Intel Sees the Light

FrgMstr

Just Plain Mean
Staff member
2FA
Joined
May 18, 1997
Messages
58,694
Intel Sees the Light.

We figured it out six years ago. Benchmarks suck when it comes to actually showing you how your gaming box is going to work out for you. It took us a few more months to figure out what we were going to do about it.


We are now going to have to tell you what kind of experience the hardware is going to deliver instead of simply showing you frame rates and whether or not there were jaggies and blurry textures. While all of this might sound tremendously difficult, it is really quite easy to accomplish. The reviewers out there are going to have to play the popular games and give their opinions of how the hardware in question handled the task.


Jump back to today, and even the guys are Intel are starting to see it our way.


Kernel benchmarking has its place, but benchmarking total (or "real") application performance is critical for accurately evaluating HPC systems.


And even guys at Purdue University!


Measuring High-Performance Computing with Real Applications

Anyway, it is good to see a company like Intel finally seeing the light even if this is specific to High Performance Computing. AMD and NVIDIA are still stuck in the days of the 3dfx framerate instead of the experience it can deliver to the customer.
 
Conclusive proof that Kyle was a being from the future way back then LOL!
 
But only a select handful of people understand the difference between a synthetic benchmark and a legit review of real-world use.
 
I bet it's nice to feel validated by one of, if not the most significant PC hardware companies in the world.
 
Kyle, I might have missed one, but could you do another one of your great scaling articles concerning cpu and video. I was always interested where they crossed over. I'd like to see where we have gone in the last few years. Please and Thankyou.:)
 
FINALLY!

Frankly, what good is a device that runs synthetic benchmarks well, when "real world" apps make it falter and fall? Much has been written before about how videocard manufacturers "cheat" with their drivers in order to run benchmarks.

What people often fail to understand, is why this is such a scandalous thing. The bottom line is that it unfairly allows and encourages the development of technology to follow what could be a "dead end," as oppsed to giving the end users the quality of "foprward-looking" technology they pay for!

Look at what's happening around us right now! With the advent of 64-bit technology, software development is falling behind rapidly, for the first time in commercial computer history. Practically speaking, the only segment that's driving the hardware industry is gaming. (Yeah, I know, what about video and audio development? That customer base is dwarfed by gamers.S'Truth!) The majority of PC users are still very well served by 32 bit hardware and software. The only real crisis for them is if their equipment breaks, and are forced to upgrade to a more expensive solution because of the inavailability of otherwise fine equipment.

There's a limit, however, to noticeable differences in framerate and detail. The human eye can only distinguish so much detail...after that limit's reached, anything else is wasted.

(Actually, I thought we'd have already reached the point, by now, where highly-detailed bitmaps could be successfully and rapidly paged for gaming...but it seems that we're still slaves to vector graphics, and however finely detailed they are, they can't match photo quality. If they could, filmmakers would have completely abandoned "real life" long ago; cartoons work for cheap, and they don't have a union.)

So what happens when gamers are led down the wrong path by subjective testing and software? What are the implications for both the future hardware and software developers?

Way to go, Kyle. Remember, I was one of your first supporters in this cause...but I didn't invent the internet. :)
 
Kyle, I might have missed one, but could you do another one of your great scaling articles concerning cpu and video. I was always interested where they crossed over. I'd like to see where we have gone in the last few years. Please and Thankyou.:)

With the industry and ad sales slowing down it has actually allowed us some "free time" to do this. Brent is working on it now. Already getting some interesting results. Will be a LOT of data to analyze.
 
Maybe this is why Intel's chips are so much faster than amd?
 
Anyway, it is good to see a company like Intel finally seeing the light even if this is specific to High Performance Computing. AMD and NVIDIA are still stuck in the days of the 3dfx framerate instead of the experience it can deliver to the customer.

This comes back to min, max, and average to me. Frankly, I would gladly run at a constant 35fps than min of 0, max of 9,000,000 and an average of 120 if I'm allways getting those drops to 0 every few mins (or seconds). But so many review sites only show averages that I think ATI and Nvidia "get it" and are willing to do anything to pump up the average frame rates, I mean just look at the massive shadder power of the 4870. There are a lot of times where it'll be +/-10%, then you get to a zone where it's not that intensive on non-shadders and it zooms to 10,000 fps which drives it's average through the roof. Please, Nvidia and ATI make cards that have higher mins, not just higher averages!
 
With the advent of 64-bit technology, software development is falling behind rapidly, for the first time in commercial computer history.

