AMD Radeon HD 6850 Overclocking Roundup: Asus, XFX, & MSI
by Ryan Smith on November 8, 2010 12:40 AM ESTFrom a variety perspective, the Radeon HD 6800 series is certainly the most interesting Radeon *800 series launches in recent history. AMD typically launches with (and only with) reference cards, and then in time partner-customized cards show up as AMD approves the designs and partners have the time to do the engineering legwork to make custom cards. In the case of the 5800 series this was a particularly long period of time, as TSMC’s production shortage meant that AMD was intentionally shipping out reference cards as fast as humanly possible; and as a result we didn’t see our first custom 5800 series card until 6 months later in February of 2010. It was a much more controlled launch than normal for AMD.
The 6800 series on the other hand turns that on its head, giving us a much more liberal launch when it comes to card designs. While the 6870 series launched and is still all-reference, the 6850 is the opposite, having launched with a number of custom designs. In fact you won’t find a reference 6850 in North America unless you’re a hardware reviewer. With an all-custom launch the door is opened to a wide variety of cards with a wide variety of performance characteristics, so we have wasted no time in collecting a few cards to see what they’re capable of – after all we’ve seen what the non-existent reference card can do, but how about the cards you can actually buy? And how about overclocking, do the latest 6850 cards continue the tradition of the *850 being strong overclockers? Today we’re going to answer all of that and more.
| AMD Radeon HD 6850 | XFX Radeon HD 6850 | MSI R6850 OC | Asus EAH6850 | |
| Stream Processors | 960 | 960 | 960 | 960 |
| Texture Units | 48 | 48 | 48 | 48 |
| ROPs | 32 | 32 | 32 | 32 |
| Core Clock | 775MHz | 775MHz | 820MHz | 790MHz |
| Memory Clock | 1GHz (4GHz effective) GDDR5 | 1GHz (4GHz effective) GDDR5 | 1.1GHz (4.4GHz effective) GDDR5 | 1GHz (4GHz effective) GDDR5 |
| Memory Bus Width | 256-bit | 256-bit | 256-bit | 256-bit |
| Frame Buffer | 1GB | 1GB | 1GB | 1GB |
| FP64 | N/A | N/A | N/A | N/A |
| Transistor Count | 1.7B | 1.7B | 1.7B | 1.7B |
| Manufacturing Process | TSMC 40nm | TSMC 40nm | TSMC 40nm | TSMC 40nm |
| Price Point | $179 | ~$189 | ~$199 | ~$185 |
The first wave of 6850 cards launching were stock-clocked cards. Our intention had been to grab all stock-clocked cards, but manufacturers have been racing to get factory overclocked cards out the door, and we ended up with 2 overclocked cards after all: the Asus with a token 15MHz core overclock, and the MSI with a more serious 45MHz core and 120MHz(480MHz effective) memory overclock. Expect to see many more overclocked cards soon, as manufacturers are eager to get their more profitable overclocked cards out, typically rolling them out along with additional levels of customization such as custom PCBs.
As we’ll see in our performance results, it’s interesting to note that while no two cards are alike in terms of temperature and acoustics, the resulting overclocks were all highly similar. At stock voltage all of our cards could hit at least 850MHz core, and with 6870 voltages (1.172v), all of them hit 940MHz. At even higher voltages such as 1.22v we’re able to push a couple of these cards up to 960MHz core, but it looks like 940-950MHz is the sweet-spot for the 6850 based on the results we’re seeing today. Meanwhile the memory hits a solid wall at 1150MHz (4.6GHz effective); none of our cards would do 1200Mhz (4.8GHz effective) which makes sense given that AMD purposely used a slower memory controller as a tradeoff for a smaller die.
It’s also interesting to note that while the load voltage on our reference 6850 was 1.094v, all of our cards here today (even the non-overclocked XFX) feature a higher voltage of 1.148v. At this point we’re still trying to get to the bottom of this, as AMD hasn’t been able to get back to us with a reason for why we’re seeing this discrepancy. The load voltage is a significant factor for the amount of power drawn (and heat generated) by cards, which means none of our partner 6850s have been able to match the reference 6850 in this aspect. We’re trying to make sure that 1.094v is indeed the 6850’s stock load voltage, or if we need to revise our previous results.
