5850 just as fast as a 5870?

GodisanAtheist

Diamond Member
Nov 16, 2006
7,496
8,279
136
http://www.xbitlabs.com/articles/video/display/radeon-hd5850_10.html#sect0

In this review, the 5850 was clocked to the same speed as a 5870 (850/4800) and when the smoke cleared, it was a whopping 2% slower than its older brother (they also test a 1010/4960 OC, but that's besides the point). The 850/4800 clocks were achieved using stock voltage no less.

The 5870 has 160 more SP's (+10%), 8 more TMUs (+10%) and is clocked at 850Mhz Core (+10%) & 1200Mhz Mem (+17%) than the 5850, yet only manages to be +2% faster with the same clock speeds. This indicates that those extra SPs and TMUs are more dead weight than actually adding to the performance of the 5870 card.

I would really like to see this review's methodology replicated in another review and tested for repeatability, but for the sake of making inflammatory statements, I will jump to conclusions and say that 5870 is the 2900xtx revisited.

This only further fuels my desire for a 1280 SP 5830. How far from the 5870 would it lie with identical clocks?

5850 is by and far the card to get.

Discuss.
 
Last edited:

BFG10K

Lifer
Aug 14, 2000
22,709
2,998
126
The 5870 has 160 more SP's (+10%), 8 more TMUs (+10%) and is clocked at 850Mhz (+10%) than the 5850, yet only manages to be +2% faster with the same clock speeds. This indicates that those extra SPs and TMUs are more dead weight than actually adding to the performance of the 5870 card.
Or the drivers aren’t using the extra units properly, or aren’t using them at all. This could be intentional, or a genuine driver issue.

I’ve been saying almost right from the start that I believe the entire 5xxx series is underperforming and I expect future drivers to correct this, probably when Fermi arrives. Either that or there’s a severe hardware bottleneck that’s holding back performance.
 

Painman

Diamond Member
Feb 27, 2000
3,728
29
86
I doubt it's a memory thing. Look at the go-for-broke OC numbers for the 5850: it trounces the stock 5870, but the mem frequency is only 160 MHz (effective) higher than 5870's. A 3.3% increase, but the mean performance increase in the massively OC'd scenario is way higher than that.

The GPU OC is more along the lines of 18-19% of 5870 freq, but of course, the 5850 has fewer shader processors. I believe the Wolfenstein #s show that the mem bandwidth is adequate. That's Doom 3 engine. If there was a mem bandwidth bottleneck, I don't believe you'd see that kind of increase. As far as that game is concerned, it seems to be responding quite well to the overall OC. As one would expect, since it's an older, and presumably GPU-limited game.
 

stag3

Diamond Member
Feb 7, 2005
3,623
0
76
but then you can just o/c the 5870 also no? to gain that advantage back
i wanted a 5870 but it wouldn't have fit my case
 

smackababy

Lifer
Oct 30, 2008
27,024
79
86
The problem is nobody looks at the min. frame rates. They will be significantly lower on the 5850. Now, dipping low on those makes a huge difference in how smooth a game plays. And you can always just O/C a 5870 and gain a bigger margin.

With that said, I bought a 5850 anyway.
 

T2k

Golden Member
Feb 24, 2004
1,665
5
81
Or the drivers aren’t using the extra units properly, or aren’t using them at all. This could be intentional, or a genuine driver issue.

I’ve been saying almost right from the start that I believe the entire 5xxx series is underperforming and I expect future drivers to correct this, probably when Fermi arrives. Either that or there’s a severe hardware bottleneck that’s holding back performance.

I agree, my bet is also on the drivers... my other theory is that once devs master DX11 better (simplified codepath etc) it will jump ahead there as well.
 
Last edited:

GodisanAtheist

Diamond Member
Nov 16, 2006
7,496
8,279
136
The problem is nobody looks at the min. frame rates. They will be significantly lower on the 5850. Now, dipping low on those makes a huge difference in how smooth a game plays. And you can always just O/C a 5870 and gain a bigger margin.

