Archive Home arrow Reviews: arrow 1120-Core "Fixed" Radeon HD 6850 Review Samples Shipped to Media
1120-Core "Fixed" Radeon HD 6850 Review Samples Shipped to Media E-mail
Written by Olin Coles   
Saturday, 23 October 2010

1120-Core "Fixed" Radeon HD 6850 Video Card Review Samples Shipped to Media

Websites publish Radeon HD 6850 reviews with inflated results as a result of hidden 6870 GPU.

Shortly after AMD launched their new Radeon HD 6000 series, which featured their 960-core AMD Radeon HD 6850 and 1120-core AMD Radeon HD 6870 video cards, some reports surfaced that some retail manufacturers had shipped 6850 test samples with the same 1120 shader core Barts GPU that comprises the more powerful Radeon HD 6870 video card.

This creates a major problem for review websites, because many of them had unknowingly published their Radeon HD 6850 video card reviews to the public. Fortunately for AMD, although quite unfortunate for NVIDIA, many curious readers have now incorrectly perceived the Radeon HD 6850 to be capable of performance levels not possible from its correct GPU - even with maximum overclocking. While this incident excludes all unbranded AMD reference samples, partners such as Sapphire, HIS, PowerColor, and XFX appear to have sent "overpowered" samples to reviewers.

AMD-Radeon-HD-6870-Top.jpg

So far, several reviews of Sapphire, HIS, and XFX Radeon HD 6850 products have been discovered to be affected. Benchmark Reviews received a 1120-core Radeon HD 6850 video card from Sapphire a week prior to the launch, and on the same day we also receiving an email from our marketing contact requesting to immediately return the sample for different reasons:

The card has been sent out to you yesterday and you will get it soon.
This is a media sample which does not contain the driver CD as I mentioned my last email.

There is another issue from the factory that there is a possibility of the thermal grease to get dried after shipping to you.
It will affect the card to get hot. Here are the serial no. which has to be recalled.

Q104100000044
Q104100000073
Q104100000014
Q104100000069
Q104100000012
Q104100000068
Q104100000062
Q104100000064
Q104100000019
Q104100000078
Q104100000061
Q104100000040
Q104100000045
Q104100000082
Q104100000013
Q104100000011

If the sample you received is in this list, pls. let me know. I will send you the replacement.
Otherwise, you will be OK.

If your card is not in this list, but if you have a problem due to a heat issue, pls. send me the status in detail.
I will follow up personally.

I am very sorry about this incident.

SAPPHIRE Technology

As it turns out, our Sapphire Radeon HD 6850 was listed in their 'recall' and was returned to them prior to publishing a full review. We did, however, manage to obtain a GPU-Z screen shot prior to shipment which confirms the problem existed:

Sapphire-Radeon-HD-6850-Video-Card-GPU-Z.gif

While I don't have firsthand experience with how HIS or XFX might have handled this, there was not attempt by Sapphire or AMD to alert reviewers of the problem. As an experienced reviewer, I'm skeptical of this incident. It's unclear how many samples were sent to reviewers with this problem, and it's difficult to determine if the action was intentional.

There's no doubt that several of the published Radeon HD 6850 reviews have been tainted with inflated results. Readers should take caution, because there are many single-card and CrossFire reviews that were published using Sapphire, HIS, and XFX Radeon HD 6850 video cards. These articles may have consequently reported abnormally high test results, which could lead to a skewed public perception of the product's natural performance.

Until merchants begin shipping the retail products to consumers, we won't know how for this problem extends. For many, this raises question and concern over unethical sampling from AMD's board partners. At the same time, now there's a consumer interest in unlocking the other 160 shader cores on the Radeon HD 6850.

Official AMD Response:

Apparently a small number of the AMD Radeon HD 6850 press samples shipped from AIB partners have a higher-than-expected number of stream processors enabled.

This is because some AIBs used early engineering ASICs intended for board validation on their press samples. The use of these ASICs results in the incorrect number of stream processors. If you have an HD 6850 board sample from an AIB, please test using a utility such as GPU-z to determine the number of active stream processors. If that number is greater than 960, please contact us and we will work to have your board replaced with a production-level sample.

All boards available in the market, as well as AMD-supplied media samples, have production-level GPUs with the correct 960 stream processors.


Related Articles:
 

Comments 

 
# RE: 1120-Core "Fixed" Radeon HD 6850 Review Samples Shipped to MediaBETA911 2010-10-23 06:14
I'm curious if you can flash the HD6850 with a HD6870 bios and get the full 1120 shaders enabled...
Report Comment
 
 
# RE: RE: 1120-Core "Fixed" Radeon HD 6850 Review Samples Shipped to MediaOlin Coles 2010-10-23 08:24
It doesn't appear that retail 6850's will be able to flash up to the 1120 shader cores on the 6870. This incident occurred with media samples only.
Report Comment
 
 
# fixed?Mikael 2010-10-23 07:27
Fixed? Where does that word come in?
Report Comment
 
 
# RE: fixed?Servando Silva 2010-10-24 08:24
It probably means those samples have more shaders to get better/fixed results.
Report Comment
 
 
# RE: fixed?Olin Coles 2010-10-24 09:03
Fixed, as in a guaranteed win, or doped.
Report Comment
 
 
# RE: 1120-Core "Fixed" Radeon HD 6850 Review Samples Shipped to Mediaaberkae 2010-10-25 16:03
it's called gpu wars! i love it, and this time AMD is playing dirty.
Report Comment
 
 
# At the writer of this articleDaniel 2010-11-22 12:03
Both companies stated that the cards were not what was meant to be shipped to reviewers. Yes, AMD should have checked and made sure that they were not accidentally sending out engineering samples, but they did. Let it go, even at 960 shaders the 6850 is faster than the 460 1gb as it was expected to be.
Report Comment
 
 
# RE: At the writer of this articleOlin Coles 2010-11-22 13:11
Thank you for addressing your comment directly to me, so that I won't feel intrusive for correcting your inaccurate remarks. First, AMD had nothing to do with sending out these samples or giving them added cores. The partner companies did this, and only did it with review media. Your assertion that they should not have been shipping out engineering samples is way off base, as nearly all launch-time products we've received for testing over the past four years have been qualified engineering samples.

Second, this article is one month old. It was written on launch day, because almost 1/3 of the 6850 reviews published used 'juiced' samples. As a result people like you think that the 6850 beats a 1GB GTX 460, when it actually competes with the 768MB version. Just take a look at our own results, which used a legitimate 960-core sample.
Report Comment
 
 
# RE: At the writer of this articleOlin Coles 2010-11-22 13:15
Finally, you fail to grasp the purpose of this article: to educate readers of the error and possible cover-up by card partners. Sapphire still denies it ever happened, and even knew of the problem before the launch. They could have alerted media, but they decided it would be better to simply have us publish inflated results and make their product look better. Even after the launch, they refused to admit the mistake and make mention of the problem to those reviewers who published results with their product.

Again, AMD didn't have a hand in this, but their partners did. My opinion is that the three companies who did it (Power Color, XFX, and Sapphire) all did so with the intention of stepping ahead of their competition by intentionally clocking down samples with too many cores.
Report Comment
 

Comments have been disabled by the administrator.

Search Benchmark Reviews
QNAP Network Storage Servers

Follow Benchmark Reviews on FacebookReceive Tweets from Benchmark Reviews on Twitter