Radeon 4830 512MB to Radeon 6870 1GB Review! Now With OC Benchmarks!!

[TurtleCrusher]

The 6870!

I am replacing my 4830 today for what I hope is a much more powerful video card.

Here is the box it came in.

The card itself.

The card sideways.

Size comparison to the 4830 and a dollar bill.

Bundled accessories - 2x Molex to PCI-E 6-pin, HDMI Cable (wow!), Mini Display port to Full Size Display port DVI to VGA Adapter, and a Cross fire Connector.

Test Setup:


CPU: Intel Q8200 @ 2.8Ghz - 400x7 - 1.09v
Ram: Patriot 4GB (2x2) PC6400 DDR2 5-5-5-12 1.9V
Motherboard: Gigabyte EP45-DS3L ATX
Hard Disk Drives: Windows Drive: Western Digital 640GB Black AALS
Gaming, Programs and Data Drive: Western Digital 1TB Blue EALS
Sound: Audigy 1 Gamer PCI
Power Supply: Antec True Power Trio 550w - 3x 18amp 12v rails
Case: Antec P180 ATX case

Drivers Used. 4830: 10.10, 6870: 11.2.

All Benchmarks were performed at 1680x1050. Each test was done three times besides the Battlefield Bad Company 2 Bench. The lowest minimum and highest maximum bench was thrown out of each.

Street Fighter 4 - Built In Benchmark Tool

Ran at 8xAA (24x Edge Detection, Super Sampled Via CCC), 16xAF, Posterization Detail Setting, Completely Maxed Out.

Radeon 6870 Stock (900/1050)
Minimum: 23
Maximum: 63
Average: 51.202

Radeon 6870 Overclocked (1003/1050)
Minimum: 27
Maximum: 65
Average: 55.797

Grand Theft Auto 4 - Built In Benchmark Tool

MLAA is Enabled via CCC

Radeon 6870 Stock
Minimum: 18
Maximum: 54
Average: 35.119

Radeon 6870 Overclocked
Minimum: 19
Maximum: 54
Average: 36.575


Left 4 Dead 2 Single Player - Plantation Level finished.

Ran at all highest possible settings. 8xAA 16xAF Motion Blur.

Radeon 4830
Minimum: 14
Maximum: 129
Average: 54.6

Radeon 6870
Minimum: 33
Maximum: 127
Average: 75.4

Radeon 6870 Overclocked
Minimum: 41
Maximum: 125
Average: 71.579

Benchmarks and Review Carries to Post #3!

 

[Termie]

Nice comparison. Didn't need the new card for Source, but you sure did for BC2. Awesome improvement! Now turn it up to 4xAA or even 8xAA and make sure you have HBAO on.

You're in DX11 for Dirt2 on the new card, right? You should be getting some improvement in graphics over the 4830. I think you might actually be CPU-limited in Dirt2. Here are my results on an oc'd HD5850, which is pretty much identical to your HD6870:

1680x1050/4xAA
Min Max Avg
61 87 74.879
1680x1050/0xAA
Min Max Avg
73 91 81.133

P.S. I actually installed that very same HD4830 in my original homespun rig that I gave to my bro. A very nice card for the price.

 

[TurtleCrusher]

Counterstrike: Source - Office Level 9 Bots and Self, Three Rounds.

Ran at all highest possible settings. 8xAA 16xAF HDR Multicore Enabled.

Radeon 4830
Minimum: 66
Maximum: 176
Average: 118.02

Radeon 6870
Minimum: 72
Maximum: 178
Average: 122.81

Radeon 6870 Overclocked
Minimum: 71
Maximum: 180
Average: 120.628

CounterStrike: Source Video Stress Test

Ran at 8xAA (24x Edge Detection, Super Sampled Via CCC), 16xAF, Completely Maxed Out.

Radeon 6870
Minimum: 44
Maximum: 97
Average: 63.255

Radeon 6870 Overclocked
Minimum:47
Maximum: 104
Average: 67.546

Battlefield Bad Company 2 - Single Player Heavy Metal (Mission 8) Four Checkpoints Benchmark - Atacama Desert 30 Minute Multiplayer Run With 31 Players + Self.

SINGLE PLAYER BENCHMARKS:

Ran at all high, 4xAA 16xAF

6870 Stock 900/1050
Minimum: 33
Maximum: 85
Average: 59.044

6870 Overclocked
Minimum: 38
Maximum: 107
Average: 65.814

MULTIPLAYER BENCHMARKS:

Ran at all high + 2xAA 16xAF.

Radeon 4830
Minimum: 9
Maximum: 50
Average: 32.37

Radeon 6870
Minimum: 43
Maximum: 124
Average: 69.8

Radeon 6870 Overclocked
Minimum: 37
Maximum: 169
Average: 54.909

DiRT 2 - Bettersea 2 Laps Rallycross With Full Lineup

Ran at maximum allowed DX10 settings (High/Ultra), 4xAA.

