Printable View
You need to be more balanced in your postings.Quote:
Originally Posted by roofsniper
SuperPi itself is not biased, but obviously a biased person could spin the results that both Intel and AMD get on that benchmark.
I am not advocating that anyone determine a buying decision on SuperPi alone and would encourage a range of benchmarks be used.
Hopefully Deneb will be a winner for AMD and can topple Penryn in gaming tasks and challenge Nehalem too for gaming.
I see questionmarks above my head when I look in the mirror and think of this thread...
SuperPI is USELESS and you look pretty dull if the first bench you ask for is SuperPI. SuperPI is useless, results in SuperPI do not translate to equal results in real programs. Since Intels Core 2 line has a huge improvement over AMD on single threaded apps, but multi-threaded wise the roles have turned but the lack of multi-threaded benched vs the overhyped SuperPI benchmarks this does not make it biased how?
It might or might not be biased from the people behind SuperPI its self, but it's biased by the people who actively use it to show how much their PC does vs the rest. Compared on the same platform, no problem. Compared on different platforms, useless.
Ok fair enough.Can you please tell me what use do you find in SUperPi,what does it do for you and what kind of conclusion about one uarchitecture can you draw by looking at its SuperPi results?Quote:
Originally Posted by Chad Boga
And please give me a straight anwser not some dodging bs this time.
Thanks
SuperPi is biasedQuote:
Originally Posted by Chad Boga
SuperPi is written with INTEL Complier, it pretty much has direct contact with Intel chips
AMD chips has to do a translation
That's why there is major difference in SuperPi times
No spinning , just a fact
Well not really.What really happens is that its dataset is small thus it benefits the chips with big caches(fits right in them) and second is it uses specific legacy x87 instructions which have very low latency on intel hw( a perfect combination of those that suits intel uarch.).It's practically a loop bench that runs awesome on CPU designs with big caches and low latency on specific x87 instructions(that no one uses since all software is migrating to SSE anyway,another point in favor of useless argument)
That being said,it's the most useless "CPU benchmark" that has near zero practical or theoretical use.One practical use is to run 1M test and see if your OC is even remotely stable,or run 32M and see if your ram OC speed is remotely stable.Theoretical use is the fact that it can be disassembled and studied to see what kind of designs it prefer(look 1st paragraph).
super pi is useless for judging amd cpus, it greatly favors intel's large extremely fast l2 cacheQuote:
Originally Posted by Glow9
Guys I'm afraid that I have to agree with Lestat.
First of all, there's no absolute "unbiased" program. Period.
Second, if the developers decided to use the A tool instead of the B tool because they could do their job easier or faster or whatever with the A tool, that doesn't make the program or the programmers biased to a company.
Third, there is NO single benchmark that can represent the performance of a processor in all tasks and scenarios.
Fourth... AMD did more than well when they introduced the Athlon 64's in SuperPi, and were the king of SuperPi until people started benching Dothans ( the C2D design "starts at" Dothan's design ) [ in a way ].
Last for now... if you want to compare two CPUs, the only way to have a good image of their performance is testing them in a dozen benchmarks, two or three per scenario, 3 different number crunching benchmarks, 3 different CAD programs, 3 different database servers & patterns, etc.
Check this post out now with more results
http://forum.coolaler.com/showpost.p...4&postcount=73
exactly, you have to get a wide spectrum of results. But another reason why I think super pi is useless for amd is because phenom is only a second or two faster at the same clock where as it is much faster than k8 in multicore extensive tests and a a lot of fp tests (not a good representation of core improvement)
and how is this flame baiting? or giberish? you have show in your own text that you intend to ridicule any person that disputes this post, by saying that the pics are fake.Quote:
Originally Posted by MrMojoZ
people are entitled to express an opinion here without feeling that excisting forum members will belittle them, disagree you can.
bellitle and ridicule you cannot, there are rules here for that, that i have allready used against you via a mod, you can try and use those rules against me all you want, but i stand true here. there is nothing to punish.
you are the flame baiter, twice now in as many days, i will not be happy until they ban you here, until then i will point every bulliy post you mmake for everyone to see.
