Techpowerup: Powerecolor HD6870 PCS++

[3DVagabond]

http://www.techpowerup.com/reviews/Powercolor/HD_6870_PCS_Plus_Plus/1.html

I don't typically post reviews of cards. This card's a bit different, though. It comes with a healthy O/C (975MHz) and very good out of box performance, but that's not the big deal, IMO. I'd imagine almost any 6870 could be O/C'd to match it. No, this 6870 has 2x Crossfire fingers. Used to be that was standard, but on this gen of cards it has been reserved only for the high end cards. (The same with SLI, unfortunately.)

Seems like some people don't consider Powercolor an upper echelon AMD partner. They reserve that for Sapphire, Gigabyte, and XFX. Powercolor does make some interesting enthusiast cards though. From their passive cooled HD6850 and LCS (water cooled cards), to the HD6950 PCS++ which comes "pre-unlocked" for those who are concerned with doing it themselves. Now the tri/quad crossfire capable 6870. I think they deserve some enthusiast consideration. They seem to be listening to what people want.

 

[cusideabelincoln]

Well, that's a pretty large jump in performance, especially in all of the DX11 games compared to the DX9/10 games. Were there different drivers used for the 6870++ than the other 6870? He doesn't specifically state which drivers were used for the ++. I have a suspicion he used the same Beta drivers as he did for the 6990 while the other Radeons were on 11.1.

And he did get the core up to 1020 and memory up to 1210. Not terrible, and the scaling looks good with the one bench he did. At these speeds it would put the card ahead of the 5870, and it looks like most especially in DX11 titles.

 

[nemesismk2]

I am a big fan of Powercolor video cards so would have to consider this Powerecolor HD6870 PCS as a future upgrade from my Powercolor Radeon 5770.

 

[wahdangun]

wow faster than GTX 560Ti, and cheaper,

 

[Gloomy]

Wow, that's a really impressive card.

 

[nanaki333]

they're ok cards.. they just don't have an OK warranty. 2 year warranty is laughable when a lot of other equally priced manufacturers do 3 year or lifetime. in some cases, double lifetime.

 

[badb0y]

Won't three of these put the AMD 6990 to shame?

 

[-Slacker-]

Something is wrong with those benchmarks. The PCS gains anywhere from 10% to 20~25% performance increase on just an 8% core bump. That's just doesn't seem possible, especially for radeons.

Also, the gtx 470 is handily faster than the stock 6870 in avp and metro, which goes against almost every review taken in the last several months.

 

[bryanW1995]

they're ok cards.. they just don't have an OK warranty. 2 year warranty is laughable when a lot of other equally priced manufacturers do 3 year or lifetime. in some cases, double lifetime.

lifetime warranty is usually just for the lifetime of the card these days; ie, if your gtx 560 ti breaks in 4 years most manufacturers won't warranty it unless tsmc still hasn't figured out 28nm. You know what, it probably will still be under warranty after all, carry on.

EVGA and XFX have traditionally been the 2 best for warranty service/repairs. However, as we saw with BFG, things can go down in a hurry in this crowded market. Don't get me wrong, I'd definitely go with EVGA/XFX if all else is equal, but often another vendor partner is $10-$20 cheaper. In those cases (as long as it's a reference design) I'll usually go with the cheapest.

 

[nanaki333]

touche bryan.

i really don't have a problem with HIS or powercolor and their low warranty anyway. i usually replace my cards once or twice a year. my current 6970s and the 6970 in my wife's computer is powercolor. they'll be replaced by winter if something new comes out

 

[wahdangun]

lifetime warranty is usually just for the lifetime of the card these days; ie, if your gtx 560 ti breaks in 4 years most manufacturers won't warranty it unless tsmc still hasn't figured out 28nm. You know what, it probably will still be under warranty after all, carry on.

EVGA and XFX have traditionally been the 2 best for warranty service/repairs. However, as we saw with BFG, things can go down in a hurry in this crowded market. Don't get me wrong, I'd definitely go with EVGA/XFX if all else is equal, but often another vendor partner is $10-$20 cheaper. In those cases (as long as it's a reference design) I'll usually go with the cheapest.

yup totally agree,

and I think powercolor have become more like asrock on graphic card, and if the dual crossfire finger is working then its really make it more valuable than GTX 560TI, and if that was possible we just need to wait for 6890 from powercolor for truly quadCF card config on the go lol

 

[zebrax2]

lifetime warranty is usually just for the lifetime of the card these days; ie, if your gtx 560 ti breaks in 4 years most manufacturers won't warranty it unless tsmc still hasn't figured out 28nm. You know what, it probably will still be under warranty after all, carry on.

EVGA and XFX have traditionally been the 2 best for warranty service/repairs. However, as we saw with BFG, things can go down in a hurry in this crowded market. Don't get me wrong, I'd definitely go with EVGA/XFX if all else is equal, but often another vendor partner is $10-$20 cheaper. In those cases (as long as it's a reference design) I'll usually go with the cheapest.

