Thread Rating:
  • 0 Vote(s) - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Nvidia > ATI/AMD?
#1
Lightbulb 
Firstly, my hardware.

Desktop: stock i7 920 (2.8GHz w/TB), stock 6850 (775/4000MHz)
Laptop: stock i7 2630 (2.6GHz w/TB), OC'd 525M (700/2000MHz)

Quote:EDIT: If you want a little backstory on what inspired this thread, please read through THIS ONE.

Today I was trying to figure out just how closely the second gen i7's match the first gens in terms of PCSX performance. Turns out pretty close. I tested 5 random games and they all had very similar performance.

Below you'll find comparison screens of some of PS2's most demanding games. Notice the EE at 100% - there's no GPU bottleneck in the laptop despite the low-end card. Here are the results in text form.

God Hand
2.67ghz (no TB) 920: 74fps
2.60ghz 2630: 75fps

Zone of the Enders
2.67ghz (no TB) 920: 100fps
2.60ghz 2630: 97fps


I had the same, extremely close results in all other games, except one! Dun dun dun...

[Image: coverzoe2.jpg]

Yup. Turns out one of the absolute harshest PS2 games isn't really all that harsh if you have the right brand of card. Take a look at these results and LOL with me.

Zone of the Enders 2
2.8ghz (TB) 920: 27fps
2.6ghz 2630: 42fps Pac-Man (50% increase)

Quote:EDIT: Here are some other tests I ran to reaffirm that it's a GPU problem, be it shaders, drivers, whatever.
4.0ghz (OC) 920 + stock 6850: 36fps
4.4ghz (OC) 920 + stock 6850: 36fps
4.4ghz (OC) 920 + OC'd 6850: 42fps

The plugins and settings used were identical on both the laptop and desktop. The only major difference was the GPU. The desktop had a much more powerful AMD card, while the laptop had a weak Nvidia card.

Anyway, just something to consider next time you're upgrading.
Reply

Sponsored links

#2
Wait, are you saying my MSi R6970 Lightning 940/1375 (pre overclock) is weaker than a puny gtx 460? In pc games, it takes the gtx 460 in SLI to match the HD 6970 performance. You take a screenshot of the training mode in zoe 2, let's see how high your fps is in that scene.
Reply
#3
(09-12-2011, 03:03 AM)synce Wrote: The only major difference was the GPU.

Uh no? different chipset, different CPU, different drivers, different GPU, probably different ram too

Sandy bridge can be beaten by old Nehalem processors in different tasks, it's just that sandy bridge is generally faster. You are testing different PCs in different circumstances and you may just be seeing the difference between the different chipsets, specially since the scene you seem to be testing is not the same frame to frame.

Unless someone can make a test with same PC but different cards this is just guesswork, just like in the GPU-Benchmark thread which is much much limited to just the GPU can still show significant differences between the same card in different PC and even then the results are not bound to be a general rule for all cards.

While testing some different nvidia drivers some time ago I checked the slowest parts of the Xenosaga I intro (when the GS% is at 99%) and turns out different drivers gave a much significant increase (182.47 up to 191.07 I could get 55FPS at the lowest point but with the current 27x.xx-28x.xx I get only 37FPS) and while that was true for my PC, it turns out Squall-Le had different results with the latest drivers and getting what seems to be faster speeds when using A.G.W.S. (basically the same slowdown problem). While I got that increase with the older drivers in that game I also got lower speed with the GPU benchmark thread, seems the drivers started giving more CPU load to decrease a bit of the GPU side but it isn't even a general rule compared to other nvidia GPU/drivers. (what could it be said in that case when same CPU/chipset/ram/gpu but different drivers are what made the difference... Nvidia > Nvidia? Tongue2)
Core i5 3570k -- Geforce GTX 670  --  Windows 7 x64
Reply
#4
sandy bridge is roughly 10% faster clock for clock, but clocking the nahalem slightly higher usually offsets it.

what you will find is that SB's dual channel is faster at memory writes than nahalems trichannel if both are running at their official rated spec, you need 1600mhz ddr3 on the trichannel to keep up with 1333mhz on the SB dual channel.
Reply
#5
As I said I tested several games and they all had similar fps except for ZOE2, and it just so happens that ZOE2's fps can't manage to go higher than 36fps during the most intense parts, whether the 920 is at 4ghz or 4.4ghz. However! OC'ing the 6850 will make it go up to 42fps. So that's how I came to the conclusion it's not a CPU issue but a GPU issue. As for RAM, the sticks in my desktop are faster and have tighter timings, so that's out of the question. I'll take some more screens in a minute...
Reply
#6
k here's a comparison of a less intense scene. EE is at 100% (no bottleneck) and FPS between the two CPUs is similar, like the rest of the games I tested. When stuff starts exploding though, well, refer to the first post... Biggrin

Now of course it'd be best if I could compare Nvidia vs AMD on the same rig but to me these results are conclusive enough to steer me clear of any AMD cards in the future.
Reply
#7
timings are worth ***** on i7, you don't benefit much at all from tightening them bandwidth or latency wise.

intels main praise point in regards to the imc on i7 has been its awesome latencies.
Reply
#8
The MTVU hack gave me a lower fps for some reason. Synce, try it without the hack, let me know if you get higher or lower fps.
Reply
#9
(09-12-2011, 05:05 AM)synce Wrote: Now of course it'd be best if I could compare Nvidia vs AMD on the same rig but to me these results are conclusive enough to steer me clear of any AMD cards in the future.

Soooo-- We're on about this again? Tongue
Reply
#10
Q9550 @3.97GHz MSi R6970 @940/1375 (pre-overclocked)

Pcsx2 R4866 (1024x768)
Zoe 1 JP training mode MTVU hack on: 44.90 fps
Zoe 1 JP training mode MTVU hack off: 60.66 fps

Zoe 2 JP training mode MTVU hack on: 61.89 fps
Zoe 2 JP training mode MTVU hack off: 75.80 fps

Q9550 @2.83GHz (stock) MSi R6970 @940/1375 (pre-overclocked)

Pcsx2 R4866 (1024x768)
Zoe 1 JP training mode MTVU hack on: 35.52 fps
Zoe 1 JP training mode MTVU hack off: 43.22 fps

Zoe 2 JP training mode MTVU hack on: 45.79 fps
Zoe 2 JP training mode MTVU hack off: 53.64 fps

@synce
Your fps in zoe1 training mode is in the 90's, was it at stock speed or what clock speed was it at? I used to have the GTX 460 (768 MB, 192 bit) and tested out the ZOE series, but it didn't gave me any extra boost. However, the fps did improved compared to the 8600gt or the ATi x1600 Pro 512mb . This was when I still had my e6300 and the asus p5k-e, not sure if that was the reason why I didn't see any fps improvement on a "nvidia" card...

Reply




Users browsing this thread: 1 Guest(s)