With that said, I bought a 5850 anyway.

-I want everyone to read this post. Go ahead, read it. This, ladies and gentlemen, is a forum user who decided he was better than the source, decided not to verify facts, decided to jump to his own conclusions. This is misinformation.

If you're the type of individual that wants, nay, NEEDS more out of life, you haven't taken the OP at face value and looked at the individual benchmarks in the source and SEEN WITH YOUR OWN EYES that xbit includes min. framerates. If you are one of these people then you KNOW that the 5850's minimum frames match those of the 5870 at normalized clocks.
 
Last edited:

shangshang

Senior member
May 17, 2008
830
0
0
-I want everyone to read this post. Go ahead, read it. This, ladies and gentlemen, is a forum user who decided he was better than the source, decided not to verify facts, decided to jump to his own conclusions. This is misinformation.

If you're the type of individual that wants, nay, NEEDS more out of life, you haven't taken the OP at face value and looked at the individual benchmarks in the source and SEEN WITH YOUR OWN EYES that xbit includes min. framerates. If you are one of these people then you KNOW that the 5850's minimum frames match those of the 5870 at normalized clocks.

No need to be ultra exuberrant in calling his post out. Ugh.. somehow I sense that you are having something personal against him and just waiting for a chance to lash out?
 

GodisanAtheist

Diamond Member
Nov 16, 2006
7,496
8,279
136
No need to be ultra exuberrant in calling his post out. Ugh.. somehow I sense that you are having something personal against him and just waiting for a chance to lash out?

-No need, more for my own enjoyment. I have a bit of a pet peeve concerning people not reading a source then spouting anecdotal evidence that is directly addressed in the source...

I only have 23 posts on this forum, hardly enough to form any sort of grudge against anyone.
 
Last edited:

bunnyfubbles

Lifer
Sep 3, 2001
12,248
3
0
They must have a super duper 5850 to achieve 850MHz at stock volts. That or its not stable and they just didn't care to figure that out. Now my own 5850 will easily do 850MHz with a voltage bump to match the 5870's.

There are some pitfalls to the 5850 being clearly the better option, the biggest is its stock reference cooler which has only two heatpipes. Its good enough for stock operation at reasonable noise levels, but just about any overclocking/overvolting will require some hell-scream level of fan noise.

The second biggest pitfall is properly cooling the VRMs. When I got my 5850, I was quick to swap out the terribad stock cooler for an Accelero S1 which I planned to strap a 120mm fan to. That worked out great at first glance, as my GPU temps are about 45C under load. The problem is the little dink tack-on VRM heatsinks are nowhere near enough to handle the amount of heat coming off those pesky VRMs. What I ultimately had to end up doing was remove the back plate from the stock cooler which acted as a giant heat spreader for the VRMs and RAM and use that in tandem with the S1.

That being said, I'm not so sure its worth the extra money and hassle at the current price points. When it was $260 vs. $380 it was a no brainer for me, but @ $310+ vs. $410+ I'm not so sure.
 

DonInKansas

Senior member
Feb 25, 2008
607
0
76
I'm on the driver bandwagon. Once fermi drops I bet ATi has drivers in the tank that will make these cards smoke more than they already do. No reason to shoot all your wad early with no competition.
 

AzN

Banned
Nov 26, 2001
4,112
2
0
Memory bandwidth?

It really is bandwidth though considering the bottleneck of RV870 design. When you think about the core clocks of the 5870 it should be faster than 4870x2 all around. But it's not. They perform about same or 5870 a slightly slower.

I don't know why BFG thinks it's drivers when Radeon drivers are mature. There's no evidence of magic drivers out of hat. If driver optimization happen it will also be applied to RV770.
 

bunnyfubbles

Lifer
Sep 3, 2001
12,248
3
0
No need to be ultra exuberrant in calling his post out. Ugh.. somehow I sense that you are having something personal against him and just waiting for a chance to lash out?

