Forum Settings
       
This Forum is Read Only

Official FFXIV Benchmarking program.Follow

#352 Jun 19 2010 at 9:59 PM Rating: Decent
**
353 posts
Quote:
The part I bolded in the quote above is false. The CPU isn't the limitation in the above screenshot, it's the GPU. Notice how that GTX 470 has an absolutely massive overclock on it (930 core, 1860 shader, and 1100 memory). That alone is what is giving him 5104 on that benchmark. The same could be achieved by any Core i7 clocked at 3.33Ghz with a similarly overclocked GPU.

AMD X2 4800+ @ 2.64Ghz + 7800GTX 512MB @ Stock clocks = ~1000 on Low (The NVIDIA 7800 GTX 512MB is GPU limited.)
AMD X2 4800+ @ 2.64Ghz + GTX 470 @ Stock clocks = ~2300 on High | 2300 on Low (The AMD X2 is CPU Limited)
Core i5-750 @ 3.2Ghz + GTX 470 @ Stock clocks = ~3300 on High (GPU Limitation only)
Core i5-750 @ 4.0Ghz + GTX 470 @ Stock clocks = ~3300 on High (GPU Limitation only)
Core i5-750 @ 3.2Ghz + GTX 470 @ 800 core, 1600 shader, 950 memory = ~4300 on High (GPU Limitation only)
Core i5-750 @ 4.0Ghz + GTX 470 @ 800 core, 1600 shader, 950 memory = ~4300 on High (GPU Limitation only)
Core i7-980x @ 3.33Ghz + GTX 470 @ 930 core, 1860 shader, 1100 memory = 5104 on High (GPU limitation only)

Unfortunately, I don't have water-cooling to bring the GTX 470 to 930 core, 1860 shader, and 1100 memory, but if I did, I'm sure I could achieve 5000 on high even with a $200 Core i5-750. In any case, this Core i5-750 computer was something I built for someone else as a low-budget non-gamer machine. The GTX 470 which I only bought for curiosities sake got returned yesterday, and an ATI 5750 will being going in instead.

If you have a Nehalem based Intel CPU (Core i3/i5/i7), your only real limitation is your GPU until you hit ~5500 on high. If you have a previous generation Intel or pretty much any AMD, you'll likely run into a CPU limitation before you max a high-end GPU on the FFXIV Bechmark. NVIDIA Fermi based GPUs (GTX 470/480) are just slower in this benchmark then their ATI 5850/5870 counterparts.

As for CPU limitations, it really comes down to FFXIV basically being a single-threaded application. It has a main thread which handles rendering that only maxes out a single core (100% / Number of cores or 25% CPU utilization on a Quad-core) , and a secondary thread (DAT loading thread?) which has ultra-low utilization the entire time. If Square-Enix ever adds real multi-threading to FFXIV, almost nobody with a modern 2.4Ghz dual-core or above would be having any sort of CPU limitation.


Mark where I bolded as I have a comment on that. Either you are biased towards intel CPUs or you are clueless. At stock speeds, the AMD phenom II x6 1090T scored about 200 less than the i7 975 on 3dMark06 and higher than the rest of the i7 chips (excluding the i7-980X which ranked 1st). (http://4.bp.blogspot.com/_9vgJ1nwu_xA/S9azBFw37yI/AAAAAAAADAw/0j0mzYSILVk/s1600/3d+mark.JPG). Furthermore, if you look here http://3.bp.blogspot.com/_9vgJ1nwu_xA/S9ay-RIFPNI/AAAAAAAADAg/nBPQhmHzL4g/s1600/Crysis+WARHEAD.JPG, it shows that at high resolution and game settings, the i7-980 extreme, i5 650, and Phenom II 1090T performed about the same with marginal differences in FPS.

Again AT STOCK SPEEDS, not everyone is after overclocking and cooling. But I just wanted to show that you will never experience CPU bottlenecking with either CPU. However, everyone should note that a 1,000 dollars processor is not giving a large performance boost over the cheaper 300 dollars processors at 1920x1080 resolution.

