Overclock.net › Forums › Industry News › Hardware News › [Nordic] RV770 has 800 SPs, breaks 1.2TFLOPS, final clocks
New Posts  All Forums:Forum Nav:

[Nordic] RV770 has 800 SPs, breaks 1.2TFLOPS, final clocks - Page 11

post #101 of 108
*sigh*

Now I have another dilema. G92 8800GTS or 4850. Hopefully some sort of comparison benchmarks with both cards at stock settings and overclocked will show up before my birthday in July.
Original Nutta!
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel i3 530 Gigabyte H55M-USB3 Sapphire 6850 Toxic Edition 2x2GB Corsair Dominator 
Hard DriveOptical DriveOSMonitor
60GB Muskin Callisto Deluxe, 640GB WD Caviar Black LG GH22NS50 22X DVD Writer Windows 7 Ultimate x64 LG W2242T 
KeyboardPowerCaseMouse
Razer Arctosa SilverStone Strider 750w Gold Corsair Graphite White 600T Cooler Master Storm Sentinel Advanced 
Mouse Pad
Thermaltake Dasher 
  hide details  
Reply
Original Nutta!
(13 items)
 
  
CPUMotherboardGraphicsRAM
Intel i3 530 Gigabyte H55M-USB3 Sapphire 6850 Toxic Edition 2x2GB Corsair Dominator 
Hard DriveOptical DriveOSMonitor
60GB Muskin Callisto Deluxe, 640GB WD Caviar Black LG GH22NS50 22X DVD Writer Windows 7 Ultimate x64 LG W2242T 
KeyboardPowerCaseMouse
Razer Arctosa SilverStone Strider 750w Gold Corsair Graphite White 600T Cooler Master Storm Sentinel Advanced 
Mouse Pad
Thermaltake Dasher 
  hide details  
Reply
post #102 of 108
Quote:
Originally Posted by RobRedbeard View Post
Please show me clearly what numbers made you think otherwise, perhaps I'm wrong, but from my experiences and from that web site you provided, it seems to me that you're saying the sky is yellow.
The article I linked had almost 30 more pages of benchmark results and you only looked at the first page that was linked, that's the problem

To be specific as to what I was looking at on that first page, I was looking at the numbers when both cards were subject to the most AA or AF. If the Radeons suffer from such a massive performance hit when you turn AA on, you would think they would be at their absolute worst performance (I'm comparing to Nvidia here) when AA is maxed out. If you look at the first AA benchmark, Anno 1701, at 24xAA the HD3870 retains ~47% of its performance with 1xAA. If you compare this to the maximum 8800GTX and 8800GT performance penalty, 38% and 40.5% respectively, you find that Nvidia actually suffers a greater performance drop from AA. There are similar results with FEAR and Oblivion. As for AF in the same game, worst case scenario at 16xHQAF it only retains 66.5% of its performance compared to the 8800GTX's 77.3% and the 8800GT's 75.1%. STALKER has the most dramatic performance impact delta, but CoH exhibits similar characteristics. Now, the only thing these benchmarks prove is that ATi's cards are faster at AA than Nvidia's but slower at AF. I did also make the claim that AF is the weak spot for RV6xx based cards.

Now, to better illustrate both my points, I've taken the liberty of extracting the most relevant numbers to show here.

First of all, I'm going to be using the benchmark from Clive Barker's Jericho because the 3870 and 8800GTX happen to have the exact same resolution scaling so that eliminates the resolution from skewing the results:


Now, first of all, lets take a look at the results without any AA or AF at 1280x1024:

As you can see, there is no major framerate advantage by any of these cards. But, now lets turn up some AA and AF:

All the GeForce cards now have a pretty clear lead, and the HD3870's performance was basically cut in half. However, AA and AF were activated at the same time, so we don't know which brought the Radeon down. But, if we crank the AA up to 8x and leave AF at 16x, the GeForce cards all take a pretty significant framerate hit while the Radeon's are largely unaffected, and the 3870 comes pretty close to matching the 8800GTX:


If you turn the resolution up to 1600x1200, you can see the same thing happen, and at 8xAA, 16xAF the 3870 actually bests the 8800Ultra:


They also tested Gothic 3 using only AF so this comparison completely eliminates AA from the equation:


The GeForce cards go largely unaffected by turning on AF while the Radeon suffers a comparatively massive performance hit.

If you look at the whole article, you will find that what I have illustrated here is a recurring theme. The Radeon tends to to compete much better at 8xAA/16xAF than it does at 4xAA/16xAF. If indeed antialiasing was the major weak point with RV6xx, then this wouldn't make sense at all. The only way to account for this is that the RV6xx's stronger AA performance is able to outweigh the AF performance hit at this point.

