Ok so many that have talked with me one-on-one (and many that haven't) know that i'm a big anti-overclocking anti-benchmarking kind of pc person.

First off i'll cover benchmarking programs. You have a program that tests the maximum possible speed of your processor, or it's processing power, whether using bignum arithmetic or straight 64-bit ALU processing (not much RAM involved)

The question is...outside of these programs, how often is that processing power utilized? If a program isn't programmed with the best possible practices and optimization to be sure that it runs the fastest, does your 13000 PC marks really make a difference? or a 23.4 second 1MB SuperPi calculation mean that you open firefox (or of course Opera) faster?

personally i believe that for one, some benchmarking programs are tailored to the hardware developers that want big numbers so they will make sales, this is partly proved with the big fiasco where nvidia had replaced .dll files in 3dmark (i believe that's the program, ask me and i'll dig up the article) in order to out-perform ATi by a broader margin.

so my question to you is this: Do you use benchmarking programs, why? and do you really care about the results?

we will leave O/Cing for another topic :a022:

somethin like SiSandra soft? i gues i didnt know how to use it mayB tat is a reason why many people dont use them..
.. some dont care, dey say watever..

..u cant really expect people to play by the game if dey didnt know the rules [fully] (many of my friends dont even know, sad fact.. some of dem are engineers)

So, wat s/w do u summon? how to use 'em??

I never bother with benchmarking programs since I left the standard hardware setup of the Amiga.

That machine had a set of chips that you couldn't change unlike a PC where almost everything can be upgraded.

I believe also that PC benchmark apps would be tailor made for the chipset by the company that made tham to give the best results which is why I don't bother as it only gives you an idea of what things could run like.

If anyone wants to see their machines pushed to the limit visit pouet.net and download any production from the top groups.

Here are a couple I think are great...
http://www.pouet.net/prod.php?which=30244
http://www.pouet.net/prod.php?which=9450
http://www.pouet.net/prod.php?which=52938 (this beauty is only 4K!!!)

Download/YouTube links are in the bottom-right of the top box which gives information about the demo.

Processor benchmarking programs can be useful for people who do a lot of audio and (particularly) video encoding, both of which rely mainly on the processor's abilities. And in these cases, special benchmarking software isn't really needed. The actual encoder can be used to compare processors. For more average activities, like web browsing, processor speed matters much less of course.

that's how i feel, when i'm looking for good hardware reviews, i'm always looking for tests such as the fastest opening time for programs i might use, such as web browsers, or say adobe software (that could be a benchmark in itself )

i've also seen some benchmark tests where they encode a 100MB mp3 file and then a wmv or avi showing the lowest scores, i just don't see the need for all these new benchmarking apps, the only people that really use the processors to that extent are supercomputers, and they have better things to do than see how many points they can get

The only time I use a benchmark app is if I make a hardware change I might do a benchmark before and another one afterward. In this case I know I'm comparing eggs with eggs.

I agree benchmarking can be very deceptive, it often doesn't reflect the difference between 2 similar ability hard drives where one of then has a superior onboard buffer for example or different architectures and cache levels on CPU's. I agree has also been perverted by hardware manufacturers who will use it untruthfully or just not tell the whole story by selective benchmarking.

I think Maz and Newartriot have got it right by using an actual application load time or process time as a benchmark although results may vary considerable when a program is loading from a prefeched state or where half its resources are in virtual memory/swap/pagefiles etc.

I'm not a high-end hardware purchaser so I'l never loose sleep wondering if my system is as fast as another. In answer to the original question, I would never take a benchmark result over and above a positive report by a genuine user on a discussion forum. Also you won't find benchmarking software loaded on any of me PC's.

There are synthetic benchmarks -- programs that are designed to give you a magic number performance rating. And there are daily-use benchmarks that just run a suite of regular apps and record the time taken to do a set of tasks. The synthetic ones are easier to manipulate, so those are the ones to be wary of, but in general, most regular stuff doesn't really come close to using all the modern processors' power so you won't really feel the difference between a current cpu and a 4 year old one on regular stuff.

i5 / P55 Lab Update - Now with more numbers
^ I'm just linking that one because I read it yesterday and it sparked a couple of thoughts about benchmarking:

(1) The new Core i5/i7 Lynnfield processors do an on-the-fly overclock, called Turbo Mode, which improves the 'snappiness' of the system as the overclock happens instantly when you open an application. This is provoking debate about the benchmarks because AMD still has the traditional, manual-overclocking-needed processors (maybe AMD can adapt to the new situation via their AMD Overdrive tool, though). That snappiness is mentioned by the reviewer - it's not apparent from any of the benchmark numbers themselves.