I've read in this forums a few occasions wherein their card was replaced by a card of similar (or slightly higher/lower) performance since their original cards wasn't on stock or made anymore. Have they stopped doing this?

 

[Gloomy]

Is it possible to do tri-SLI with this card as the centerpiece and two reference designs?

 

[Daedalus685]

Is it possible to do tri-SLI with this card as the centerpiece and two reference designs?

In theory it should be fine, the crossfire stubs are parallel to each other. However, you'd need a very long CF bridge to bend in a way as to actually ever plug it in, certainly a standard ribbon would not work.

 

[MentalIlness]

Something is wrong with those benchmarks. The PCS gains anywhere from 10% to 20~25% performance increase on just an 8% core bump. That's just doesn't seem possible, especially for radeons.

Also, the gtx 470 is handily faster than the stock 6870 in avp and metro, which goes against almost every review taken in the last several months.

Performance for the 6xxx cards over the last few driver updates has seen quite a boost in performance. I suspect that is why.

 

[tincart]

Something is wrong with those benchmarks. The PCS gains anywhere from 10% to 20~25% performance increase on just an 8% core bump. That's just doesn't seem possible, especially for radeons.

You're forgetting that there's about 9.5% increase in the memory bandwidth over stock as well. Where are you seeing 25% increases? In most cases they are around 12-13%.

Also, the gtx 470 is handily faster than the stock 6870 in avp and metro, which goes against almost every review taken in the last several months.

I don't understand this sentence. Are you saying that the 470 is faster than 6870 in other reviews and so the TPU reviews go against other results? The 470 does "handily" beat the 6870 in some of the TPU benchmarks but not others and I recall Radeon cards doing particularly well in most recent AVP benchmarks. AVP is one of the titles where AMD has managed to get fairly robust FPS gains across several driver updates. Things can change.

 

[3DVagabond]

Well, that's a pretty large jump in performance, especially in all of the DX11 games compared to the DX9/10 games. Were there different drivers used for the 6870++ than the other 6870? He doesn't specifically state which drivers were used for the ++. I have a suspicion he used the same Beta drivers as he did for the 6990 while the other Radeons were on 11.1.

And he did get the core up to 1020 and memory up to 1210. Not terrible, and the scaling looks good with the one bench he did. At these speeds it would put the card ahead of the 5870, and it looks like most especially in DX11 titles.

Drivers might very well be different. w1zzard isn't known for updating old tests because drivers have changed. I doubt he would roll back to any previous drivers instead of just leaving the betas in if he wasn't having any issues either.

edit: I just wish they'd quit doing the cheesy, "make the HS cowling look like a car". LOL

 

[badb0y]

Definitely looks like he used the new "Mjolnir I" beta drivers.

 

[-Slacker-]

Aw, so they only test out the new card and then compare the results with older ones from older cards... I just assumed they would test all the cards in their inventory on every new review, in order to keep the comparisons fair...

You're forgetting that there's about 9.5% increase in the memory bandwidth over stock as well. Where are you seeing 25% increases? In most cases they are around 12-13%.

1920x1200, normal 6870 vs pcs++:

avp---26.2 vs 34.2 ---30.5%

bfbc2--- 50.6 vs 59 --- 16.6%

battleforge---42.3 vs 51.8--- 22.4%

mw4 ---116 vs 123.4 ---6.3%

CoJ2---99.4 vs 108.3 ---9%

civ5---38.5 vs 43.5 --- 13%

crysis ---30.2 vs 33.8 ---12%

f12010 --- 53.3 vs 58.8 --- 10%

FC2 ---75.8 vs 88.5 ---17%

hawx --- 78 vs 85.3 ---9.3%

metro ---14.6 vs 18.5 --- 27%

ut3 --- 148 vs 155 --- 5%

wow ---67 vs 74 --- 10%


I didn't forget that there was a memory bump too, although, according to my previews post, I might as well have. Apologies for not making myself clearer. Anyway, an overclock on the core usually implies a similar % overclock on the mem - if you want to grasp at straws and tackle that then be my guest, I'm not going to argue. In fact, if you want to grasp at straws, then go ahead and tackle this whole post; I wasted quite a bit of time calculating what should have been obvious, and I'm going to leave it at that.

 

[3DVagabond]

Aw, so they only test out the new card and then compare the results with older ones from older cards... I just assumed they would test all the cards in their inventory on every new review, in order to keep the comparisons fair...

He'll usually update the scores 3 or 4 times a year. With 12 (+Hotfixes, etc.) new AMD drivers a year, and whatever nVidia does (Not as many typically, but they've released a whole slew since the 590), and the number of benches he runs, it would be extremely time consuming. Especially since there's typically not any meaningful differences. (I know some people will argue this, but in reality differences are mostly in the realm of "margin of error".) Last year though when he updated his tests, at one point, he ended up using the driver release that broke crossfire in a lot of games (11.4, IIRC? It was the ones that [H] revealed as broken. I don't recall which they were for sure.) and ran with those for a few months. Caused more than a few arguments over 5970 performance.