3DMark Windows update: hardware monitoring fixed, Ice Storm Unlimited added

We've also added the Ice Storm Unlimited test for benchmarking Windows 8 tablets and other mobile devices. You can compare Ice Storm scores from the latest Windows, Android and iOS devices.

The Steam version of 3DMark updates automatically; the standalone will prompt you to install the update when you next run the program.

New in 3DMark Windows v1.2.250

  • Adds Ice Storm Unlimited test for comparing Windows 8 tablets with the latest Android and iOS devices.
  • 3DMark now uses technology provided by TechPowerUp for improved GPU hardware detection.
  • Fixes hardware monitoring. The performance graphs now show clock speeds and temperatures for the CPU and GPU again (with compatible hardware in 3DMark Advanced and Professional Editions only).


6

Belgium Massman says:

"Yeey" for better hardware detection. "Meh" for the additional Ice Storm test. "Why still not" for not having cheat checks in the GUI, for not showing version number in the score window, for not showing system info in the score window, for not checking rendered image quality (to void lod, tesselation).

Finland FM_Jarnis says:

In-GUI (offline) cheat checks are pointless as they are too trivial to hack. We could spend a month of dev time to make one, then hear two weeks later how there is a result that 3DMark GUI says is fine, but in reality is completely bogus, because someone "tweaked" the GUI code a bit. Only way to be sure is to do any checks away from the hands of the "enemy" (ie. system/user being checked for cheats), so on a server. We offer that. "Online Validation". Shows even in the GUI when result passes it - but yeah, it requires online connection for obvious reasons. Also if someone presents such a screenshot, you can obviously then ask for score validation link to verify it yourself (from 3DMark.com) as there then is one. That link is also visible on the "View Result Details" page. As for versions, check the result screen. Each workload has little "v1.1" (except Ice Storm is now "v1.2") next to the test name. That is the workload version which changes whenever there is any score-relevant changes to the benchmark. As for system information, for that we have the "view result details". Open that. There was no way to fit all that information together with the result on the main result page of the UI.

United States Gunslinger says:

"Why still not" for not having cheat checks in the GUI, for not showing version number in the score window, for not showing system info in the score window, for not checking rendered image quality (to void lod, tesselation).


+1,000,000

In-GUI (offline) cheat checks are pointless as they are too trivial to hack. We could spend a month of dev time to make one, then hear two weeks later how there is a result that 3DMark GUI says is fine, but in reality is completely bogus, because someone "tweaked" the GUI code a bit.


We, the paying customer, have been asking for this very "trivial" cheat check to be displayed in the GUI for quite some time.

I work at a manufacturing facility, and we take our customers product requests very seriously, no matter how trivial they may seem. We are in business only because of them, if we don't give them what they want, they will find someone who is willing to do so.

FM now has some competition from Unigine and the folks from Catzilla, so maybe it's high time you guys started listening a little closer to your customers.

Finland FM_Jarnis says:

It is not trivial to do it. In fact, it would require quite a bit of code porting from server side to client side and some of the checks could not be done at all (ones based on comparative data). If it were really trivial to add it to the client for offline, we would have done it years ago.

It is trivial to HACK it if one were to be added. Hence the current view of FM is that it would be a lot of effort for minor benefit. It would also open the ability to have unverifiable screenshots that claim the result to be okay, except since no copy of the result exists on our server, there is absolutely no way to verify such validity after the fact.

The client already has a check (result showing "valid result") if the result is validated but this requires online connection. No, it currently doesn't have a way to check for NV LOD thing - this is based on the fact that this requires registry editing to do (while comparable AMD tessellation thing is a driver setting that any normal user may do by accident). Argument can be made that we should flag them anyway - feel free to send feedback on the subject. I personally think it would be a good idea to do but it is not my decision.

3DMark is *not* hardened against determined on-purpose cheating (as opposed to a system configuration that just isn't comparable that could happen without on-purpose cheating).

Such a thing cannot be done without heavy-handed monitoring (punkbuster-style) of processes that would most likely influence benchmark scores due to system resources required. Only way to root out those is by checking the result against a large pool of results from similar hardware. Can't be done without our result database, so can't be done offline. Key part of this is accurate hardware detection - and we've licensed the current "gold standards" on both CPU and GPU side to improve that because it is important (and because re-inventing the wheel is a waste of resources)

United States Gunslinger says:

There is a saying "locks only keep honest people out"

If someone is determined to cheat/hack a benchmark they are going to do it regardless.

The changes we've requested, we've requested for our needs and desires. Nobody said it was going to be easy to implement. If you're using the excuse that it's too much work for little benefit, you're missing the message that this is what your paying customers want.

GENiEBEN says:

FM_Jarnis said: Such a thing cannot be done without heavy-handed monitoring (punkbuster-style) of processes that would most likely influence benchmark scores due to system resources required


Although I agree with you, this is a necessary evil. Too late to implement on current benchmark(s), but a must-have in next iterations.

Btw Jarnis, check your PM later.

Please log in or register to comment.