(2) If you look at the Games benchmarks, there's something odd that we haven't seen for a few years: the AMD processor with an nVidia graphics card, is kicking those Intel processors' rears quite significantly for some reason. As this is totally unexpected (nearly all games are showing the same thing), the benchmarking shows that there is either a problem or an upset in the Force. AMD used to always be ahead for games back in the day, so it's not really that strange, but it's odd considering today's cpu situation anyway.

I don't think an individual benchmark can say something valid, but benchmarking a set of regular daily-use apps should start to give an idea of how processors stack up against eachother. That said, the set of apps used and the small amount of testconfigurations, can still leave a bias in the results -- and that snappiness that you seek, still isn't properly quantified in any benchmark, so point conceded

that actually is odd considering AMD always worked so well with ATi in the past, but before i get on the topic of nvidia (which i'll start soon)

Personally i would like to see more open-source benchmarking apps (read memtest, though it's not exactly benchmarking, it can be used as one) as then we can be sure there is unbiased coding, such as no SSE4/SSE5 or whatever instruction set that intel has that amd doesn't or vice versa, it can be very difficult to code for CPU's nowadays and put the test on "equal grounds" but throwing every benchmarking app as most news and review sites seems to do is definitely not the answer.

i was hoping for more of an argument, but it seems that most are in agreement, so i'll leave it at that unless someone has anything else to add :)

I too am anti-synthetic benchmarks. They don't really serve a purpose other than measuring e-peen at LAN parties. Because there are so many variables in today's systems, you cannot really brag about your 12 billion 3dmarks when a system with 4 billion 3dmarks gets higher fps in game x or performs application y better because game z has dismal performance and application α runs like crap. While most games do not utilize all 4 threads, there are more and more 64 bit applications being produced which can use as many threads as it can which is why we're seeing a trend to more cores (hexa-core processors coming in about 6 months) Another reason you keep seeing synthetic applications in reviews is that they prove to be bias in that they're written for 32 bit only and the producers of those utilities refuse to make a 64 bit version.

I do however see a point in real world benchmarks as long as they're performed well; trans coding video and audio is a very good benchmark and will usually max out your hardware. Last time I used the Vegas studio, it maxed out my machine for several hours and maxed out my roomates quadcore in less than half the time. Many of the tech sites I frequent transcode some video, convert a CD format, pack up a file in winrar (heavily intel based btw), and do something with database like a microsoft excel operation along with the synthetics and games. All these, when combined, paint a more complete picture of a single hardware difference.

Of course new hardware gives more responsiveness (snappier) and faster everything: if it didn't, no one would buy it nor would companies produce anything new. Due to current system bottlenecks, there is only so much to be gained for an individual hardware upgrade which I believe is why you think testing applications', games, and other frivolous tests tries to identify. A few more megahertz, or a few more transistors, or a few more shaders can only do so much on their own.

Thank you for this forum by the way, I can already tell I'm going to love it.

...yeah, tats 'nother way of sayin watevr

if thats de case.. isnt thr one benchmarkin software tat can be true to its wits?

while recommending some hardware for some guys, i usually read through some reviews in some magazines(digit : http://www.thinkdigit.com/) i gotto say.. at times it did work, dey speak of the pro's & con's of certain hardware (i dont have de least idea if dey were affiliated to some company.. putting out some partiality factor)

can u name some forums where dey publish some ratings like these on the net? (i donno if dey have it on the net, but i can upload some old pdf's of the magazine if required), dey speak of some benchmarkin software like SiSandra soft etc.. so, i really wanted to know if something good is out thr.. (some B'markin s/w)

i also wanted to know if i can rely on somethin like tat.. or not
based upon ur reviews, notin matters.. its all but fake, of course not many(or noboDy) uses a supercomputer.. who wud rely on the verdict by these kinda s/w's But isnt thr something tat can be true.. atleast 80%?? ..i know its all but crap.. a piece of sh!t.. but some sh!t outta be good.. rit?!..

you mean to say that the "synthetic benchmarking" apps don't have 64 bit versions? or did i read it wrong?

