What good are benchmarks when a PC is ‘good enough’?

[jaydeejohn]

"So - how do benchmarking companies/sites continue to exist in a year or two when most games run at >60FPS on a i3-randomnumber/Llano? "
My response
Sorry for the non quote.
I'm referring to sites that still thrive in this direction, compared to those that have gravitated away and adapted, as the enthusiast market shrinks

[The Crouch]

Originally Posted by lurrker

<sweepingGeneralisations...>
The PC market (as viewed from a benchmarking perspective) can be viewed as around 5 segments:
- basic requirements (i.e. any current PC will do. Would expect to be able to do web browsing, typical MS Office-type functions, watch movies/videos)
- basic gaming (i.e. as above but with sufficient GPU power to drive a 720p display at >30fps for MOST games at playable graphics settings)
- gaming/high-end office (i.e. fast CPU/GPU)
- ultra highend (top of the range everything....)
- specialist task (i.e. server or workstation where workload is well known and bottlenecks have been analyzed)
</sweepingGeneralisations...>

I would guess that the two basic categories would cover >40% of the market, with specialist taking another 50% and the remaining categories filling out the remaining 10%.

The bottom of the range Core2 and later CPU's have fulfilled the basic category CPU requirements and current generation APU's are there or thereabouts for covering basic CPU/GPU requirements. In addition, non-PC devices are eroding this market.

The specialist task and ultra highend markets will remain, but costs are likely to go up to maintain performance increases as much of the low hanging fruit (more cache/cores/higher clocks/die shrinks) have been used to achieve current performance levels leading to larger, more complex (and more expensive) chips in the future.

For gaming/highend office functions, you no longer need the best equipment to carry out the PC functions. Consoles have reduced and "dumbed down" the PC gaming market and many games look very good without using the ultra graphics settings if your GPU is unable to support them at playable frame rates.

So - how do benchmarking companies/sites continue to exist in a year or two when most games run at >60FPS on a i3-randomnumber/Llano?

Thinking of the Tech HQ's of the companies I worked the longest at (Saab Auto and Volvo cars), everyone had a pc, some even two. None of those were high spec, even when new.

For workstations, I'm not sure how you meant to define them. As far as I know, Most CAD sw run under windows now (for example UGS, Catia, ProE) and have been for at least 5-10 years. They would still have a Quadro card (big Nv dominance...), but the rest would b PC components. Hopefully better grade, but still "just" PC components. Some of the people where I am now (Volvo trucks) run ProE on laptops.

Anyway, if I use the definition of workstation as running sw other than MS office apps, basically counting a CAD station as a workstation, regardless of the hw, the number of workstatons would still be very hard pressed to reach even 10% share.

[Copper]

Originally Posted by jaydeejohn

"So - how do benchmarking companies/sites continue to exist in a year or two when most games run at >60FPS on a i3-randomnumber/Llano? "
My response
Sorry for the non quote.
I'm referring to sites that still thrive in this direction, compared to those that have gravitated away and adapted, as the enthusiast market shrinks

ah too funny. I saw a behavior change in a couple of those sites about a year ago. It's going to get worse as there is consolidation. I'm glad to be a bystander in many ways.

[caveman-jim]

Originally Posted by lurrker

So - how do benchmarking companies/sites continue to exist in a year or two when most games run at >60FPS on a i3-randomnumber/Llano?

Because this is not true.

[vain]

I didn't know where else to post this, so out of stupidity, most likely, I'm posting it here.

Had some fun with a KillAWatt...

MSI-X370==E-450==13.4"==M4-SSD

Idle + Display off completely = 6.3 Whats
Idle + Lowest screen bright = 8.3 Whats
Idle + Full screen brightness = 11.6 Whats
Idle + turn on WLAN + display full = 12.1 Whats
CPU stress test +WLAN on + display full ~ 17.5 Whats
CPU/GPU stress test + WLAN on + display full ~20.3 Whats


It is undervolted, here are the P-states:

P-State 1650 MHz - FID 0x100 - VID 0x1C - IDD 5 (16.50x - 1.200 V)
P-State 1320 MHz - FID 0x102 - VID 0x28 - IDD 3 (13.20x - 1.050 V)
P-State 825 MHz - FID 0x300 - VID 0x34 - IDD 2 (8.25x - 0.900 V)

Original vcore was I think...

P0 of 1.35v
P1 of 1.20v
P2 of 1.05v

Could restart without BrazosTweaker if anyone really wants to know, but I'm just posting this to provide some numbers towards the article. Would be nice if I could figure out a way to get power draw of just the idle CPU.