Maybe he was a bit harsh but I think he was perfectly justified in doing it. People talking out of their ass and spreading misinformation is not good for a forum. I was about to call him on it as well but was beaten to the punch.
 

T2k

Golden Member
Feb 24, 2004
1,665
5
81
It really is bandwidth though considering the bottleneck of RV870 design. When you think about the core clocks of the 5870 it should be faster than 4870x2 all around. But it's not. They perform about same or 5870 a slightly slower.

I don't know why BFG thinks it's drivers when Radeon drivers are mature. There's no evidence of magic drivers out of hat. If driver optimization happen it will also be applied to RV770.


Ehhh? Perhaps because we have seen this several times before, last time as late as August?


This month we are seeing a massive performance increase with a whole host of games as compared to the ATI Catalyst 9.7 driver. Detailed release notes are available for most of the game optimizations; here are the highlights:


  • Battleforge DirectX 10/DirectX 10.1 performance improves of up to 50% with the largest gains in configurations using ATI CrossFireXTM technology.
  • Company of Heroes DirectX 10 performance improves of up to 77%.
  • Crysis DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 10% and quad mode performance improves of up to 34%.
  • Crysis Warhead DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 7% and quad mode performance improves of up to 69%.
  • Far Cry 2 DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 50% and quad mode performance improves of up to 88%.
  • Tom Clancy’s H.A.W.X. DirectX 10/DirectX 10.1 performance ofATI CrossFireX technology in dual mode improves of up to 40% and with quad mode performance improving of up to 60%.
  • UnigineTropics OpenGL performance improvements of up to 20%.
  • UnigineTropics DirectX 10 performance ofATI CrossFireX technology in quad mode improvements of up to 20%.
  • World in Conflict DirectX 10 performance improvements of up to by 10%."
 

dguy6789

Diamond Member
Dec 9, 2002
8,558
3
76
It really is bandwidth though considering the bottleneck of RV870 design. When you think about the core clocks of the 5870 it should be faster than 4870x2 all around. But it's not. They perform about same or 5870 a slightly slower.

I don't know why BFG thinks it's drivers when Radeon drivers are mature. There's no evidence of magic drivers out of hat. If driver optimization happen it will also be applied to RV770.

It has been tested and proven already that the performance limitation is not memory bandwidth. Overclocking the memory by huge amounts nets little performance gain on the 5870. Increasing the GPU clock increases performance in a normal manner.

The only reasonable explanation is drivers.
 

T2k

Golden Member
Feb 24, 2004
1,665
5
81
Just looked up the news about my last fav Catalyst upgrade, exactly from a year ago: http://techreport.com/discussions.x/16033

Catalyst 8.12 drivers bring CAL runtime, speedups
by Cyril Kowaliski — 10:47 AM on December 10, 2008