...as far as AMD's are concerned.. i know i thing for sure. dey bring on the heAt. a common misconception among many of my friends, dey dont wanna use AMD, cos' dey say it gets very hot 'n the processor wud worn out soon.. i donno anybody speaking to me of a burn-out, fried processor. hey im one of dem, who wud use a intel compared to AMD considering the fact that many of my friends dont even care to close their psu cabinets, dey just leave it open.. dey say macha(dude) it wud cool down the pc, or the pc needs more air.. sorta thing..

and i f dey say, games run wen the processor can churn out bits of data in a faster manner, mayB tats why AMD's are so good in the gaming industry <- jus a theory(or a hypothesis)

so, are u sayin, if u wanna BE GAME go for AMD's ??
.....jus now i saw it.. AMD - MAD | wow!!

QUOTE

you mean to say that the "synthetic benchmarking" apps don't have 64 bit versions? or did i read it wrong?

wat?!

..i just know benchmarkin, synthetic??
i didnt read muff's comment :(

..i wanted to know if there are some forums where dey do these kinda benchmarking stuff and publish it at the net, so tat we could get a better idea on wat hardware to buy next..

i think you missed the idea of the topic rottzy, this is more about benchmarking software and the futility of most of them, not amd vs intel or what the top hardware is, most of that is covered in regular tech help regularly

Yeah, tomshardware commented on how users were commenting on the lack of 64bit OS in their testing of new hardware (as it usually yields better performance). The reviewers are sometimes forced to use windowsx86 just to run a specific benchmark rather than run what most people are using or wanting.

I don't remember which synthetics they were but there were a few that they kept using and commenters kept complaining about the lack of x64 OSes.

well think about it like this

if you run 64-bit vista, and the benchmarking is on 32-bit spec, then you are not only throttling it, but you are emulating the 32-bit processing, which can yield even worse results, which is why i also believe testing should be done on normal applications, since if there is no 32-bit foobar2000 or lame.exe then who cares if a 64-bit processor gets more points than my athlon, if they both run the 32-bit apps at the same speed

and since most common apps are running in 32-bit anyways (your average browser is not 64-bit regardless of ability to utilize it, as far as i know) i think that a 32-bit testbed would be better anyway, not to mention if you are on vista, xp-32 is going to be a better choice as it has more room for resource usage than vista does, which allows for better scores


hopefully my point came across clear

I understand what you're saying, but in just about every aspect, (at least for win7) 64 bit is the way to go for real world stuff. Understandably some of the more common applications like internet browsing, and even word processing do not really need 64 bit as the performance difference would be negligible, but stuff which almost requires a lot of RAM (apps come to mind like Vegas, CS4, CAD/CAM, games) obviously benefits from 64 bit and hardware related capabilities.

A simple app such as ConvertXtoDVD is a nice little benchmark tool. Since I'm an avid OC'er, I can see direct results using the plethora of benchmarking software available, in turn I'm also able to see these results in the time it takes to decode a Video using ConvertXtoDVD, since decoding is a processor intensive task.

I only used ConvertX as an example since it's quite popular, but the same results can be had with many different apps, such as video editing, muxing/demuxing, and rendering types of software.

that's actually a good point JulesandAl, most everyone i know here uses that app, and comparing speeds on it is a great way to benchmark your current setup, if an OC doesn't help the speeds on that program, but it does on a benchmarking tool, is there really a point to it? i'd like to think there isn't :)

Just because an Overclock doesn't help a single program doesn't mean the net gain is zero. It would be like modifying the engine on your car (with intake, exhaust, and computer chip) without getting a decrease in your 0-60 time, without improving handling, but you get an increase in towing capability. Perhaps this is a stretch of an analogy, but an overclock does not help single programs as each program utilizes each part of a computer differently such as ram speed, ram timings, front side bus, number of cores, number of threads, CPU architecture, GPU architecture, effective CPU speed, and even GPU speed.

granted, but if it doesn't help the programs you use, then why do it? increasing the performance in anything without increasing the quality of the internal parts is just asking for breakdowns, as in your analogy, putting high compression heads on a car without forged pistons is asking for a blow out :) same thing with computers

True, however overclocking doesn't necessarily mean your things will break down quickly. I had a friend have more motherboards, power supplies, and RAM simply stop working on factory clocks and voltages than I have. (My PC has had an overclock for the last 4 years) Perhaps I'm one of the lucky ones that didn't have my stuff meltdown, but I have decided that an overclock has benefited me in games, Sony Vegas, SQL database, filesharing, as well as decreases in windows boot times.

