I'm happy to see your 2011 GPU bench is up, I was wondering when you would get the newer cards on there. But one thing I would love to see is average framerates across all the games you tested, once for each of the three tested resolutions. I'm not as concerned with how well a card performs at individual games as I am with its overall performance.
The relative performance of the cards really varies depending on the resolution. If you averaged out the data, you'd completely miss the fact that AMD cards tend to drop off in performance a bit more slowly than NVIDIA cards with higher resolutions. As a result 2560 is a much different case than say 1680.
See that? The benches are so corrupt! There's an Atom N450 in second place, right next to a corei3. Then there are some corei5's at the bottom right inbetween more atom processors! Unbelievable!
I really like how the Anand Bench is improving. Some areas needs work though. For example, values for GPU idle noise levels are pretty useless. I assume these are from reference cards, but if you actually take a measurement from a "real-world" cardd like Asus, XFI, Saphire, etc, the idle noise for them can go well under 20dBa, and not 40+dBa as shown in the Bench.
This is because the way Ryan measures noise levels for GPUs is at a distance of 12" from an open case, and the noise from system fans contributes to the overall noise levels. So in this particular benchmark, the noise floor is ~40dB, though if you close the case and measure at a distance of 3' you'd drop to probably 30-33dB. (It would still be a "tie" however.)
Just to chime in with Jarred, our speakerphone volume level in the smartphone bench has a noise floor of 51.8 dBA. I'm still adding lots of data in there as well.
The N8 should be in there as the N8-00, I had that in there but didn't have it set to live. We're still adding lots of benches to the Smartphone bench in particular.
The smartphone bench is great. Thanks for the creating it. AT has reviewed the Dell Streak and I was hoping you would list it on bench--preferably with the results of it running Froyo.
I think that the bench, in it's current form, doesn't make use of the database as good as it could. I'm trying to find out what graphics card to buy...
But how do I do that? I don't have much use of comparing all cards in a single test, nor do I have much use of looking at all tests of a single card. The only thing usable is to compare two cards, but I feel that is rather limiting.
I would like to be able to choose any combination of tests, and any combination of cards. A list with multiple choices shouldn't be too hard, am I right? But it needs to be a tree instead of a list, don't lose the grouping of tests into categories...
Of course, if you could add pricing too, it would be even better. It would be the ultimate hardware comparison tool!
What's better, I think it would be great if I could just upload a picture of myself to the Bench and have it automatically figure out exactly what CPU and graphics card would be perfect for me, find the nearest retailer, and automatically purchase them for me and have them delivered to my house within 1-2 business days. That wouldn't be too hard, would it? You just have to do some slight tinkering... it would be so great, I wouldn't even have to think before I buy my graphics card and computer.
I have used the Anandtech Bench many, many times to help people choose the right card for their build. Truly this is one of the best features of Anandtech that sets it apart from all the other review websites. While I hate to add more suggestions to the pile to give you more work there are a couple.
1. I know that not every card can be retested for say the 2011 GPU bench, but there are some cards and configurations in tests that don't make the bench, like the 5770 CF which I had to go to the article to compare would be useful. I don't know how many aren't in there, but is it much harder to add every one?
2. It is occasionally useful to know the integrated graphics performance of a CPU. In days past they were are basically terrible, but since AMD and Intel are putting emphasis on the GPU as part of the chip, could the integrated graphics of the CPU be tested and added to the GPU bench? As time goes on the graphics performance of the CPU will only increase and a baseline to compare would be fantastic.
I was just curious as to what the reasoning was to split the GPU benchmarks into two separate categories. Your CPU bench is all in one list, making comparisons between anything rather easy. But the GPU bench is split up, meaning that for older cards, you have to do a two part comparison (comparing an old card to a card on both lists, and then comparing that card to the new card you want to examine).
It seems like it could be because you changed the benchmarks somewhat between years, but again, I look at the CPU bench page. Have you been using the exact same benchmarks for the CPUs this whole time? It just seems like you could at least find some benchmark overlap between most of the cards between the two years.
And that's pretty much my long winded way of complaining that I can't easily compare my 4850 to the new GPUs.
Also, some way of comparing multiple (more than two) of any class of hardware would be excellent.
