the course... Maybe I should have said MY personal record to clarify?

Russ scores 17.986 with a Thuban Hex-Core. So mine falls right in place no? He's using slightly better memory so it would make some sense.

 

Yeah he did, you're right, now I look at it. I think the result for the E6750 must be wrong (not incorrectly recalled, but some other issue with the test) - the E6750s were good, but they weren't better than 4Ghz Phenom II good...

 

Remember there is a huge Intel bias in the program. So only really useful for comparisons within the same brand and for testing OCing differences.

I mean a HUGE bias.

 

That may well be true, but 13.880/10.148 would make my i5 only 37% faster than Russ' E6750.
By knowledge of how Core and Lynnfield compare, the E6750 would have to have been clocked at around 4.1Ghz to achieve that. Doable, but I didn't recall it getting that high from memory - more like 3.7-3.8 was it?
I suppose a 10% variation isn't much to be concerned with in an antiquated single-thread benchmark.

 

Jeff,

My best Super Pi 1.5 score so far, was 17.072s, which is almost 1 second faster.

Russ

 

Sam,

Actually I miss-copied! I wrote 13. instead if 14. I have to shamefully admit that 13 had been crossed out on my hand written log, not once but twice! Dyslexic fingers I guess!

Russ

 

Yep the newer Thuban core and better memory timings would be the cause there.

 

Steve,

I use IBT, 5 pass, maximum stress!

Russ

 

Same here. I use 5 passes max stress to test new settings and a 20 pass burn-in to make sure I'm stable. Have gotten 4GHz to do 20 passes at normal stress but will not do it at max stress.

 

X264 is all the stress test I need Unless I'm going for high clocks. Then I'll begin with Prime

 

Any good encoding software can be a valid stress test. Any distributed computing program works great as well. Folding@Home, Seti@Home, BitCoin, etc. Using GPU Compute(OpenCL, Cuda) makes them a good graphics and overall system test as well.

 

IBT is much better than SuperPi or P95 but running AIDA64 CPU Stress Test (100%) at the same time run Unigen Heaven will test the system much better and will cover multiple graphic cards better, also it will stress your buses better covering AVX/ADS and so on.... PCMark7 is another decent tool for measuring as these tools go.

However like I have said before I can test quickly when encoding like Kev does but also just stoking up Angry Birds, go figure!