In any case, today we’ll be looking at 3 partner cards alongside our reference 6850: the XFX Radeon HD 6850 (HD-685X-ZNFC), the Asus EAH6850, and the MSI R6850 OC. This represents a diverse group of cards, ranging from short & stubby cards to longer cards with custom PCBs, and everything in between.
Facebook
Twitter
RSS
93 Comments
View All Comments
tech6 - Monday, November 8, 2010 - link
Nice work Ryan.Let's all turn it down a notch about the 460 OC cards. This was never an issue and it still isn't. A number of available cards are chosen for comparison purposes but you can never include all possible variations. So those that are seeing some sort of conspiracy that has Anand taking bags of cash from a vendor in return for leaving out some card in his comparisons - it's time to take off you foil hats and come out from your basements. AT is an excellent and thorough tech review site and to throw out unsubstantiated charges of bias is just a cheap shot. If you have some sort of persuasive proof to contrary then please share it with the rest of us.
silverblue - Monday, November 8, 2010 - link
...it may be better to just compare the 6870 and its OCed variants to the best 460s around in a separate article. It's obvious (at least, to me) that in terms of performance, the OCed 6850s cannot match the 460 FTW and that the true competition would be the OCed 6870. In terms of power and temperature, the 6850 will win, but that's not of much concern when people want the fastest card.Veroxious - Monday, November 8, 2010 - link
Well I guess it's a case of damned if you do and damned if you don't..... IMO you should have included the superclocked EVGA GTX460 (focusing the article on OC results) as it would then be an apples to apples comparison. If one does refer back to the 6850/6870 article you will see that the 2 best overclocking cards (Asus and more so MSI) are neck and neck with the EVGA GTX460 FTW. So buying any of these cards would be an excellent buy.The DIFFERENCE is while the EVGA is arguably a limited version/best case scenario for the GTX 460 it costs $240. There is no price for the MSI card but the Asus card is $185 (out of stock currently on Newegg). As you can see in the article ALL the 6850 cards could hit 940Mhz on stock volts.
So you have a 185$ card that will give you the performance of a $240 card? No wonder there is no stock. At this price ($55 less) it gives the term "value for your money" real meaning. Can I have 2 please?
El_Capitan - Monday, November 8, 2010 - link
I'm sorry, but the EVGA GTX 460 1GB FTW isn't the best overclocking GTX 460 1GB card. It may be the best factory overclocked card, but the winner goes to:1. ASUS from $199.99 to $219.99 where available, hitting 1015MHz core clock.
2. MSI Hawk from $189.99 to $199.99 where available, hitting 960MHz core clock.
Comparing those prices to any of this review's HD 6850 cards that range from $184.99 to $189.00 plus shipping, the prices are equal.
Stuka87 - Monday, November 8, 2010 - link
Am I the only one who is not seeing *ANY* XFX scores on the benchmarks page?None of the graphs have it there??
Lazlo Panaflex - Monday, November 8, 2010 - link
I'm not seeing them either...WTH?Ryan Smith - Monday, November 8, 2010 - link
The XFX is stock-clocked, and when overclocked all of our cards reach common clocks. So we don't break out the scores separately.rpmrush - Monday, November 8, 2010 - link
I love the in-depth vid card reviews here. I really wanna c the custom cooled 6870 cards. I'm waiting to pull the trigger on the one that is quietest with decent overclocking head room.danielkza - Monday, November 8, 2010 - link
At page 8:"All of the cards could hit 850MHz core at stock clocks"
Didn't you mean:
"All of the cards could hit 850MHz core at stock voltages"
tomoyo - Monday, November 8, 2010 - link
It's funny to me that it'll seem like our wants are flipflopping, but the issue is pretty clear.An article about a brand new video card that isn't overclocked, should be compared against the same type of video card.
An article about overclocked video cards should be compared against overclocked video cards of the same market.
Simply put, you unfortunately made the wrong mistake both times, but at least this issue is pretty easy to correct. I hope you also add overclocked GTX 460 results into this one as an update, it does make sense to. The issue originally was that you compared an overclocked GTX 460, but did not compare it to an overclocked Radeon 6850.