Radeon 4830 DX 10
Minimum: 26
Maximum: 54
Average: 43.183

Radeon 6870 DX 11
Minimum: 40
Maximum: 79
Average: 59.386

Radeon 6870 Overclocked DX 11
Minimum: 47
Maximum: 86
Average: 64.409

Temperature/Heat: The 4830 when idle downclocks to 160/250. This keeps it cool usually around 35C in my case with the fan off. The 6870 idles at 100/300 and keeps around 32C by default fan speeds.

At load (BF:BC2) the 4830 hits 72C with the fan at 100%. The 6870 hits 58C with the fan at 65%.

These were conducted in an Antec P180 case with three 120mm fans. The Fan on the 6870 is quite audible in my case above 60%, and is very loud at 100%. 50% fan speed is adequate fort everything I've done so far.

Conclusion for the 6870: The games that I was able to benchmark all saw an increase in FPS. game like Counterstrike Source only saw an increase in averages, telling me that I am CPU bottlenecked in that title. Games like DiRT 2 and Battlefield Bad Company 2 saw a dramatic increase in FPS and a game play experience. Left 4 Dead 2 saw a massive increase minimum fps and that resulted in much easier game play during hordes of zombies. Otherwise the game seemed indistinguishable compared with the 4830.

While Overclocking the 6870 the games that I observed CPU bottlenecks on showed no benefit. Be it an oddity or not, performance was usually a fps or 2 lower on average. I do consider this well within the margin of error. I did notice Marked benefit in DiRT2, as minimum and average frames made this game absolutely playable at my resolution.

Battlefield Bad Company 2 is a bench that I don't have the time to do three tests on each, so my overclocked result wasn't exactly in line with the stock benchmarks. Most likely this game can vary just by how you play the game. If you sit out and snipe all game long I can see the fps remain consistently high. As I played in mostly close quarters combat in my overclcoked bench, fps was lower aroudn the board as I was near action the entire time. It was still a very playable experience. The single player part of the game produced very constant results on benchmarking. All scores were so close that throwing out the highest and lowest didn't change a thing. The single player supposedly hits 33-38fps at a minimum at times, which i never noticed. 60fps average at least at 4xAA, which is very nice. I now notice I am not CPU limited in single player as my overclock of 11% gains of 11.1%, which is exactly the percentage overclock on my GPU. Very surprising the gains were linear with the clockspeed increase. By this finding I believe it is safe to assume Battlefield Bad Company 2 is core speed limited rather than bandwidth limited while using the Radeon 6870.

Overall, I like it the increases, even with games that are mostly CPU bound. There will be more stock vs overclock benchmarks to come.

 

[MrK6]

Great review, nice organization! Glad your new card impressed. Try cranking IQ settings in CPU bound games :thumbsup:.

 

[TurtleCrusher]

@Termie Thanks. The minimum increase on BC2 was amazing. It felt soooooo much better to play.

You're right, DiRT2 was running DX11 instead of 10. I'll keep them at the same settings just to show the difference in parts, as you're not going to run DX10 on 58xx/68/9xx series cards. I hope I'm not CPU limited in DiRT 2, as it's a multithreaded game, but if it is the performance is more than satisfactory.

The 4830 is still a decent card. Mid level ATI cards (x5/6xx series) cards have almost caught up to it. It could run all games on high settings, but some lagged down. There isn't a game is CANT play yet. I'll miss it, maybe!

@MrK6 That's the plan! 8xAA 16xAF in almost all games without a hitch. I never thought I'd buy a card that would make my trusty Core 2 Quad the limiting factor in multi-threaded games, but that's life.

 

[drizek]

You might be able to run Battlefied with 16x AA. I don't know what my framerate is, but with vsync on I never notice stuttering.

I upgraded from that same 4830 too. Mine was incredibly loud though. I never hear anyone else mention it, but I had to hook mine up to a fan controller

 

[MrK6]

That's the plan! 8xAA 16xAF in almost all games without a hitch. I never thought I'd buy a card that would make my trusty Core 2 Quad the limiting factor in multi-threaded games, but that's life.

If you ever wanted to go wild with the voltage, I'm sure you could take that quad further. Unfortunately, the Q8x00 chips weren't very efficient overclockers, although great bang-for-your-buck :thumbsup:.

 

[TurtleCrusher]

If you ever wanted to go wild with the voltage, I'm sure you could take that quad further. Unfortunately, the Q8x00 chips weren't very efficient overclockers, although great bang-for-your-buck :thumbsup:.