Conclusion: The GPU only matters and we all know that the GTX480 is underperforming. It is also better to go with AMD 1090T and save the money and get a nice GPU. It just makes more sense UNLESS YOU PLAN ON GAMING AT LOW RESOLUTION AND VIDEO SETTINGS. But seriously, in this year, who does?





Edited, Jun 20th 2010 12:08am by Imaboomer
#353 Jun 19 2010 at 10:58 PM Rating: Decent
Imaboomer wrote:
Mark where I bolded as I have a comment on that. Either you are biased towards intel CPUs or you are clueless. At stock speeds, the AMD phenom II x6 1090T scored about 200 less than the i7 975 on 3dMark06 and higher than the rest of the i7 chips (excluding the i7-980X which ranked 1st). (http://4.bp.blogspot.com/_9vgJ1nwu_xA/S9azBFw37yI/AAAAAAAADAw/0j0mzYSILVk/s1600/3d+mark.JPG). Furthermore, if you look here http://3.bp.blogspot.com/_9vgJ1nwu_xA/S9ay-RIFPNI/AAAAAAAADAg/nBPQhmHzL4g/s1600/Crysis+WARHEAD.JPG, it shows that at high resolution and game settings, the i7-980 extreme, i5 650, and Phenom II 1090T performed about the same with marginal differences in FPS.

Again AT STOCK SPEEDS, not everyone is after overclocking and cooling. But I just wanted to show that you will never experience CPU bottlenecking with either CPU. However, everyone should note that a 1,000 dollars processor is not giving a large performance boost over the cheaper 300 dollars processors at 1920x1080 resolution.


Just for the sake of accuracy, the x6 1095T that you're referencing as having outperformed all i7 CPUs (exc. i7 980X) in 3D-Mark was actually overclocked to the tune of 30%. The stock speed on those processors is 3.2Ghz...the score you're referening is for a CPU running at 4.1Ghz. The stock speed 1095T on that list scored 25% lower than the overclocked one and 30% lower than the i7 980X. Also, the Crysis performance chart you referenced is an example of a GPU bottleneck. Just like the XIV benchmark, low resolutions shift bottleneck potential to the CPU. Bump up the resolution and the bottleneck source becomes the GPU. So yes, you will experience a CPU bottleneck with the x6 1095T at stock speeds on a low res XIV benchmark or, at the very least, it will post scores lower than the i7 980x to a significant degree.

If you're talking performance/dollar, the x6 1095T is the hands down winner. In terms of performance as a 6 core CPU, it's a disappointment. The reason why Intel can continue to sell their Extreme series CPUs for the prices they're asking is because AMD can't touch them. Again, I appreciate the validity of your point in terms of a performance/$ situation, but let's not take it too far and claim performance that isn't there.
#354 Jun 19 2010 at 11:14 PM Rating: Decent
**
353 posts
Yeah but I was pointing out this quote " or pretty much any AMD, you'll likely run into a CPU limitation before you max a high-end GPU on the FFXIV Bechmark" which isn't true on high resolution.
#355 Jun 19 2010 at 11:24 PM Rating: Decent
*
209 posts
So I ran the test again this time I over clocked the processor to 3.8GHZ and opened the task manager. I got the same old song and dance, it did nothing to my benchmark. Again I scored 2700, but this time I seen the processor is running @ 18%~30% with occasional spikes to 40%. Its 100% clear that I am GPU bound and not CPU. The 470&480GTX are scoring low as well too. Its defiantly a Nvidia problem and not a problem with my CPU!
#356 Jun 19 2010 at 11:29 PM Rating: Decent
TheBSTGuy wrote:
So I ran the test again this time I over clocked the processor to 3.8GHZ and opened the task manager. I got the same old song and dance, it did nothing to my benchmark. Again I scored 2700, but this time I seen the processor is running @ 18%~30% with occasional spikes to 40%. Its 100% clear that I am GPU bound and not CPU. The 470&480GTX are scoring low as well too. Its defiantly a Nvidia problem and not a problem with my CPU!