As for UT3, I have never played the game so I cannot say for sure, but I do recall there was quite an outcry about the only way to use AA would be to force it in the Catalyst control panel. Doing this would use AA techniques that are not optimized for the specific game and a severe performance penalty and input lag makes sense under those conditions.
Edited by darkcloud89 - 6/10/08 at 12:25pm
post #103 of 108
Quote:
Originally Posted by Vietblood315 View Post
wow!!that is nice..
u guys think i can drop a 4850 on this mobo? M2R32-MVP
Viet I would upgrade this mobo altogether lol.
New rig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon64 X2 5000 BE @3.3GHz ASUS M2R32-MVP HD3870 512MB CrossFire GeIL Esoteria 4x1GB DDR2-1100 
Hard DriveOptical DriveOSMonitor
WD Raptor 150GB Lite-On LH-20A1L DVDRW Vista Ultimate 32-bit ViewSonic 22" HD LCD 1680x1050 
KeyboardPowerCaseMouse
Logitech Media Elite PC Power & Cooling Silencer 750W Quad TT Armor Extreme Full ATX Logitech G5 Laser 
  hide details  
Reply
New rig
(13 items)
 
  
CPUMotherboardGraphicsRAM
Athlon64 X2 5000 BE @3.3GHz ASUS M2R32-MVP HD3870 512MB CrossFire GeIL Esoteria 4x1GB DDR2-1100 
Hard DriveOptical DriveOSMonitor
WD Raptor 150GB Lite-On LH-20A1L DVDRW Vista Ultimate 32-bit ViewSonic 22" HD LCD 1680x1050 
KeyboardPowerCaseMouse
Logitech Media Elite PC Power & Cooling Silencer 750W Quad TT Armor Extreme Full ATX Logitech G5 Laser 
  hide details  
Reply
post #104 of 108
Quote:
Originally Posted by darkcloud89 View Post
If you look at the whole article, you will find that what I have illustrated here is a recurring theme. The Radeon tends to to compete much better at 8xAA/16xAF than it does at 4xAA/16xAF. If indeed antialiasing was the major weak point with RV6xx, then this wouldn't make sense at all. The only way to account for this is that the RV6xx's stronger AA performance is able to outweigh the AF performance hit at this point.
First, for my 100th post, I'd like to begin by saying that I really appreciate the effort and attention you put into your response. I now understand your point much better than before.

I think my problem is that I haven't played most of the games listed, so I haven't experienced these drops from AF. However, using different levels of AF in the Orange Box, Trackmania Nations, UT3, UT2004, Assassin's Creed, and (I believe) Crysis, didn't result in a noticeable drop in performance, so I maxed it out to 16xAF. However, so much as enabling AA seemed to affect all but Trackmania too much for me to leave it on.

I'll do some further testing, but one thing is for sure: more texture units is a good thing, and better shader performance would be very welcome. I'm really looking forward to seeing some real tests of the 4870.

Rob
Endor-IV
(13 items)
 
  
CPUMotherboardGraphicsRAM
QX 9650 9x400 @ stock volts 1.25v Gigabyte EX38-DS4 Sapphire 4870 Toxic 1GB 2x2GB Corsair Dominator PC2-8500 
Hard DriveOptical DriveOSMonitor
1TB Seagate Barracuda 12 (7200.12) LG GSA-H42L Windows 7 Ultimate 64-bit Samsung 2493HM 
KeyboardPowerCaseMouse
Razer/Microsoft Reclusa Huntkey F1 620W Huntkey H001 Razer Lachesis 
Mouse Pad
Desktop 
  hide details  
Reply
Endor-IV
(13 items)
 
  
CPUMotherboardGraphicsRAM
QX 9650 9x400 @ stock volts 1.25v Gigabyte EX38-DS4 Sapphire 4870 Toxic 1GB 2x2GB Corsair Dominator PC2-8500 
Hard DriveOptical DriveOSMonitor
1TB Seagate Barracuda 12 (7200.12) LG GSA-H42L Windows 7 Ultimate 64-bit Samsung 2493HM 
KeyboardPowerCaseMouse
Razer/Microsoft Reclusa Huntkey F1 620W Huntkey H001 Razer Lachesis 
Mouse Pad
Desktop 
  hide details  
Reply
post #105 of 108
nice info, first time saw performance diff with both aa and af...
one more + to the ATI...and one more reason to buy 4850 in my future config...
    