And Im getting sick of people constantly bickering off topic in almost every thread this :banana::banana::banana::banana: has nothing to do with the OP take it to PM's or get a room.:mad:
I really want to know how durable AMD's silicon is to even attempt, boot, and stay stable enough on air* to run long enough at 1.6vcore to get that screen shot LOL!
I'd like to know how high I can get it at 1.4vcore. If I can do 3.4Ghz I'll be happy as can be. If Deneb gives us IPC enhancement it will be AS FAST AS NEHALEM in real world gaming (not in Intel optimized synthetic tests, though, of course). Why do I say this? Well Nehalem much slower than Core 2's in non-multithread optimized games, just like K10 ironically. Seems like "server optimized" architectures aren't the best for gaming but they are the wave of the future. When software is optimized for Nehalem, K10/.5s will pay you dividends.
Proof: http://74.125.93.104/translate_c?hl=...5JudaW9pwJxSeg
I wanna know if they really used 1.6v for that or its just a cpuz bug!Quote:
Originally Posted by Mechromancer
No, i7 is nearly as fast/slightly faster as Yorkfield, until you use SMT.
Since there's no easy way to toggle SMT in Windows (only in BIOS), welcome to the new world of tradeoffs.
Now let me guess... most review sites use SMT exclusively for productivity benchmarks, turn SMT off for games, and hail i7 as teh winRAR. :D
Oh, and think of it. 1.6V on AIR?
@ Titan, those benchmarks are from a few months ago. thats when we first were shown the huge power drop the 45nm were gonna see.
as far as a benchmark being biased, hows the new 3dmark vantage cpu benchmark handle? i hear its biased based on the video card offering up to 10% difference on cpu score by having different video cards used.
Oh, BTW the Vantage CPU benchmark uses... PhysX engine. Which uses...
x87 legacy code again. (Hardly little SSE optimization for CPUs; they want to sell their PPU/GPGPU)
That explains why SMT on i7 has such a big lead- x87 is primitive code that isn't used much by any other routine.
From what i heard 45nm Phenoms/Shanghais will be very cool chips,even with high volts>The current leakage has been well addressed at 45nm node and power consumption will be a lot lower than at 65nm with comparable clocks and/or volts.AMD also feel confident about their shipping parts so it's wait and see game right now.Not much more to wait,maybe until middle of November or the end of it for Deneb.
By the way, in case some of you are unaware of the existence of three SuperPi versions, SSE, SSE2, and SSE3.
I think with the early steppings going that high most of us don't doubt that, the hot topic now is the HOW.Quote:
Originally Posted by informal
Even if AMD's 65nm was dastardly crappy, 45nm wouldn't automatically give results that good either. They've used Immersion Litho since 65. Same for Strained silicon, perhaps even earlier.
But it's definitely not HK/MG. That's been delayed to 32 and no buts. :p:
Reported for flame baiting.Quote:
Originally Posted by soundood
I don't know anything about what hardware matters in 'Fritz chess' but it doesn't look good. Is it a fair comparison?
AMD:
http://img.inpai.com.cn/2008/8/5/0f5...c2fae025f0.gif
Intel:
- Middle of site
From what i read the CPU is running on a 780G /SB700
That initself is quite a feat 4ghz and tbh i didnt think a 780G would be capable of pushing 1.6v + without good PVRM. Not to mention the 780G is DDR2.
Also air cooled.
Even with all the leaks deneb remains an enigma.
Sooo AMD is lacking something in design. I'm guessing a better l2 would benefit AMD I don't see how this makes super pi biased, it's because AMD is lacking in cache.Quote:
Originally Posted by AliG
really old post im not sure if it is true but i saw that months back.Quote:
Originally Posted by Eson
and seriously guys i would stop posting in this section for awhile because anyone that has any common sense would know that super pi obviously gets better results for intel. you could underclock some intel cpu down really low and run a ton of benchmarks and the amd cpu would win every time but on super pi intel could still possibly win. i bet even intel fan boys have to agree with this.
Remember it's a C0,the first 45nm stepping that was produced(could be better score with C2 since we don't know how bug free was C0 to begin with).
As for Fritz,it is one application that ,like PovRay or Cinebench10,prefers intel hardware.I don't know how much the Fritz engine is used in chess games-or for anything else,but i do know that the former 2(PovR and C10) are not,how to say it polite, preferred choice amongst the professionals.