Can anyone explain or propose ideas as to what 'hardware-based' frame metering technology actually means? I know NVIDIA have been very protective of whatever it is they're doing in the 690 (or, if you accept the above PC Perspective interview, in 'all 600 series cards'), but at a more general level I cannot for the life of me think how frame metering can take place in anything other than a software/driver context.
At 1:15 min mark Linus is discussing methodology. "We run all graphics cards overclocked because overclocking is free...."
https://docs.google.com/spreadsheet/ccc?key=0AmLKAgEko3SAdHFtby1jNTFvaTF2UHhaMzdMQ0FNM2c#gid=0
HD7950 OC can beat GTX680 OC at 1440p.
And his methodology is flawed since overclocking is not guaranteed. Nice try.
Also, a single 680 or 7950 is too slow for 1440p MAXXED OUT, especially in newer games. But nice try cherrypicking that res again.
lol so true. its amazing that you can go from 25-30 fps to well above 60 fps in most games by just turning down a few settings. heck in some games just lowering the shadows one notch can double the framerates in some spots.Yet it's something quite a few people are prepared to undertake, so for them it's relevant. People who refuse to overclock can make their purchasing decision based on the massive amounts of data that exist on cards run at stock. Everyone wins.
Fixed it for you. I run a 7970 OC'd at 1440p and...(drumroll)...turn down my settings to reach my desired framerate.
Anyone who insists on playing everything at absolute max, framerate be damned, is a fool. These benchmark results are just that. They're there to help us compare cards in different situations. They're NOT necessarily done at settings that any sane gamer should use.
lol so true. its amazing that you can go from 25-30 fps to well above 60 fps in most games by just turning down a few settings. heck in some games just lowering the shadows one notch can double the framerates in some spots.
Can anyone explain or propose ideas as to what 'hardware-based' frame metering technology actually means? I know NVIDIA have been very protective of whatever it is they're doing in the 690 (or, if you accept the above PC Perspective interview, in 'all 600 series cards'), but at a more general level I cannot for the life of me think how frame metering can take place in anything other than a software/driver context.
Can anyone explain or propose ideas as to what 'hardware-based' frame metering technology actually means? I know NVIDIA have been very protective of whatever it is they're doing in the 690 (or, if you accept the above PC Perspective interview, in 'all 600 series cards'), but at a more general level I cannot for the life of me think how frame metering can take place in anything other than a software/driver context.
In fact, in a bit of a shocking revelation, Petersen told us Nvidia has "lots of hardware" in its GPUs aimed at trying to fix multi-GPU stuttering. The basic technology, known as frame metering, dynamically tracks the average interval between frames. Those frames that show up "early" are delayed slightlyin other words, the GPU doesn't flip to a new buffer immediatelyin order to ensure a more even pace of frames presented for display. The lengths of those delays are adapted depending on the frame rate at any particular time.
Yet it's something quite a few people are prepared to undertake, so for them it's relevant. People who refuse to overclock can make their purchasing decision based on the massive amounts of data that exist on cards run at stock. Everyone wins.
Fixed it for you. I run a 7970 OC'd at 1440p and...(drumroll)...turn down my settings to reach my desired framerate.
Anyone who insists on playing everything at absolute max, framerate be damned, is a fool. These benchmark results are just that. They're there to help us compare cards in different situations. They're NOT necessarily done at settings that any sane gamer should use.
And his methodology is flawed since overclocking is not guaranteed. Nice try.
Also, a single 680 or 7950 is too slow for 1440p, especially in newer games. But nice try cherrypicking that res again.
Nice call out.
My problem is the same as what everyone else with similar setups experiences. It's pretty well documented by now, I'm sure you're aware of it so I'm not sure why you felt the need to attack me.
What's your point? HD7950 OC beats GTX680/7970GE at 1080P too:
http://www.legionhardware.com/artic...z_edition_7950_iceq_xsup2_boost_clock,13.html
This has been known for the last 12+ months. Welcome to 2012.
You keep saying overclocking is luck of the draw but that's exactly why people share their experiences with overclocked parts, why people upgrade their cooling systems, etc. Certain SKUs on average overclock better because they either have higher binned chips or are better build quality (like MSI TF3 7950 overclocking cards).
My point is that those results are not representative for the majority of users. Either because they don't overclock in the first place or because they do not reach those clocks that he got. There always are good and bad clockers out there, even within an SKU.
Therefore his methodology is flawed. I have nothing against separate OC reviews or sections in the review where OC is analyzed. But to make it ONLY about OC is just stupid. And irresponsible btw. Overclocking & stability can be tricky at times (not to mention it increases power draw and noise), and I would never recommend it to someone who is hesitant about these things or has no idea what he/she is doing. It seems you want everyone to overclock, just because AMD cards are good clockers I presume. That is irresponsible and reckless, forcing your view on others. I say let people decide themselves if they want to OC or not instead of advertising it to death in VC&G.
And what do you want with a 7950 OC vs a 680 non-OC? Making uneven comparisons seems to be a serious fetish of yours
Turning down shadows by 1 setting in most games is sufficient for 20-30 fps in many games, as you mentioned. Crysis 3 is another great example, as is metro last light. Crysis 3 in particular looks *EXACTLY* the same with the quality setting at high as it does at very high, yet the framerate difference is huge. I mean, I actually spent time looking at screenshot differences between "very high" and "high" and there isn't a difference. Not even a minute difference, the game looks the same.
most users do overclock.
a PC user who is knowledgeable enough to assemble his own PCs and buy GPUs and install can easily handle light overclocking. its not rocket science. the real enthusiasts who spend big bucks and go all out for max clocks with custom air cooling or watercooling and upgrade frequently (yearly) might be a minority.
for any chip there is the concept of average OC. for HD 7950 boost thats 1100 - 1150 mhz. a really good HD 7950 does 1200 - 1225 mhz. a golden HD 7950 chip is one which 1250 -1300 mhz. 75% of HD 7950 boost cards will hit average OC. the really bad clockers (1000 - 1050 mhz) and really good clockers will be around 10%. the golden would be 5%.
there is no forcing one's views. anybody who is testing hardware provide their testing results and opinion. you can take it or ignore it. the HD 7950 has been the best value for money GPU for the last 9 months. clock for clock HD 7950 is 5% slower han HD 7970, but much cheaper.
Wow thanks for the Metro 2033 tip, playing through it now and its butter smooth! Much better! Hd6970Yeah, there are strange OCD people who are obsessed with maxing games out - i've said this before but there are SO MANY games where you can turn down 1-2 settings and literally double your framerate while the game looks ABSOLUTELY no different.
Turning down shadows by 1 setting in most games is sufficient for 20-30 fps in many games, as you mentioned. Crysis 3 is another great example, as is metro last light. Crysis 3 in particular looks *EXACTLY* the same with the quality setting at high as it does at very high, yet the framerate difference is huge. I mean, I actually spent time looking at screenshot differences between "very high" and "high" and there isn't a difference. Not even a minute difference, the game looks the same. Then there's metro 2033 - Turning ADOF off ups your framerate by 30-40fps....But then you have people paying 1000$ more in hardware just so they can "max" it out when the game looks no different. Just so they can instill a sense of self-pride for having that very high settings. Because you can't technically "max out" a lot of games with any single GPU card at 2560x1600, nor do you need to. It's just the OCD obsessed people who want 8x MSAA w/ very high quality even though the game looks no different with 2-3 settings turned down 1 notch. By all means they can waste 1000-1500$ in hardware just to have that sense of self pride....some people just love wasting money I suppose. Or another way to look at it, the OCD people are supporting the industry I guess? Either way I find the entire thing hilarious.