Short memory? With each hardware advance software lags and for good reason. Recall the the transition to 32bit and the silliness of extended memory managers, Win16->Win32S->Win32, CGA->EGA->VGA->SVGA, etc. Same can be said of instruction set extensions like MMX, 3dNow, SSE(1-5). Lag is built into economy.

Practically speaking, the only segment that's driving the hardware industry is gaming.

Can you support that statement? How much hardware R&D dollars are spent targeting gaming advances vs say increasing efficiency for better battery life? How do you classify the migration toward general purpose computing in GPUs?

So what happens when gamers are led down the wrong path by subjective testing and software? What are the implications for both the future hardware and software developers?

Ever hear of peer review and herding cats? If not, let me summarize: some people are intelligent and people disagree. The implications are left as an exercise for the reader.

Discussing the merits of synthetic vs 'real world' benchmarks is academic. The difference is point of view. Know and understand what is being benchmarked and the possible conclusions that can be drawn. Because that's not always easily known or explain, specific 'end user' applications are used and phrases like 'real world' are coined to engender confidence in measured results.
 
With the advent of 64-bit technology, software development is falling behind rapidly, for the first time in commercial computer history. Practically speaking, the only segment that's driving the hardware industry is gaming. (Yeah, I know, what about video and audio development? That customer base is dwarfed by gamers.S'Truth!) The majority of PC users are still very well served by 32 bit hardware and software. The only real crisis for them is if their equipment breaks, and are forced to upgrade to a more expensive solution because of the inavailability of otherwise fine equipment

Your conclusion is correct (software development is lagging behind), but your premise is false (that 64-bit is the reason). Also, software development *ALWAYS* lags behind (tools have to catch up first, THEN the commercial software).

Software development is lagging behind the multi-core trend. It is a radical shift from previous technologies, and will take massive innovation in the tools area to really take advantage of it on a widespread basis.

There really isn't anything to take advantage of by moving to 64-bit. A ton more RAM address space is pretty much it, but not many programs come anywhere close to the 2GB RAM limit of 32-bit programs (OS limitation). The increased INT size is rather "meh" since the few apps that need to be able to represent a number greater than 2,147,483,648 can just use long long, which is 64-bit on 32-bit platforms. The speed impact is minimal since there aren't many apps that need numbers that large to begin with. Hence consumer and developer adoption of 64-bit is slow, since there aren't any real gains to be had for 95% of the programs out there (even if they are converted to 64-bit, that is).
 
Kyle... you do realize that AMD is one of the contributers to the Phoronix Test Suite? http://www.phoronix-test-suite.com/ : The test that emphasized benchmarking SYSTEM performance over graphics performance? Say, with a starting date BEFORE April 2008? Like, over a calendar year ago? The test suite that did what Futuremark refused to do?

Sorry, but AMD's history invalidates your pot-shot about them being stuck in the 3DFX days.
 
Short memory? With each hardware advance software lags and for good reason. Recall the the transition to 32bit and the silliness of extended memory managers, Win16->Win32S->Win32, CGA->EGA->VGA->SVGA, etc. Same can be said of instruction set extensions like MMX, 3dNow, SSE(1-5). Lag is built into economy.



Can you support that statement? How much hardware R&D dollars are spent targeting gaming advances vs say increasing efficiency for better battery life? How do you classify the migration toward general purpose computing in GPUs?



Ever hear of peer review and herding cats? If not, let me summarize: some people are intelligent and people disagree. The implications are left as an exercise for the reader.

Discussing the merits of synthetic vs 'real world' benchmarks is academic. The difference is point of view. Know and understand what is being benchmarked and the possible conclusions that can be drawn. Because that's not always easily known or explain, specific 'end user' applications are used and phrases like 'real world' are coined to engender confidence in measured results.


Excellent post.
 
Yay to Intel for delivering a better gaming experience to the customer by using their IGP, not like AMD and NVIDIA, which are still stuck in the days of the 3dfx framerate for their discrete GPUs. :p
 
Kyle... you do realize that AMD is one of the contributers to the Phoronix Test Suite? http://www.phoronix-test-suite.com/ : The test that emphasized benchmarking SYSTEM performance over graphics performance? Say, with a starting date BEFORE April 2008? Like, over a calendar year ago? The test suite that did what Futuremark refused to do?

Sorry, but AMD's history invalidates your pot-shot about them being stuck in the 3DFX days.


Yes, I see it leading their marketing effort all the time. :rolleyes:
 
ITT: Kyle was the first to discover benchmarks with real applications before everyone else ever and even owns the patent.
 
Back
Top