Thursday, June 16th 2016
MSI and ASUS Send VGA Review Samples with Higher Clocks than Retail Cards
MSI and ASUS have been sending us review samples for their graphics cards with higher clock speeds out of the box, than what consumers get out of the box. The cards TechPowerUp has been receiving run at a higher software-defined clock speed profile than what consumers get out of the box. Consumers have access to the higher clock speed profile, too, but only if they install a custom app by the companies, and enable that profile. This, we feel, is not 100% representative of retail cards, and is questionable tactics by the two companies. This BIOS tweaking could also open the door to more elaborate changes like a quieter fan profile or different power management.
MSI's factory-overclocked GeForce GTX 1080 Gaming X graphics card comes with three software-defined clock-speed profiles, beginning with the "Gaming Mode," which is what the card runs at, out of the box, the faster "OC mode," and the slower "Silent mode," which runs the card at reference clock speeds. To select between the modes, you're expected to install the MSI Gaming software from the driver DVD, and use that software to apply clock speeds of your desired mode. Turns out, that while the retail cards (the cards you find in the stores) run in "Gaming mode" out of the box, the review samples MSI has been sending out, run at "OC mode" out of the box. If the OC mode is how the card is intended to be used, then why make OC mode the default for reviewers only, and not your own customers?Above, you see two GPU-Z screenshots, one of the TPU review sample, next to the retail board (provided by Nizzen). Flashing the retail BIOS onto our review sample changed the clocks to match exactly what is shown on the GPU-Z retail screenshot.
In case of the GTX 1080 Gaming X, the "Gaming mode" runs the card at 1683 MHz core and 1822 MHz GPU Boost; and the "OC mode" runs it at 1708 MHz core and 1847 MHz GPU Boost. The cards consumers buy will run in the "Gaming mode" out of the box, which presumably is the default factory-overclock of these cards, since they're branded under the "Gaming series."The "OC Mode" is just there so consumers can overclock it a little further at the push of a button, without having any knowledge of overclocking. Now if the OC mode is enabled for review samples of one company and not for the others, this means that potential customers comparing reviews will think one card performs better than the other, even if it's just 1%, people do base their buying decision on such small differences.
With the case of the GTX 1080 at hand, we started looking back at our previous reviews and were shocked to realize that this practice has been going on for years in MSI's case. It looks like ASUS has just adopted it, probably because their competitor does it, too, "so it must be ok."It's also interesting to see that not all cards are affected, whether this is elaborate or by accident is unknown.
While we don't have any statements of the companies yet, the most likely explanation is that reviewers usually don't install any software bundled with the graphics card, yet the companies want the cards to be tested in OC mode, which provides higher performance numbers, beating their competitors. That's probably how this whole thing started, nobody noticed and the practice became standard for reviews moving forward.
This issue could affect upcoming custom GeForce GTX 1070 review samples too, we will be on the lookout.
MSI's factory-overclocked GeForce GTX 1080 Gaming X graphics card comes with three software-defined clock-speed profiles, beginning with the "Gaming Mode," which is what the card runs at, out of the box, the faster "OC mode," and the slower "Silent mode," which runs the card at reference clock speeds. To select between the modes, you're expected to install the MSI Gaming software from the driver DVD, and use that software to apply clock speeds of your desired mode. Turns out, that while the retail cards (the cards you find in the stores) run in "Gaming mode" out of the box, the review samples MSI has been sending out, run at "OC mode" out of the box. If the OC mode is how the card is intended to be used, then why make OC mode the default for reviewers only, and not your own customers?Above, you see two GPU-Z screenshots, one of the TPU review sample, next to the retail board (provided by Nizzen). Flashing the retail BIOS onto our review sample changed the clocks to match exactly what is shown on the GPU-Z retail screenshot.
In case of the GTX 1080 Gaming X, the "Gaming mode" runs the card at 1683 MHz core and 1822 MHz GPU Boost; and the "OC mode" runs it at 1708 MHz core and 1847 MHz GPU Boost. The cards consumers buy will run in the "Gaming mode" out of the box, which presumably is the default factory-overclock of these cards, since they're branded under the "Gaming series."The "OC Mode" is just there so consumers can overclock it a little further at the push of a button, without having any knowledge of overclocking. Now if the OC mode is enabled for review samples of one company and not for the others, this means that potential customers comparing reviews will think one card performs better than the other, even if it's just 1%, people do base their buying decision on such small differences.
With the case of the GTX 1080 at hand, we started looking back at our previous reviews and were shocked to realize that this practice has been going on for years in MSI's case. It looks like ASUS has just adopted it, probably because their competitor does it, too, "so it must be ok."It's also interesting to see that not all cards are affected, whether this is elaborate or by accident is unknown.
While we don't have any statements of the companies yet, the most likely explanation is that reviewers usually don't install any software bundled with the graphics card, yet the companies want the cards to be tested in OC mode, which provides higher performance numbers, beating their competitors. That's probably how this whole thing started, nobody noticed and the practice became standard for reviews moving forward.