In your case, it's an old GPU problem, regardless of who nmade it. The low performance from the 470s/480s is an entirely different issue.
#357 Jun 19 2010 at 11:35 PM Rating: Default
*
209 posts
Quote:
In your case, it's an old GPU problem, regardless of who nmade it. The low performance from the 470s/480s is an entirely different issue.
Lets not fight anymore, do you think it would make a huge difference if I put in a ATI 5870? Right now I have a 275GTX. It would only be $50 more then me going out and buying a 250GB PS3.
#358 Jun 19 2010 at 11:42 PM Rating: Decent
TheBSTGuy wrote:
Quote:
In your case, it's an old GPU problem, regardless of who nmade it. The low performance from the 470s/480s is an entirely different issue.
Lets not fight anymore, do you think it would make a huge difference if I put in a ATI 5870? Right now I have a 275GTX. It would only be $50 more then me going out and buying a 250GB PS3.


You could probably expect high res scores in the neighborhood of 4000 with your existing rig swapping the 275 for a 5870.
#359 Jun 20 2010 at 12:08 AM Rating: Default
**
353 posts
I dont see the point in upgrading from a GTX275 honestly.. its a huge waste of money. Why not just wait for the next generation of cards in 2011.
#360 Jun 20 2010 at 12:25 AM Rating: Decent
*
209 posts
Quote:
You could probably expect high res scores in the neighborhood of 4000 with your existing rig swapping the 275 for a 5870.


OK, ty 4k is not enough of a upgrade. I am going to stick with the original plan and buy a PS3 and wait for technology to evolve.
#361 Jun 20 2010 at 12:26 AM Rating: Decent
*
209 posts
Quote:
I dont see the point in upgrading from a GTX275 honestly.. its a huge waste of money. Why not just wait for the next generation of cards in 2011.
I am not going to do it, its not enough of a upgrade. I am just going to buy a PS3.
#362 Jun 20 2010 at 1:03 AM Rating: Decent
*
59 posts
Quote:
Yeah but I was pointing out this quote " or pretty much any AMD, you'll likely run into a CPU limitation before you max a high-end GPU on the FFXIV Bechmark" which isn't true on high resolution.


I'm not Intel biased at all, in fact for the past 9 years I've only had AMD CPUs in my personal computers. Even being an AMD guy, I do have to admit that for the past few years Intel has had superior high-end CPUs with AMD playing catch-up.

The only reason I said that was because of the horrible CPU bottleneck from my AMD Athlon X2 and it appearing that FFXIV may be favoring Intel Nehalem-based CPU's single-threaded performance. Again, remember that FFXIV is basically a single-threaded application, so multi-threading performance means almost nothing in the benchmark.

I've gone back and edited my post to from pretty much any AMD CPU to any any AMD which isn't a highly-clocked Phenom II to be a little more clear about what I was talking about. I still would not be surprised if a $113 dual-core Core i3-530 offered near-identical or possibly even better performance then the highest end hex-core Phenom II X6 carrying a $300 price-tag or highest end quad-core Phenom II X4 carrying a $245 price-tag in the FFXIV Benchmark. Single-threaded performance is king, and anything beyond a dual-core is unneeded and unused by FFXIV Benchmark.

Edited, Jun 20th 2010 1:02am by Cyberbeing
#363 Jun 20 2010 at 2:38 AM Rating: Decent
**
353 posts
Wrong.. FFXIV supports multicore.
#364 Jun 20 2010 at 3:03 AM Rating: Decent
*
59 posts
Quote:
Wrong.. FFXIV supports multicore.