Yes, if all I did was use the internet and play media then an over clock would be pointless but since I do use my PC for a broad range of applications which do utilize an overclock it is actually beneficial to me. The crux of the matter is that programs do not have a decal on their box pointing out that "if you overclock you will see benefits when using this program" so it is nice to see if your personal over clock is helping a program or set of programs. If I save a few seconds loading into left4dead, trim 10 minutes off of editing a video, and can launch Microsoft word faster then I feel there is a net gain despite some people feel a few seconds or minutes to be trivial. Free performance is free performance making my "investment" worth just that much more and last just a little bit longer. It is true that parts do have their limits in just how much more performance you can get out of them; but isn't any performance for free better than none?

probably true, but with those overclocks, i must ask, do you have any extra coolers than the stock fans you got with your setup? most don't, which adds usually a minimum of 20-30$ for a good aftermarket cooler for the CPU to keep the temps down, when you could spend an extra 30$ from the start and buy a faster processor

of course, it might be just me, but my Athlon XP 2700+ stock has been running since the 2700+'s release without a hitch, (and i don't mean it sits around turned off but still works, i mean it's still running 24/7 as a server now) heck the mobo manufacturer does not even make them anymore (chaintech) but then again, we can't predict whether it would have failed on OC or not

but back then, it was far more difficult to overclock without so much risk as you can today

Nope, all stock cooling with stock case fans even. (Thermaltake Armor) I did try a higher CFM fan on the stock cooler which did drop the core temps by about 3-4 degrees but it was just a little too loud and now I'm back on the stock fan.

You have to look at the high end parts as well; compare the core i7 920 and 975EE. On some air cooled 920's they can be seen working daily at around 4ghz. The 975EE does a bit higher around 4.2 and is arguably easier to overclock due to the unlocked multiplier however the $1000 price difference isn't justifiable for a few hundred (if that) more megahertz. The 920 stock clocks at 2.66ghz while the 975EE comes out of the box at 3.33ghz. For 670mhz, is that $1000+ premium worth it at stock speeds? Without considering ultra high end stuff like that, we can look at AMD's two flagship consumer processors; the Phenom II 955BE and 965BE at only ~$45 difference, there is only a 200mhz clock difference. Without spending much or anything on aftermarket cooling, you can easily save the $45 and overclock the 955 to operate safely at a 965 speed. Heck, for $45 in aftermarket cooling added to the 955 you can surpass anything the 965 can do out of the box. Once again this is a statement of "speed costs money" but if you can safely get "free speed" why not?

However you do point out the big elephant in teh corner; your 130nm processor probably couldn't work daily with a 15% overclock and a stock cooler due to manufacturing advances since 2002. My processor is on the rather old 90nm process and came with a stellar stock cooler as well and overclocks terribly compared to the 65 and 45nm stuff out today. It's all about doing things within a certain set of parameters or within reason; I am comfortable with overclocking my parts and am confident that this overclocked 4400+ will go on for another 4 years as long as the other hardware holds up.

some valid points muffin_man :) point conceded there

Elaborating on muffin_man's point, I'd argue that overclocking is just reverse cpu-binning as the cpu model is partly, if not largely, determined by marketing decisions rather than the small variations in performance that occur between cpus. The easy unlocking of extra cores on Phenom2s illustrate this best, as some perfectly good cpus get their parts disabled to fill the high consumer demand at a particular pricepoint.

For overclocks, benchmarking can show the stability of an overclock a lot faster, though those apps tend to be synthetic because of that non-regular-use nature. Conclusion: synthetic for testing, regular-use benching for comparing?

I wouldn't mind some "free" extra speed either, and in some cases sure, the stock fans may be good enough, and you may get some "free" speed, however, in the long run, generally someone with a fetish for over-clocking (or was the word hobby...) is definitely going to spend more money on their rig than the average bloke because, that is just the way of things.

There are the exceptions, but "free" is definitely not the rule.

Muffin man has the right idea here. If you are a gaming enthusiast, the communities will always relay information as to which chips will give the most bang for the buck.

Example; an E7500 2.93Ghz @ $120.00 can easily clock to 3.8Ghz, while the E8600 3.3Ghz @ $270.00 doesn't really like to be over 4.2Ghz, with the exception of the E8600's added 6MB L2 cache, the E7500 is definitely a better deal.

I find the OC problems of today are more centered around NB and RAM, rather than the CPU itself. I found it quite easy to keep my CPU cool with a decent HSF, but a quality Motherboard is required to get a healthy and stable OC past the NB.