The scores in CPU Bench don't "go bad"; however for GPU Bench we have to constantly rebuild things as driver releases change performance. To include your 4850 I'd have to rebench it; as it stands I already have 31 different configurations and this doesn't include lower-end cards such as the GT 430.
As a result it's not very practical to cover every last card in existence, so for cards 2+ generations out you're only going to see 1 or 2 cards represented. It's not ideal, but if you know how a 4850 performs relative to a 4870, then you have enough information to make an informed decision.
Using the 50% level of each phones brightness setting doesn't make sense as there is no correlation between the brightness level setting for each phone. What that level represents is specific to each device. Instead, actual measured brightness levels should be used or some method that makes more sense.
For a decently design SoC with decent OS/drivers and application, the screen backlight should be the dominant power user of the system while web browsing or any activity the require only a few seconds of processing for a much large time of the user reading while the CPU can be idle the majority of the time. I suppose Flash advertisements on some pages may keep the CPU from being idle as often.
I'm very disappointed to see 1080p is still not included in the GPU bench. No HDTV runs any higher, so nothing else really matters. Except maybe 900p and 720p.
I also want cards tested all the way down to the HD5670 at least. The 5770 is still out of some peoples price range. Though this is more minor since adding cards is just a matter of running the tests. GTS450?
I'd really like to see an averaged score at each resolution, or just 1080p, for each card across all games you test on. As it stands now I have to go through copy down all the scores and do the math myself every time you release new bench or new cards. Ofcourse when I do it I balance minimum frame rates from Crysis at 60% and the rest at 40% and I only do 4-6 games total; Crysis counts as 2.
Then I'd just like to see Sandy Bridge in the mobile bench. Specifically the 15.6" Clevo boxes with the GTX460M and the GT540M as well as the Compal with the GT540M. Though I'm sure you guys will get on that ASAP all on your own.
Plz add Core i5-2400 to the bench. Is it wise to wait for all the mobos to be replaced or just go and get one and use SATA 6 gbps coz i think i been waiting long enough (2-3 months now) to buy sandybridge
Anandtech would be the last place I would look for unbiased benchmarking. People looking for "professional" reviews should try Techreport, Phoronix, Bit-Tech or any other.
Smartphone bench is awesome. Apart from wanting to see more Graphics benchmarks, it would be great if you can add links in the results from product name to a spec sheet with cpu core, frequency, battery rating, gpu core, frequency, ddr frequency, etc.
<a href="htpp://www.spela-casino.org">casinospel</a> Note, that there are also some desktop processors in the list (because some laptops are equiped with desktop CPUs and for comparison).
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
34 Comments
Back to Article
sircod - Tuesday, February 1, 2011 - link
I'm happy to see your 2011 GPU bench is up, I was wondering when you would get the newer cards on there. But one thing I would love to see is average framerates across all the games you tested, once for each of the three tested resolutions. I'm not as concerned with how well a card performs at individual games as I am with its overall performance.Ryan Smith - Tuesday, February 1, 2011 - link
The relative performance of the cards really varies depending on the resolution. If you averaged out the data, you'd completely miss the fact that AMD cards tend to drop off in performance a bit more slowly than NVIDIA cards with higher resolutions. As a result 2560 is a much different case than say 1680.ProDigit - Saturday, February 5, 2011 - link
See that?The benches are so corrupt!
There's an Atom N450 in second place, right next to a corei3.
Then there are some corei5's at the bottom right inbetween more atom processors!
Unbelievable!