Its my board that's holding me back. I pushed it to 3.3ghz once and it seemed stable 1.3v, but my motherboard isn't up to snuff. It's a P45, but has the much smaller heatsinks on it. My caps near the CPU were going brown after an hour or so. 3ghz is as far as I dare now. 2.8 seems fast enough.

 

[TurtleCrusher]

Nice comparison. Didn't need the new card for Source, but you sure did for BC2. Awesome improvement! Now turn it up to 4xAA or even 8xAA and make sure you have HBAO on.

You're in DX11 for Dirt2 on the new card, right? You should be getting some improvement in graphics over the 4830. I think you might actually be CPU-limited in Dirt2. Here are my results on an oc'd HD5850, which is pretty much identical to your HD6870:

1680x1050/4xAA
Min Max Avg
61 87 74.879
1680x1050/0xAA
Min Max Avg
73 91 81.133

P.S. I actually installed that very same HD4830 in my original homespun rig that I gave to my bro. A very nice card for the price.

Looks like I'm not too CPU limited in DiRT2 as I thought. Bettersea (the track I benched) might just be the most demanding track there is. For instance, I just did the first Rally in Morocco with ALL settings on the absolute highest and it was:

1680x1050 8xAA (16xAF in CCC)
Min Max Average
56 82 70.714

All other tracks perform nearly the exact FPS so far!

 

[Termie]

Looks like I'm not too CPU limited in DiRT2 as I thought. Bettersea (the track I benched) might just be the most demanding track there is. For instance, I just did the first Rally in Morocco with ALL settings on the absolute highest and it was:

1680x1050 8xAA (16xAF in CCC)
Min Max Average
56 82 70.714

All other tracks perform nearly the exact FPS so far!

Good bench, looks like you're right. Just picked a tough level to start with, it seems.

 

[nyker96]

I can see a 50%-100% jump for you, excellent upgrade. try the game Metro 2033, that one really slows my gtx 460 down, I wonder how your hd6870 will fare.

 

[TurtleCrusher]

You might be able to run Battlefied with 16x AA. I don't know what my framerate is, but with vsync on I never notice stuttering.

I upgraded from that same 4830 too. Mine was incredibly loud though. I never hear anyone else mention it, but I had to hook mine up to a fan controller

I ran it today at 8xAA, and it was still really smooth. FPS typically around 57-63fps with drop to 40 in serious action... Which mirrors my 2xAA scores. I'll be benching 8xAA on BC2 soon.

My P180 is a really loud case as I keep all fans on high. The 4830 was the last thing I had to worry about on sound. Didn't ever hear it really.

I can see a 50%-100% jump for you, excellent upgrade. try the game Metro 2033, that one really slows my gtx 460 down, I wonder how your hd6870 will fare.

Yeah it is great so far. Once I get some more $$ I'll try and get Metro 2033 if reviews are positive. Unlike many PC gamers out there, I only play what I've bought!

 

[MrK6]

Its my board that's holding me back. I pushed it to 3.3ghz once and it seemed stable 1.3v, but my motherboard isn't up to snuff. It's a P45, but has the much smaller heatsinks on it. My caps near the CPU were going brown after an hour or so. 3ghz is as far as I dare now. 2.8 seems fast enough.

You might have to put some heatsinks on the mosfets, but my EP45-DS3P was a great clocker. Either way, do what you're comfortable with and DON'T fry your gear :awe:.

Yeah it is great so far. Once I get some more $$ I'll try and get Metro 2033 if reviews are positive. Unlike many PC gamers out there, I only play what I've bought!

It goes on sale for $10-12.50 from time to time - keep your eye on slickdeals or the Hot Deals section.

 

[drizek]

I got in on Metro for $7.50 I think. Buy it from somewhere that gives you a cd key and then activate it on Steam.

It's not bad on my GTX 460 at all. Maybe I just have more tolerance for low framerates.

Does the P180 use Tricool fans? I can't imagine running mine on High. I keep it at low and push it up to medium when gaming.

 

[TurtleCrusher]

I got in on Metro for $7.50 I think. Buy it from somewhere that gives you a cd key and then activate it on Steam.

It's not bad on my GTX 460 at all. Maybe I just have more tolerance for low framerates.

Does the P180 use Tricool fans? I can't imagine running mine on High. I keep it at low and push it up to medium when gaming.

Yes they are the tri-cool fans. I dont mind the sound too much. My desk/computer area is in a really cold, secluded walk-in closeted area so my temps stay really low as it is. I'm just looking to make my hardware last as long as possible.

Edit: Wow the sound is so much lower with the fans on low. This is better. I still cant hear my GPU fan until 50%.

 

[drizek]

You really don't need them running that high. Low works fine for me, I can run the OCCT power supply stress test on it and it doesn't even overheat. I set it to medium just to be safe, but I can get away with low all the time.