I do use benchmarking products because, honestly 3d mark is a decent way of rating non-OCed systems. The results really aren't important except in tweaking, or deciding what to buy. Yeah each manufacturer tries to stack numbers in their favor but I had a decent idea of the FPS I was going to get in certain games by looking at the 3d mark scores of people with similar systems to the one that I was building and comparing them to mine. Honestly though the best way to benchmark a machine is to run a game under similar circumstances and rate it by FPS and benchmarking programs like PC and 3d mark are very controlled ways of doing that.

3d mark produces an arbitrary number that doesn't reflect much. Far cry 2 favors ATi GPUs while Crysis favors nVidia. Systems can have a very similar 3d mark score but vary greatly in these 2 games just by swapping the video card out.

exactly my point, benchmarking applications are 'controlled' they are manipulated to produce results to the highest bidder, or the one that modifies the program upon execution (read: nvidia)

if you want to know the 100% pure FLOP count of your card, then that's a good way to get a real-world number instead of the estimation by the company, but if you really want to know which video vard/cpu/mobo/ram combination is going to give you the best performance in-game, then benchmarking programs are really just a measuring tape

I use benchmarks - they tell me if something produces a change and how much the change is,
it helps me to tune servers, and design systems - every once in a while I'll overclock something,
either because I'm curious, but more than likely its because the client thinks he can get a savings.
Benchmarks are tools, its good to know how to use them, its better to know when they are useful,
I remember someone who spent a lot of time bragging at how powerful his car was, he'd talk and talk...
rather good at talking, finally someone quietly said - let's race - he may have had a powerful car,
but it turned out he really had no idea how to drive it.

.

...sometimes benchmarks help us better drive our 'puters.

regards

Im thinking that it depends alot on which company makes the bench mark software and if they favor amd over intell or vise versa. If you are not a tweek freek then these tools dont realy have much relevence in the real world but that's the point though overclocking isnt real world for most we like to see an improvement in the apps we use and if i upgrade to a faster cpu then i expect to see an increase in speed to lets say winrar if it unrar's at 5-10 seconds faster or im converting and it takes 5 mins less then im happy I could over clock and see an improvement but to keep the system stable i would need a mobo with a high fsb speed and probably an added fan and maybe a new cooler, or i could upgrade to a slightly more expensive cpu which is better?

Plus the main issue is, is my rig stable what's that burning smell i hope someone downstairs is making toast, and i don't want to be watching the dual core center more than i look at the the time.

A page full of numbers i don't understand is not going to put my nerves at rest. In my opinion for me at least is that if cpu's were built for these overclocked speeds wouldn't the makers of them just set them to the advertised overclocking speed and sell them as a better chip

Will Intel be forced to remove the "cripple AMD" function from their compiler?. An interesting blog article by Agner Fog

i read about this just this week, nothing new in my experience, when nvidia (as discussed in this topic) did the same thing with benchmarking programs. Just think of all the benchmarking software if compiled with that software, it automatically lists intel as the best because it's compiled with a bias

i can't remember the software, but there's one in particular that regularly lists intel as the best, even when all others list amd better, i'll try to find it and see who makes it and what they use to compile if possible

Sorry to hijack guys but shouldnt benchmarking really only be used to be double sure you havent just wasted another £300 or more on your computer?
I use pc mark when i get a new piece of hardware for various reasons.
1, To make sure i havent just be conned out of my hard earned cash.
2, To see how far i can push the new hardware without frying it or becomeing unstable, bottlenecking etc.
3, Because curiousity killed the cat so to speak.

People give out all these benchmark scores but they forgot to mention that every batch of hardware will react a little diffrently. Like my old q6600 go stepping i could not get it over 3.5ghz but alot of others did etc etc. As well as the fact it also makes a huge diffrence on how old the other hardware is (is it worn down a bit).
So surely at the end of the day its back to the thing of have i just been conned or do i have what i was expecting from that piece of hardware.

Thats just my thaughts on it. Apart from that i can see no use whatsoever of benchmarking anything.

very true

thats an awesome benchmark! It doesnt even run! :D My computer sucks

However, the Intel CPU dispatcher does not only check which instruction set is supported by the CPU, it also checks the vendor ID string. If the vendor string says "GenuineIntel" then it uses the optimal code path. If the CPU is not from Intel then, in most cases, it will run the slowest possible version of the code, even if the CPU is fully compatible with a better version.

If the above is not against the law, than is most certainly should be.

it is against the law in some respects, such as if that software is used for a hardware manufacturer's ad campaign slamming the competition, but they have to be caught, i know nvidia got slammed with a lawsuit for replacing certain .dll files in a 3d benchmarking software which boosted their results significantly