Hardware cheating

Source: http://videocardz.com/61121/asus-and-msi-accused-of-sending-modified-cards-to-the-press


Interesting article was just posted on Hardware.fr.

According Damien Triolet, a known and respected GPU reviewer (who, as you might remember, helped to unravel the secrets in GTX 970 memory allocation problem), ASUS and MSI are sending cards with modified BIOSes to the press. Such software enables more power on review samples, which leads to increased frequencies and better results overall.

Damien reports that those ‘optimized’ BIOSes are a very common problem in GPU industry. Manufacturers often encourage GPU reviewers to enable special overclocking presets before attempting to review those cards. Luckily, with little success.

The alternative is to supply optimized BIOSes to the press, so such settings are enabled by default. This usually means a gain of few MHz. Something that you won’t get on retail sample.

For such reason Damien asks manufacturers to supply retail BIOSes for his tests. Obviously manufactures are not eager to supply such software.

The problem was discovered with MSI GTX 1080 GAMING X and ASUS GTX 1070 STRIX, so might want to take reviews of those cards with a grain of salt. Gigabyte on the other hand does not use such practices with its G1 GAMING Series.

Damien Triolet, Hardware.fr:

Obviously, Asus and MSI have made the calculation that getting these small gains brought more benefits than criticism among some troublemakers. Sure, after all, so these bios boost perhaps the performance of the cards by 1%, but that’s no reason to look elsewhere! After 1%, it will be what? 2%? Then 3%? Then widespread cheating contest?

UPDATE: Guys over at TechPowerUP have just confirmed that the problem is present with their MSI sample as well:

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.

2 Likes

Cheeky buggers, doesn’t surprise me though, the hardware industry is half crooked. I only really respect AMD/ARM nowadays. They have the ‘cleanest’ record and seem to strive for openness the most, sharing a lot with developers and the world in general.

I don’t think it makes a difference. Either you’re an over-clocker/expert and you know about it or your ignorant of it and don’t care for 1%- 3% boost anyway.

I heard Ferrari does the same when their road cars are reviewed :wink:

Wait… you’re telling me that commercials and manufacturers claims aren’t 100% reliable?



1 Like

2 Likes

2 Likes

Lemme at’em!

There’s also a non-conspiracy possibility: late development boards are what get sent to the press, but it’s possible that the performance step-downs get added on the retail boards once all the power budgets and life-cycle testing is complete, to ensure the consumer product meets performance expectations as well as power-consumption, temperature, and lifetime expectations.

While I could certainly believe that a developer would want to get higher marks in reviews, “cheating” to just get a 1% advantage seems pretty malicious for a relatively insignificant gain, especially if everyone does it.

Just sayin’.

From reddit:

The ASUS 1080 has three different modes, silent, gaming and OC. Those are available for everyone, it’s just a shortcut on the keyboard (ctrl-alt-1 for OC). Since reviewers aren’t always the smartest I can understand that they pre choose the highest factory OC. I’d get annoyed if they reviewed and benchmarked the card at silent mode by accident.

More fuel for the fire, love neighbors

1 Like