[charlie]

Originally Posted by vain

I didn't know where else to post this, so out of stupidity, most likely, I'm posting it here.

Had some fun with a KillAWatt...

MSI-X370==E-450==13.4"==M4-SSD

Idle + Display off completely = 6.3 Whats
Idle + Lowest screen bright = 8.3 Whats
Idle + Full screen brightness = 11.6 Whats
Idle + turn on WLAN + display full = 12.1 Whats
CPU stress test +WLAN on + display full ~ 17.5 Whats
CPU/GPU stress test + WLAN on + display full ~20.3 Whats


It is undervolted, here are the P-states:

P-State 1650 MHz - FID 0x100 - VID 0x1C - IDD 5 (16.50x - 1.200 V)
P-State 1320 MHz - FID 0x102 - VID 0x28 - IDD 3 (13.20x - 1.050 V)
P-State 825 MHz - FID 0x300 - VID 0x34 - IDD 2 (8.25x - 0.900 V)

Original vcore was I think...

P0 of 1.35v
P1 of 1.20v
P2 of 1.05v

Could restart without BrazosTweaker if anyone really wants to know, but I'm just posting this to provide some numbers towards the article. Would be nice if I could figure out a way to get power draw of just the idle CPU.

Nice numbers. Can you try one test for me, try the screen off, low, high one with both a static screen (IE desktop) and with a full screen movie playing with lots of movement. I want to see not just backlight power, but what screen switching adds. Just curious mainly.

-Charlie

[vain]

Originally Posted by charlie

Nice numbers. Can you try one test for me, try the screen off, low, high one with both a static screen (IE desktop) and with a full screen movie playing with lots of movement. I want to see not just backlight power, but what screen switching adds. Just curious mainly.

-Charlie

I used a surge protector (extension) that splits 4 ways and has a dim orange light on it, plugged it into the watt meter, to which I plugged in whatever I was using.

Screen 1: Laptop LED
Screen 2: Samsung Syncmaster 216BW 22"(?) LCD connected by VGA (energystar )

Everything unplugged but extension: ~0.2 W (jumps from 0.3 to zero every so often, doesn't like reading so low?)

Plug in laptop - battery + idle + screen 1 off : 6.5 W
laptop + idle + screen 1 low : 8.5 W

Screen 1 off + video in background (baseline) : ~8.5 W
Screen 1 low + video : ~10.9 W
Screen 1 high + video :~ 14.0 W

Now put battery back in, unplug laptop from extension.

Extension cord again: ~0.2 W
Plug in screen 2 and wait for standby mode: 0.8 W
Screen 2 Idle low bright: 17.6 W
Screen 2 Movie low bright: ~17.7 W
Screen 2 Idle high bright: 40.5 W
Screen 2 Movie high bright: ~40.6 W

I had to do this multiple times, it was hard to actually see a different in a movie playing and a movie not playing. It would fluctuate every so often slightly up 0.2 or 0.3 W or so. Honestly it was really negligable, and could just be due to me being biased in looking for an increase. I did notice that when the LCD was first plugged in, it would use up to 43W, and had to wait for it as it slowly dropped down to stable. Did it three times to make sure it wasn't a fluke. Not sure if this is an issue with meter or LCD.

edit: The reason I had the numbers with surge protector was because I originally was going to do both the laptop and LCD plugged into it also, but then partway through it figured it would just be redundant data.

edit: Also, these, and the numbers before, are just taken after waiting about a minute and seeing what the meter displays. They're not really scientific in terms of a mean. I'd have to run it for a while and get Wh, but the results are good for curiousity sake.

[vain]

Originally Posted by charlie

Nice numbers. Can you try one test for me, try the screen off, low, high one with both a static screen (IE desktop) and with a full screen movie playing with lots of movement. I want to see not just backlight power, but what screen switching adds. Just curious mainly.

-Charlie

So it looks like the cpu playing the movie in the background is using up about 2 W. The display itself is maybe using anywhere from slim to half a watt at most extra from the movement. Considering margin of error, it's almost negligible.

Old LCD screens seem to be quite a draw though, it's amazing how far LEDs have come.

[charlie]

Originally Posted by vain

I used a surge protector (extension) that splits 4 ways and has a dim orange light on it, plugged it into the watt meter, to which I plugged in whatever I was using.

