Updated Smartphone & GPU Bench DBs Live!
by Anand Lal Shimpi on December 24, 2011 12:08 AM EST- Posted in
- Site Updates
- Smartphones
- Mobile
- GPUs
We've been running our public performance database tool, Bench, for a few years now. Bench allows you to have direct access to the performance data we use in our reviews, segmented according to product category. Today we have five major product categories in Bench: CPU, GPU, Smartphone, Mobile and SSDs.
CPU Bench is the longest running without significant changes. You can actually compare Intel's most recently released Sandy Bridge E CPU to an old Core 2 Duo E6550 from four years ago in a number of tests using the tool.
GPU Bench has always posed a more significant challenge. Drivers and test platforms change very frequently, which often precludes us from keeping a running tally of benchmarks similar to what we do on the CPU side. We used our Radeon HD 7970 review as the kicking off point for our brand new PCIe 3.0 testbed for GPUs going forward - a platform we hope to keep in place for at least the next 12 months. While all of this data is already in the 7970 review, if you want to do any head to head comparisons between GPUs you can now do so using our updated GPU Bench 2012 database. The previous version of the database (GPU Bench 2011) is still available.
Similarly, we've finally updated our Smartphone Bench database. Smartphone Bench 2011 features all of the tests and test data we've been using in our most recent smartphone reviews (e.g. Droid RAZR). We opted against calling it Smartphone Bench 2012 because we plan on updating some of the benchmarks in the new year (e.g. you may have seen one of our new battery life tests in our Eee Pad Transformer Prime Follow-up).
We are in the preliminary stages of planning AnandTech's site refresh for 2012, including updates to Bench. If you have any suggestions you'd like to see implemented, please feel free to leave them as comments here. We already have a number of ideas of things we're trying to implement, but your opinions are the most important here so I'm always eager to listen if you're willing to share.
Facebook
Twitter
RSS
63 Comments
View All Comments
Ryan Smith - Sunday, December 25, 2011 - link
Please keep in mind that not every video card is in Bench yet. I still have a large number of cards on hand that need to be processed, so that only scratches the surface.pc_void - Monday, December 26, 2011 - link
I sure hope so and concur with exostrife.I'm happy to see the old 8800 GT in the list! Its rather close to my 250 and is REALLY nice to comparison - otherwise the db would be almost useless.
So thanks for that :)
repoman27 - Saturday, December 24, 2011 - link
Perhaps it's all in the name of attempting to be platform agnostic, but I've always wondered why this site doesn't have a custom iOS Web Clip icon...pvdw - Saturday, December 24, 2011 - link
I can understand it being difficult to include older CPUs, and in particular to include mobile variants. But, would it be possible to link un-benched CPUs to the closest benched one with a caveat that the figures are approximations. This would help me to demonstrate how much better a new mobile i3 is over, say, a mobile T8100.Maybe link unbenched CPUs in the same family/line with average percentage faster/slower than the benched one - e.g. Athlon II X4 6xx
Thanks!
campbbri - Saturday, December 24, 2011 - link
I love the bench utility; especially the price / performance scatter chart.I have three suggestions:
1. Add benchmarks for Folding@home. I bet if you contacted Vijay Pande at Stanford saying you'd like to use one of their "test" work units and clients that you could run on multiple machines (without interacting with their servers) they could work something out. The benefits to them are increased visability for their cause and of course the actual benchmark results, which would help them see how different CPUs perform. The benefits to Anandtech are increased traffic. The FAH crowd is full of enthusiasts who go to sites like these and would read the detailed articles (and those guys really do care about a 5% performance increase). FAH benchmarks would be an advantage over competing tech sites.
2. I love the price / performance scatter plot. What would be really cool (although maybe hard to implement) is where we can choose each axis in the plot. So instead of price / performance we could look at performance / load power, or plot a heavily threaded performance benchmark against a lightly threaded bench to see how they correlate.
3. For the more casual users, you could create some Anandtech benchmark "profiles" so users can compare for their needs. For example a "gamer" profile would be a weighted average of the gamer benches. You could have separate weighted average benches for "HTPC", "productivity", "media encoder", "granny", etc.
MadMan007 - Saturday, December 24, 2011 - link
I don't like the blog-style front page in the first place, but if it has to stay, fine.What drives me crazy is trying to find a specific Anandtech review article (as opposed to a news post article.) Even using search I have to sift through too many results, and if I go to the correct website sub-sectoin (CPU, GPU, etc) everything 'tagged' a certain way is all just thrown in together: reviews, previews, news blurbs, and so on. I would like to see a separation of full reviews and all the other junk.
LedHed - Saturday, December 24, 2011 - link
It is kind of silly to list the 8800GT, 285 and 470, but leave out the 9800GTX, 295 and GTX 480. I know you guys are just starting, but I hope to see that list grow quickly and always stick to accurate data over quantity.Araemo - Saturday, December 24, 2011 - link
http://www.anandtech.com/bench/GPU12/383The Radeon 6950 has a much better minimum framerate than the radeon 6970? Were those two simply reversed when inserted into the database?
LedHed - Saturday, December 24, 2011 - link
I take back my previous comment, didn't see there were 2 GPU sections : DLedHed - Saturday, December 24, 2011 - link
Although two things I would suggest1. Allow GPU 2012 and 2011 to compare between each other
2. Put in OCd models or run a model OCd