Right now my system is essentially silent and the CPU is running at 29C and the video card is at 24C. It's kinda cold in here today, but even if they were running at 35/30, that is still very low.

Try undervolting your CPUs idle speed. You should be able to get it down quite low.

 

[nyker96]

I ran it today at 8xAA, and it was still really smooth. FPS typically around 57-63fps with drop to 40 in serious action... Which mirrors my 2xAA scores. I'll be benching 8xAA on BC2 soon.

My P180 is a really loud case as I keep all fans on high. The 4830 was the last thing I had to worry about on sound. Didn't ever hear it really.



Yeah it is great so far. Once I get some more $$ I'll try and get Metro 2033 if reviews are positive. Unlike many PC gamers out there, I only play what I've bought!

I think I bought metro 2033 for under $15 on steam, to be honest I bought about $65 when steam had sale and so far only ran metro couple of times and most played tf2, I keep buying games but never play most of them. I hope you get a chance to grab a copy to test it out, it's got good reviews so not just a benchmark game.

anyways, I have a P182 in another rig and do find the high setting on fans to be way too loud, personally I think all you need is medium setting, you are just using one video card it should be enough cooling add another 120mm in the front behind HD cage if you want more ventilation.

As per an earlier poster, I'd recommend you oc that q8200, probably pass 3,2Ghz should give your games a nice boost.

 

[nemesismk2]

thanks for the great review. interesting read

 

[TurtleCrusher]

thanks for the great review. interesting read

Thanks!

Updated OP with overclock results. Notes added in conclusion. Very soon I'll be doing single 6870 scores in other games, then adding stock vs overclock scores after that.

 

[happy medium]

Thanks!

Updated OP with overclock results. Notes added in conclusion. Very soon I'll be doing single 6870 scores in other games, then adding stock vs overclock scores after that.

Nice info and thanks.:thumbsup:

I might have missed it, what is the 6870 overclocked at?

edit: 9%? or 1000 core found it? You don't seem to be that cpu limited. Mabe just a little.

 

[TurtleCrusher]

Nice info and thanks.:thumbsup:

I might have missed it, what is the 6870 overclocked at?

1003/1050 vs 900/1050. I keep the memory stock. I changed the pic from GPUreview to include the 6870 as well.

Why not OC the ram? I've never had success OC memory on ANY card, and one died, so no interest in that

 

[Termie]

1003/1050 vs 900/1050. I keep the memory stock. I changed the pic from GPUreview to include the 6870 as well.

Why not OC the ram? I've never had success OC memory on ANY card, and one died, so no interest in that

Are you sure your settings were the same in BC2 between the stock and overclocked tests? I've never seen that much variation on a single map at the same clock speed, let along a clock speed that should have given you higher frame rates. One thing you could consider doing is just benching part of the single-player game. Choose a scene, bench for 2 or 3 minutes, then change your settings, and bench the very same scene again. This isn't accurate for testing CPU limitations, but it will show your overclock pretty well.

I just completed a round of overclock testing on BC2 with my HD5850. You can see my results here:

http://forums.anandtech.com/showthread.php?t=2020118&page=282

 

[erikistired]

i am getting ready to upgrade my 4850, and i believe i just found my new card. thanks for the info.

 

[fuzzymath10]

I upgraded my slim platform from a Q8200 to a SB i5 keeping the video card the same. I also upgraded from a 1600x900 LCD to a 1920x1080 LCD, and removed the overclock on my 5750 (from 865/1265 back to 700/1150). Even with the resolution increase and the switch back to stock GPU clocks, my frame rates were much better. Before I was hitting lag in intense scenarios (sub 30, even 20fps). Now I don't even notice. SC2 is horribly limited by the CPU, and I think only i5 or i7 CPUs are ideal.

I have the Q8200 paired with a passively cooled 4830, and I think SC2 was hardly worse than with a 5750.

 

[TurtleCrusher]

Are you sure your settings were the same in BC2 between the stock and overclocked tests? I've never seen that much variation on a single map at the same clock speed, let along a clock speed that should have given you higher frame rates. One thing you could consider doing is just benching part of the single-player game. Choose a scene, bench for 2 or 3 minutes, then change your settings, and bench the very same scene again. This isn't accurate for testing CPU limitations, but it will show your overclock pretty well.

I just completed a round of overclock testing on BC2 with my HD5850. You can see my results here:

http://forums.anandtech.com/showthread.php?t=2020118&page=282

Absolutely the same. I just played 5 mins of close combat and then 5 mins of sniping and the performance fell in line with the data in the OP. I wont be able to bench the 4830 against the 6870 for now, but I will be able to bench the 6870 and the 6870 OCed.

How do you bench single player in BC2? Once I get past a part I cant go back. It autosaves and i'd have to redo the entire campaign!