New Sapphire 5850 vs BFG 8800GTX OC Benchies :)

[dlknight]

Hello,

A short while ago I decided it was time to buy a new graphics card and give my mate Mortster back his trusty BFG 8800GTX that I have been borrowing for the last year or so.

I decided on a ATI Radeon 5850 after originally buying an XFX HD4890 from auction, unfortunately this card wasn't great (this may have been due to the non-reference design PCB) and I returned it to the seller (seller subsequently managed to get a full refund from the etailer it had originally been purchased from).

Just to clarify yes the 4890 results are dodgy and this appears to be because the card was from a bad batch of XFX cards made and sold in August 2009 - http://www.overclock.net/ati/634248-xfx-4890-bad-batch-discussion.html

So after this I was a bit dubious about buying another 4890, especially an XFX card so after looking at lots of reviews and also the price compared to cards from NVIDIA my heart was set on a 5850. I have always been a NVIDIA buyer but the quality of the 48** series cards has swayed me towards ATI.

My run of luck continued and after purchasing a Club 3D 5850 which lasted only 24 hours (refusing to POST) I received a replacement card from OCUK, a Sapphire HD5850.

Initial impressions are thumbs up, many thanks to OCUK for changing my RMA card to the Sapphire, think they were having issues getting stock of the Club 3D cards.

The image quality within Windows seems warmer and more natural than the impression the 8800 gave. Especially viewing TV recordings from my Mythtv box in VLC, I watched a few recordings of the Winter Olympics on the PC whilst doing other stuff and the colour and clarity of the BBC HD recordings was superb, it also seemed that some of the video processing was being offloaded to the GPU nice one ATI. NVIDIA have video processing in the form of VDPAU so was great to see that ATI are doing something similar.

One of the features I was keen on using is the ATI eyefinity feature as I have two Samsung 22" LCD's. After exploring the options within the Catalyst Manager I worked out that I needed to create a profile and then setup a display group within that profile. I duly did this then had a couple of hours play on MW2 in single player mode at 3360x1050 , with only two monitors its a bit tricky having the crosshair in the centre but I managed to play and it didn't annoy me one bit.

In preperation for the new video card I had performed various benchies on the 8800 and when I had the 4890 I did the same benchies on that, so here are the final results tallied together along with some fancy graphs (cause I like playing).

Tests were performed on the following system:

• ASUS P5Q-E Motherboard with Intel E6300 Conroe @ 3360MHz
• 2 x 2GB OCZ Reaper HPC Edition
• 2 x 250GB 7200rpm SATA drives in Raid 0
• Windows 7 RC 64 Bit
• NVIDIA Driver version 196.21
• ATI Catalyst 10.1
• Crysis 1.2.1
• DiRT 2 Demo
• Stalker Clearsky Benchmark

Crysis

Results in Crysis kind of speak for themselves, as you can see on some of the results with 0xAA it appears there is a bottleneck which I am guessing is the CPU. Graphs below show the benefits of the 5850 slightly better than the table:





DiRT 2

Results in DiRT 2 show the same bottleneck from the Crysis results above which again I am assuming is down to the CPU.

Online reviews/benchies showing similar results for my setup - http://www.pcgameshardware.com/aid,700780/Dirt-2-CPU-benchmarks-with-DirectX-9-and-DirectX-11-Phenom-doing-well-quad-cores-rule/Practice/

Stalker ClearSky Benchmark

As you can see significant improvements are shown in the Stalker ClearSky Benchmark







Cheers

 

[dlknight]

*BUMP* readded the post, I removed it as though I could enter it in competition but would have to do a bit more work really

 

[dlknight]

Not sure if anyone else has noticed the obvious flaws in the graphs - pretty poor.

Which flaws might they be then?

Flawed results, for a start you say dirt 2 runs in DX9 on the 5850 when it runs Dx11 at default, so I take it you edited the config file so it plays dx9 only? if that's the case you would be seeing 80-90fps even with your crappy CPU.