CPUMotherboardGraphicsRAM
E5200, @3.5Ghz, 1.385 :/ Asus P5E3 Deluxe Sapphire HD 4850 Kingston 2*1gb, 1333Mhz, CL9 temp 
Hard DriveOSMonitorPower
WD Caviar Blue 6400 AAKS Win XP SP3 dual boot Vista Ultimate SP1 Samsung T220 Thermaltake Purepower RX 600w 
Case
NZXT Tempest 
  hide details  
Reply
    
CPUMotherboardGraphicsRAM
E5200, @3.5Ghz, 1.385 :/ Asus P5E3 Deluxe Sapphire HD 4850 Kingston 2*1gb, 1333Mhz, CL9 temp 
Hard DriveOSMonitorPower
WD Caviar Blue 6400 AAKS Win XP SP3 dual boot Vista Ultimate SP1 Samsung T220 Thermaltake Purepower RX 600w 
Case
NZXT Tempest 
  hide details  
Reply
post #106 of 108
Quote:
Originally Posted by RobRedbeard View Post
First, for my 100th post, I'd like to begin by saying that I really appreciate the effort and attention you put into your response. I now understand your point much better than before.
Not a problem, I just don't like to see ATi get flak for implementing AA the correct way

Quote:
I think my problem is that I haven't played most of the games listed, so I haven't experienced these drops from AF. However, using different levels of AF in the Orange Box, Trackmania Nations, UT3, UT2004, Assassin's Creed, and (I believe) Crysis, didn't result in a noticeable drop in performance, so I maxed it out to 16xAF. However, so much as enabling AA seemed to affect all but Trackmania too much for me to leave it on.
One thing I failed to mention was AF performance in older games. I would almost expect you to see a much lesser hit from AF in older games. Smaller textures would mean it will take a lot more passes of AF before the TMUs are really being stressed and so performance would possibly become shader-bound before it's texture-bound if that happens. In this case, taking a greater performance hit from AA than from AF is a reasonable outcome.

Quote:
I'll do some further testing, but one thing is for sure: more texture units is a good thing, and better shader performance would be very welcome. I'm really looking forward to seeing some real tests of the 4870.
I'm also looking forward to seeing the HD4k series in action. I hope they did their best to prevent texturing from becoming a huge bottleneck like it was with RV670, and then they can really set the massive shader power to work
Edited by darkcloud89 - 6/12/08 at 9:11pm
post #107 of 108
Quote:
Originally Posted by IcedEarth View Post
What I am confused about is, what is different between the 4850/4870 except for core/mem frequencies?

That si the only difference i can see, which means the 4850 is just an underclocked 4870. So you should be able to just by a 4850 and overclock it?

Or are they using difference core processors? I'm confused

But either way, i really hope these take off! Seems like ATI are trying pretty damn hard this time round.
It looks promising. i must admit. this may get AMD out of the gutter...they should focus everything they've got on the GPU market now.
post #108 of 108
Quote:
Originally Posted by darkcloud89 View Post
I'm also looking forward to seeing the HD4k series in action. I hope they did their best to prevent texturing from becoming a huge bottleneck like it was with RV670, and then they can really set the massive shader power to work
All around this should be an interesting generation. There are some major architectural adjustments, unlike the rather fixes between the 2x00 and 3x00 series.

Rob
Endor-IV
(13 items)
 
  
CPUMotherboardGraphicsRAM
QX 9650 9x400 @ stock volts 1.25v Gigabyte EX38-DS4 Sapphire 4870 Toxic 1GB 2x2GB Corsair Dominator PC2-8500 
Hard DriveOptical DriveOSMonitor
1TB Seagate Barracuda 12 (7200.12) LG GSA-H42L Windows 7 Ultimate 64-bit Samsung 2493HM 
KeyboardPowerCaseMouse
Razer/Microsoft Reclusa Huntkey F1 620W Huntkey H001 Razer Lachesis 
Mouse Pad
Desktop 
  hide details  
Reply
Endor-IV
(13 items)
 
  
CPUMotherboardGraphicsRAM
QX 9650 9x400 @ stock volts 1.25v Gigabyte EX38-DS4 Sapphire 4870 Toxic 1GB 2x2GB Corsair Dominator PC2-8500 
Hard DriveOptical DriveOSMonitor
1TB Seagate Barracuda 12 (7200.12) LG GSA-H42L Windows 7 Ultimate 64-bit Samsung 2493HM 
KeyboardPowerCaseMouse
Razer/Microsoft Reclusa Huntkey F1 620W Huntkey H001 Razer Lachesis 
Mouse Pad
Desktop 
  hide details  
Reply
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: Hardware News
Overclock.net › Forums › Industry News › Hardware News › [Nordic] RV770 has 800 SPs, breaks 1.2TFLOPS, final clocks