AMD's graphics driver developers haven't been twiddling their thumbs. They've just released a new batch of Catalyst drivers that brings not only full ATI Stream support and a new video transcoder, but also a host of (claimed) performance boosts in games.
(...)
Here's what those folks can expect, in AMD's words:
Call of Duty: World at War DX9 – performance gains of up to 21% for Single and Crossfire mode. Performance gains were noticed on all Radeon HD48xx.
Crysis DX10 – performance gains of up to 25% for Single and Crossfire mode. Performance gains were noticed on all Radeon HD4xxx.
Crysis Warhead DX10 – performance gains of up to 13% for Single card mode and up to 16% for Crossfire mode. Performance gains were noticed on all Radeon HD4xxx.
Devil May Cry 4 DX10– performance gains of up to 6% for Single and Crossfire mode. Performance gains were noticed on all Radeon HD4xxx and HD38xx.
Fallout 3 – performance gains of up to 15% for Single card mode.
Far Cry 2 DX10 – performance gains of up to 10% for Single card mode and up to 57% in Crossfire mode. Performance gains were noticed on all Radeon HD4xxx and HD38xx.
FEAR DX9 – performance gains of up to 6% for Single and Crossfire mode. Performance gains were noticed on Radeon HD4870X2 and HD4870.
Hellgate: London DX10 – performance gains of up to 6% for Single card mode and up to 10% for Crossfire mode. Performance gains were noticed on all Radeon HD4xxx.
Left 4 Dead DX9– performance gains of up to 10% for Single card mode and up to 5% for Crossfire mode. Performance gains were noticed on all Radeon HD4870 series.
Lost Planet Colonies DX10 – performance gains of up to 10% for Single and Crossfire mode. Performance gains were noticed on all Radeon HD4xxx and HD38xx.
Prey OGL – performance gains of up to 15% for Crossfire mode. Performance gains were noticed on Radeon HD4870 1GB products, HD46xx, HD45xx.
S.T.A.L.K.E.R.: Clear Sky – performance gains of up to 10% for Crossfire mode. Performance gains were noticed on all Radeon HD4xxx and HD38xx.
Unreal Tournament 3 DX9 – performance gains of up to 18% for Single card mode and up to 15% in Crossfire mode, especially in cases where AA is enabled. Performance gains were noticed on all Radeon HD4xxx and HD38xx.
This was the driver after I decided to invest into my 4850 X2 2GB... if ATI can pull out a similar ~15-20% average increase in games I play (FPS, RTS) by March (Fermi's release) then I will replace my 5870 with a 5970 2GB monster. :)
 
Last edited:

AzN

Banned
Nov 26, 2001
4,112
2
0
Ehhh? Perhaps because we have seen this several times before, last time as late as August?


This month we are seeing a massive performance increase with a whole host of games as compared to the ATI Catalyst 9.7 driver. Detailed release notes are available for most of the game optimizations; here are the highlights:


  • Battleforge DirectX 10/DirectX 10.1 performance improves of up to 50% with the largest gains in configurations using ATI CrossFireXTM technology.
  • Company of Heroes DirectX 10 performance improves of up to 77%.
  • Crysis DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 10% and quad mode performance improves of up to 34%.
  • Crysis Warhead DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 7% and quad mode performance improves of up to 69%.
  • Far Cry 2 DirectX 10 performance ofATI CrossFireX technology in dual mode improves of up to 50% and quad mode performance improves of up to 88%.
  • Tom Clancy’s H.A.W.X. DirectX 10/DirectX 10.1 performance ofATI CrossFireX technology in dual mode improves of up to 40% and with quad mode performance improving of up to 60%.
  • UnigineTropics OpenGL performance improvements of up to 20%.
  • UnigineTropics DirectX 10 performance ofATI CrossFireX technology in quad mode improvements of up to 20%.
  • World in Conflict DirectX 10 performance improvements of up to by 10%."

Did those drivers improvements only apply to 5870 or not?
 

MrK6

Diamond Member
Aug 9, 2004
4,458
4
81
Pretty much every driver release since the Cat 9.9's has netted me more performance on my 5870. I'm going to go with drivers as well.
 

AzN

Banned
Nov 26, 2001
4,112
2
0
It has been tested and proven already that the performance limitation is not memory bandwidth. Overclocking the memory by huge amounts nets little performance gain on the 5870. Increasing the GPU clock increases performance in a normal manner.

The only reasonable explanation is drivers.

With all video cards bandwidth plays the equalizer to the core. Of course bandwidth is the limitation of the 5870 considering all the core doesn't have the breathing room it used to have with rv770. If it didn't it would be faster than 4870x2 by 15% or more in all cases but it's not.

The link above by OP only strengthen that concept as 5850 with 5870 clocks is only 2% slower than 5870.

What you are proposing is that 5850 have better drivers than 5870. We all know this isn't true.
 

GodisanAtheist

Diamond Member
Nov 16, 2006
7,496
8,279
136
:eek: In August....?

I'm pretty sure he means that the driver improvements weren't limited to a 4870, but applied across the 4xxx series line. In such a case, this is hardly an argument against the assertion made in the OP, as the 5850's performance will benefit just as well as the 5870's with further driver improvements.