ProDigit - Saturday, February 5, 2011 - link
too bad there is no bench done on the Atom N550 for netbooks.gevorg - Tuesday, February 1, 2011 - link
I really like how the Anand Bench is improving. Some areas needs work though. For example, values for GPU idle noise levels are pretty useless. I assume these are from reference cards, but if you actually take a measurement from a "real-world" cardd like Asus, XFI, Saphire, etc, the idle noise for them can go well under 20dBa, and not 40+dBa as shown in the Bench.JarredWalton - Tuesday, February 1, 2011 - link
This is because the way Ryan measures noise levels for GPUs is at a distance of 12" from an open case, and the noise from system fans contributes to the overall noise levels. So in this particular benchmark, the noise floor is ~40dB, though if you close the case and measure at a distance of 3' you'd drop to probably 30-33dB. (It would still be a "tie" however.)Brian Klug - Tuesday, February 1, 2011 - link
Just to chime in with Jarred, our speakerphone volume level in the smartphone bench has a noise floor of 51.8 dBA. I'm still adding lots of data in there as well.-Brian
halcyon - Tuesday, February 1, 2011 - link
Will you be adding Nokia N8 results to the smartphone bench?Brian Klug - Tuesday, February 1, 2011 - link
The N8 should be in there as the N8-00, I had that in there but didn't have it set to live. We're still adding lots of benches to the Smartphone bench in particular.-Brian
Cat - Tuesday, February 1, 2011 - link
I think.trake1 - Tuesday, February 1, 2011 - link
4890 needs to be placed in the 2011 GPU Batch, not the 2010 please.lvewell - Tuesday, February 1, 2011 - link
The smartphone bench is great. Thanks for the creating it. AT has reviewed the Dell Streak and I was hoping you would list it on bench--preferably with the results of it running Froyo.Brian Klug - Tuesday, February 1, 2011 - link
Oh good point, I need to add those numbers as well. We're still adding lots of benches to the Smartphone bench in particular.-Brian
O8h7w - Tuesday, February 1, 2011 - link
I think that the bench, in it's current form, doesn't make use of the database as good as it could. I'm trying to find out what graphics card to buy...But how do I do that? I don't have much use of comparing all cards in a single test, nor do I have much use of looking at all tests of a single card. The only thing usable is to compare two cards, but I feel that is rather limiting.
I would like to be able to choose any combination of tests, and any combination of cards. A list with multiple choices shouldn't be too hard, am I right? But it needs to be a tree instead of a list, don't lose the grouping of tests into categories...
Of course, if you could add pricing too, it would be even better. It would be the ultimate hardware comparison tool!
Just my thoughts /O8h7w
seapeople - Tuesday, February 1, 2011 - link
What's better, I think it would be great if I could just upload a picture of myself to the Bench and have it automatically figure out exactly what CPU and graphics card would be perfect for me, find the nearest retailer, and automatically purchase them for me and have them delivered to my house within 1-2 business days. That wouldn't be too hard, would it? You just have to do some slight tinkering... it would be so great, I wouldn't even have to think before I buy my graphics card and computer.MeanBruce - Tuesday, February 1, 2011 - link
Dude, you are hilarious! Yuk yuk yuk!driscoll42 - Tuesday, February 1, 2011 - link
I have used the Anandtech Bench many, many times to help people choose the right card for their build. Truly this is one of the best features of Anandtech that sets it apart from all the other review websites. While I hate to add more suggestions to the pile to give you more work there are a couple.1. I know that not every card can be retested for say the 2011 GPU bench, but there are some cards and configurations in tests that don't make the bench, like the 5770 CF which I had to go to the article to compare would be useful. I don't know how many aren't in there, but is it much harder to add every one?
2. It is occasionally useful to know the integrated graphics performance of a CPU. In days past they were are basically terrible, but since AMD and Intel are putting emphasis on the GPU as part of the chip, could the integrated graphics of the CPU be tested and added to the GPU bench? As time goes on the graphics performance of the CPU will only increase and a baseline to compare would be fantastic.
doczero - Tuesday, February 1, 2011 - link
6950 1GB Battleforge 1680x1050 Max quality + 4xAA Doesn't appear to match the scores posted in the review.Review score 65.9
Bench score 23
Thanks.
DarkForceRising - Tuesday, February 1, 2011 - link
I was just curious as to what the reasoning was to split the GPU benchmarks into two separate categories. Your CPU bench is all in one list, making comparisons between anything rather easy. But the GPU bench is split up, meaning that for older cards, you have to do a two part comparison (comparing an old card to a card on both lists, and then comparing that card to the new card you want to examine).It seems like it could be because you changed the benchmarks somewhat between years, but again, I look at the CPU bench page. Have you been using the exact same benchmarks for the CPUs this whole time? It just seems like you could at least find some benchmark overlap between most of the cards between the two years.
And that's pretty much my long winded way of complaining that I can't easily compare my 4850 to the new GPUs.
Also, some way of comparing multiple (more than two) of any class of hardware would be excellent.