This issue could affect upcoming custom GeForce GTX 1070 review samples too, we will be on the lookout.
162 Comments on MSI and ASUS Send VGA Review Samples with Higher Clocks than Retail Cards
Also, most sites review cards using their default clocks and maximum overclock. Its fine like this, and consistent.
If they did install it, should they try all profiles and post results for all 3 of them? And if not, which one should they choose? I would say the default, "gaming" one. Which you can get without the software.
I wonder, how about the all review for other cards on the past..?
is this the first time?
i'll would be surprised if other consumer electronics won't suffer from same "illness"; same shi.... like VW emission scandal.... one catched and the rest are coming...
this is the ugly face of competitiveness between companies... but let's not forget that behind any companies are people who implement their own way of thinking
"in $ we trust"
I have seen some Gigabyte products get new revisions during the product life cycle. There is nothing wrong in this in general, but sometimes the new revision is more like a new product, arguably changing the used components so much (sometimes to worse as well!), so what is sold no longer reflects the initial reviews done with REV1.0.
Pretty much all manufacturers are doing this, what could you do better to distinguish the differences to consumers who are buying your products? At the moment you can't know whether you are getting lets say REV1.0 board or REV1.1 which could have some significant changes.
At least some of the revisions are listed on your product pages with photos of the boards, so you are already ahead of others. :)
--
Worst guys in this gimping over product life cycle thing are arguably XFX and TUL corporations like PowerColor. I have also seen MSI products with cut down reference design components advertised as "Military class quality", ofc in that case its complete BS copy paste on product page. And Asus conveniently removing significant part of VReg bulk capacitance by leaving "unnecessary" components off the board after initial samples. List is long...
The workaround for people in the know is just to dial in the advertised clock rates as OC'ing in Precision X or Afterburner is going to have the same effect as the stupid gaming app. That's what I did, I have a profile that is the standard "OC" profile that was listed on MSI's website, and a profile where I crank it up, and one where it's just at the default rate.
I guess my opinion is that anyone reading all these reviews is going to install the app or overclock on their own anyway, and the ones that just buy the card and throw it in are fine with the "Gaming" profile that is default in the VBIOS. Sure, it's stupid to try to force people to install your stupid gaming app to get the full guaranteed speed of the card, but if the only change to review samples were to dial in the advertised speeds ahead of time so that the reviewers don't mess up the testing setup, I think it's understandably non-malicious. MSI just really needs to reexamine why they are adding barriers to get their card's top speed in the first place. That gaming app is useless and robs people of performance unless they aren't stupid, but MSI was not maliciously inflating performance past what any owner of the card is guaranteed to achieve.
Looking @ overclockers.co.uk just as an example it's sold @ OC clocks for what it is worth.
www.overclockers.co.uk/msi-geforce-gtx-970-gaming-edition-4096mb-gddr5-pci-express-graphics-card-gx-259-ms.html
Geforce Experience is wretched, uninstalled that a day after installing it. Gigabytes software for this motherboard I have is god-awful, uninstalled it almost as fast as I uninstalled it. They are always so clunky and bug-ridden.
Now they want you to install it so you can have your extra 30Mhz.... lol. Sad...
Those of us old enough will remember the benchmark tricks of 2003 when NVIDIA deliberately took rendering shortcuts with the likes of fixed 3DMark benchmarks to get a significantly higher score. When the view was changed, a corrupted scene was rendered (the sky I think) revealing the trickery, resulting in a scandal which changed how reviewers benchmark to this day. I know ATI wasn't totally immune from this either, but NVIDIA was worse.
And I'm a long time NVIDIA user, so no fanboyism here.
@W1zzard Will you be updating your reviews to take this into account? I'm thinking running the cards at the same clocks as a customer would get and inserting the benchmark into the results, plus a short writeup about it, with an adjustment to the review score if necessary.
I thought it was odd how the Founder's Edition cards were hitting their thermal limits like that. An unofficial overclock could explain it.
Now, if they had sent cards with a special bios that allowed a higher power envelope, or higher temp limit, or something that is not attainable to a consumer than that is different. While I think the 970 memory stunt was ridiculously douchey, people were still buying it based on the reviews which showed the card performed exactly as it should.
Hell as it stands you'd think AotS is the best game ever made.... meh.
The new ones get leds control with them but using afterburner is better because whenever i run gaming app it makes the clocks for my r9 270x go to 1050/1400 (silent mode) by default & when i close it my idle clock is 300/300 not 300/150.
I oced my card to 1200/1500 with afterburner & there is no option for us to run custom profiles from gaming app.
So yes avoid that software & use afterburner.
As @danbert2000 said, they should think hard about why users have to jump through hoops to get the extra performance.
It does depends though. How well advertised is the feature?