Crysis, why you run only high, thats' for the birds, run it at very high... a number of other dodgy looking results but CBA to go over them.

Erm yes I edited the config file and your point is?

80-90fps really? please provide the evidence to back your claims...

Again dodgy results, please elaborate...

 

[dlknight]

I suggest you go and take a look at the "stacked bar chart" template in Microsoft Excel 2007 before you post anymore crap.

Then maybe if you take a look at the key you will see that the minimum and average FPS have been combined into one bar. Hence why the FPS is shown on each segment of the bar.

How I present my results is purely up to me and didn't realise you were the master here...

I could have flooded the entire post with all my results in graph format with seperate bars for minium and average fps but then you probably would have complained that the webpage didn't fit into your crappy 15" LCD that you mommy and poppy bought you last xmas.

 

[dlknight]

Just ran Dirt 2 at 1680x1050 max settings, forced DX9 and no AA, on a 5870 clocked at 5850 speeds and an i7 920 underclocked to 2.2 GHZ and two cores disabled. Results tell me the OP has got some major performance issues..

Posting results from a totally different architecture even though you might have disabled cores/underclocked etc. Doesn't help does it? what version of Windows are you using? driver versions, disk configuration etc...

Actually looking at online reviews my results are about right - http://www.pcgameshardware.com/aid,700780/Dirt-2-CPU-benchmarks-with-DirectX-9-and-DirectX-11-Phenom-doing-well-quad-cores-rule/Practice/

But then I guess you will say these are incorrect too ?

 

[dlknight]

4890 results are very flawed, which makes me think the 5850 are too.

Just for comparision I was getting 55 fps in Crysis @ 1680x1050/High with my 5850. This was on a 3.4ghz Q9550, but since Crysis hardly uses multi-core processors, it's not going to make a huge difference.

Regardless of this though, a 4890 destroys a 8800GTX, so there's something wrong anyway.

You are correct about the 4890, did you read my first post although it may not have been entirely clear.

The 4890 appears to have been from a dodgy batch of XFX HD4890's produced and sold in August last year, hence why I returned it and they seller gave me a full refund.

http://www.overclock.net/ati/634248-xfx-4890-bad-batch-discussion.html

The guy I bought it off is twodigitz and has posted in the forum linked to above.

Your Crysis results are similar to mine at that resolution and settings.

 

[dlknight]

Just face facts your results are screwed up, noway does a 5850 get that lower FPS in dirt 2, it's just about on par with a 8800GTX according to your results, it's crrrrrrrrrrrrazy...

Have you ever heard of a system bottleneck? I'm guessing it is my CPU because if I decrease resolution to 800x600 then the FPS increases to over 100FPS.

I'm pretty happy with my 5850, I posted my results as they were reported by the software. I have been pretty comprehensive with my posting and yes maybe I could have made the graphs look a little bit better.

 

[dlknight]

But surely that indicates that you do not have CPU bottlekneck, because resolution affects only the GPU.

Ahh yes you maybe right I will check this later on when I get home and post the result of running at lower resolution. I may have changed the quality setting and the resolution when I was getting 100FPS+

 

[dlknight]

17%, which can be quite a hit with AA


dlknight - good effort. Stacked graphs are a pain and wouldnt be the way to go without some data-fiddling. Clustered bars would be better and easier with excel

I may do some clustered bars on the weekend if I feel the need for some Excel time LOL.

 

[dlknight]

If you guys think theres something wrong with my DiRT 2 results I suggest you look here - http://www.pcgameshardware.com/aid,700780/Dirt-2-CPU-benchmarks-with-DirectX-9-and-DirectX-11-Phenom-doing-well-quad-cores-rule/Practice/

Which clearly shows that DiRT 2 is heavily CPU dependent, look at the scores for an E6600 and an E8400.

Cheers

 

[dlknight]

It has been awhile but I have updated the Excel Charts to be Clustered Bar instead of my mistake of the stacked bar...

@Buckster read the first post as it explains the 4890 results caused by a faulty XFX card.