Ryan Smith - Tuesday, February 1, 2011 - link
Drivers play a big part of this.The scores in CPU Bench don't "go bad"; however for GPU Bench we have to constantly rebuild things as driver releases change performance. To include your 4850 I'd have to rebench it; as it stands I already have 31 different configurations and this doesn't include lower-end cards such as the GT 430.
As a result it's not very practical to cover every last card in existence, so for cards 2+ generations out you're only going to see 1 or 2 cards represented. It's not ideal, but if you know how a 4850 performs relative to a 4870, then you have enough information to make an informed decision.
DarkForceRising - Tuesday, February 1, 2011 - link
Fair enough. I figured there was something that I hadn't thought of.marraco - Tuesday, February 1, 2011 - link
They are seriously lacking on 560/570 SLI. There is no info, and those are one of the best bang for the buck choices.Funtastico - Friday, February 4, 2011 - link
Agreed! I have a feeling they're working on it and will continue to add upcoming cards to this list. Any comments from Dustin/Ryan?Bless your souls for the hard work you've all put into this. This site shines out amongst its competitors because of commendable work such as this.
Wishlist: Will more upcoming games be added to this list in the future?
=0)
ScruffyNerfherder - Tuesday, February 1, 2011 - link
Using the 50% level of each phones brightness setting doesn't make sense as there is no correlation between the brightness level setting for each phone. What that level represents is specific to each device. Instead, actual measured brightness levels should be used or some method that makes more sense.For a decently design SoC with decent OS/drivers and application, the screen backlight should be the dominant power user of the system while web browsing or any activity the require only a few seconds of processing for a much large time of the user reading while the CPU can be idle the majority of the time. I suppose Flash advertisements on some pages may keep the CPU from being idle as often.
7amood - Wednesday, February 2, 2011 - link
NICCCCCCCCCCCCCE...would love to c how the next generation Galaxy S would compete with the old generation galaxy s and other smart phones.
Aone - Wednesday, February 2, 2011 - link
Bench outlook might be readable better, if "Lower is better" is marked w/ color or graphically.Hrel - Wednesday, February 2, 2011 - link
I'm very disappointed to see 1080p is still not included in the GPU bench. No HDTV runs any higher, so nothing else really matters. Except maybe 900p and 720p.I also want cards tested all the way down to the HD5670 at least. The 5770 is still out of some peoples price range. Though this is more minor since adding cards is just a matter of running the tests. GTS450?
I'd really like to see an averaged score at each resolution, or just 1080p, for each card across all games you test on. As it stands now I have to go through copy down all the scores and do the math myself every time you release new bench or new cards. Ofcourse when I do it I balance minimum frame rates from Crysis at 60% and the rest at 40% and I only do 4-6 games total; Crysis counts as 2.
Then I'd just like to see Sandy Bridge in the mobile bench. Specifically the 15.6" Clevo boxes with the GTX460M and the GT540M as well as the Compal with the GT540M. Though I'm sure you guys will get on that ASAP all on your own.
ssohall - Wednesday, February 2, 2011 - link
Plz add Core i5-2400 to the bench. Is it wise to wait for all the mobos to be replaced or just go and get one and use SATA 6 gbps coz i think i been waiting long enough (2-3 months now) to buy sandybridgeredisnidma - Wednesday, February 2, 2011 - link
Anandtech would be the last place I would look for unbiased benchmarking. People looking for "professional" reviews should try Techreport, Phoronix, Bit-Tech or any other.NateSLC - Thursday, February 3, 2011 - link
Could you update your Droid X numbers with the official 2.2 OS? The numbers there don't mean much to me otherwise.vzv - Friday, February 4, 2011 - link
Smartphone bench is awesome. Apart from wanting to see more Graphics benchmarks, it would be great if you can add links in the results from product name to a spec sheet with cpu core, frequency, battery rating, gpu core, frequency, ddr frequency, etc.alfredska - Tuesday, February 8, 2011 - link
The GPU database is immensely useful.ritagreen - Friday, February 18, 2011 - link
<a href="htpp://www.spela-casino.org">casinospel</a>Note, that there are also some desktop processors in the list (because some
laptops are equiped with desktop CPUs and for comparison).
ritagreen - Friday, February 18, 2011 - link
Note, that there are also some desktop processors in the list (because somelaptops are equiped with desktop CPUs and for comparison).