Screen 1: Laptop LED
Screen 2: Samsung Syncmaster 216BW 22"(?) LCD connected by VGA (energystar )

Everything unplugged but extension: ~0.2 W (jumps from 0.3 to zero every so often, doesn't like reading so low?)

Plug in laptop - battery + idle + screen 1 off : 6.5 W
laptop + idle + screen 1 low : 8.5 W

Screen 1 off + video in background (baseline) : ~8.5 W
Screen 1 low + video : ~10.9 W
Screen 1 high + video :~ 14.0 W

Now put battery back in, unplug laptop from extension.

Extension cord again: ~0.2 W
Plug in screen 2 and wait for standby mode: 0.8 W
Screen 2 Idle low bright: 17.6 W
Screen 2 Movie low bright: ~17.7 W
Screen 2 Idle high bright: 40.5 W
Screen 2 Movie high bright: ~40.6 W

I had to do this multiple times, it was hard to actually see a different in a movie playing and a movie not playing. It would fluctuate every so often slightly up 0.2 or 0.3 W or so. Honestly it was really negligable, and could just be due to me being biased in looking for an increase. I did notice that when the LCD was first plugged in, it would use up to 43W, and had to wait for it as it slowly dropped down to stable. Did it three times to make sure it wasn't a fluke. Not sure if this is an issue with meter or LCD.

edit: The reason I had the numbers with surge protector was because I originally was going to do both the laptop and LCD plugged into it also, but then partway through it figured it would just be redundant data.

edit: Also, these, and the numbers before, are just taken after waiting about a minute and seeing what the meter displays. They're not really scientific in terms of a mean. I'd have to run it for a while and get Wh, but the results are good for curiousity sake.

Can you try without the second monitor, and no power strip? If you want to turn the backlight off, just set the machine to not sleep when the lid is closed, then press the switch that is depressed when you close the screen. That should turn backlight off but no the screen off.

Alternatively, you could put it on the second screen and not put that on the power strip, but that changes your measurement from screen switching to VGA power out measurements. Meh.

-Charlie

[vain]

Originally Posted by charlie

Can you try without the second monitor, and no power strip? If you want to turn the backlight off, just set the machine to not sleep when the lid is closed, then press the switch that is depressed when you close the screen. That should turn backlight off but no the screen off.

Alternatively, you could put it on the second screen and not put that on the power strip, but that changes your measurement from screen switching to VGA power out measurements. Meh.

-Charlie

Whenever i say screen off for the LED laptop screen, it is simply me disabling any auto suspend/hibernate in the power options and closing the lid. I guess that would more correctly be the backlight off.

The first part of the video portions (the one with the LED laptop data) was done with the LCD monitor (2) compeletely unplugged, and no VGA attachment attached to the laptop. Only when I was doing the portion with data including the LCD monitor (2) did I plug the LCD monitor into the strip and unplug the laptop charger from the strip.

So this (below) is without the LCD external monitor plugged into the strip at all. Only the laptop.

Plug in laptop - battery + idle + screen 1 off : 6.5 W
laptop + idle + screen 1 low : 8.5 W

Screen 1 off + video in background (baseline) : ~8.5 W
Screen 1 low + video : ~10.9 W
Screen 1 high + video :~ 14.0 W

This (below) is when I put the battery back in the laptop and unplugged it from the extension, then plugged in the external monitor and ran the video off the battery. This shows only the external monitor power draw, no laptop power draw whatsoever.

Now put battery back in, unplug laptop from extension.

Extension cord again: ~0.2 W
Plug in screen 2 and wait for standby mode: 0.8 W
Screen 2 Idle low bright: 17.6 W
Screen 2 Movie low bright: ~17.7 W
Screen 2 Idle high bright: 40.5 W
Screen 2 Movie high bright: ~40.6 W

In regards to the external screen and VGA numbers. I could take a measurement of just the power draw of the laptop, without plugging in the external monitor (2) into the same socket, while both the LED screen (1) and the external screen (2) are working (extended desktop or w/e). Then switch to external only. This would theoretically give a delta of the screen (1) actually off versus being on I presume? I would be running VGA during both measurements, with the only difference in the LED (1) being on or off.

I'll redo it again, with the above in mind and the surge protector in mind, when I get home tonight, but I don't consider this instrument that sensitive, as to show additional VGA output power consumption from the laptop (I'm assuming it's on the order of tens-hundred of mW).

Edit: I would think that the image on the screen, it's actual brightness and whether or not there is a lot of "white" being displayed, would create a margin of error large enough, while playing video, that things like surge protector and VGA output become meaningless. I will see tonight.