Dual-core is multi-core. FFXIV has a single high-utilization thread, and a secondary low utilization thread. The single high-utilization thread is what limits CPU performance in FFXIV. Hence why single-threaded performance is currently king in FFXIV.

As you can see from this image, on a dual-core, only a single-thread (which equals 50% CPU usage on a dual-core) from the FFXIV benchmark is doing much of anything. A secondary thread doing much of nothing, and the rest are completely idle. If FFXIV had proper multi-threading, the image below would be showing two or more of its threads with near-equal CPU utilization, yet it doesn't. CPU usage is very lop-sided with emphasis on a single thread.

http://img251.imageshack.us/img251/5046/cpulimitation.png

Edited, Jun 20th 2010 2:59am by Cyberbeing
#365 Jun 20 2010 at 3:57 AM Rating: Decent
**
353 posts
I dont know about that. I use AMD overdrive to see the cores running and 5 of them are 50% most of the time, and one peaks to 80% sometimes.
#366 Jun 20 2010 at 4:09 AM Rating: Good
*
59 posts
Honestly I would be happy if you could prove me wrong. If you really see FFXIV using 50% CPU usage on 5 cores, I'm really curious about what is going on. If you're using Windows Vista or Windows 7, I assume it's the DirectX thread or something else OS related which is causing the extra utilization and not FFXIV, but who knows. I need some screenshot proof to really see what is going on.

Please open up Process Explorer ( http://technet.microsoft.com/en-us/sysinternals/bb896653.aspx ) when FFXIV Bench is running, double click on FFXivWinBenchmark.exe, click on the threads tab, sort by Start Address, and take a screenshot of the CPU utilization of the FFXivWinBenchmark.exe threads.

Edited, Jun 20th 2010 3:22am by Cyberbeing
#367 Jun 20 2010 at 4:16 AM Rating: Decent
Cyberbeing wrote:
Honestly I would be happy if you could prove me wrong. If you really see FFXIV using 50% CPU usage on 5 cores, I'm really curious about what is going on.

Please open up Process Explorer ( http://technet.microsoft.com/en-us/sysinternals/bb896653.aspx ) when FFXIV Bench is running, double click on FFXivWinBenchmark.exe, click on the threads tab, sort by Start Address, and take a screenshot of the CPU utilization of the FFXivWinBenchmark.exe threads.


Here's my core usage during a low res bench run. Prior to starting the benchmark all eight cores were showing fully idle.

Screenshot
#368 Jun 20 2010 at 4:25 AM Rating: Decent
*
59 posts
Quote:
Here's my core usage during a low res bench run. Prior to starting the benchmark all eight cores were showing fully idle.

That doesn't really help much, other then it appearing to support my point that only two threads are being fully-utilized. Since you have a quad-core with hyper-threading enabled, 25% Total CPU usage over 8 virtual cores = 100% CPU usage on two physical cores. The Win 7 thread scheduler shifts things around between the cores to make it appear multi-threading is working, when it really isn't.

If you want to help, please do what I described with Process Explorer. That is the only way to narrow down what the threads of FFXIV benchmark alone are doing.

Edited, Jun 20th 2010 3:43am by Cyberbeing
#369 Jun 20 2010 at 4:50 AM Rating: Good
5 posts
Nice program that Process Explorer. I downloaded and ran it on a 4 core system and the benchmark does use about 5 threads at all times when the actual benchmark is running. It jumps all the way up to 10 or 11 threads at times. That only seems to happen when there is data to be loaded, mainly during the breaks. I watched it use 6 and 7 threads at some points as well. This was at the 1080p setting, and cpu usage did hit around 70-80% at a time or two. Mainly it stayed between 40 and 60 percent.

Gpu usage was between 45 and 99 percent for the most part, mainly at the higher ranges.
Phenom II 955 @ 3.5ghz
ddr3 @ 1333 , I think
Radeon 5870 @ 891,1288

What I have found strange on my runs through the benchmark is at 720p resolution my Gpu has a low usage between 30-50 percent for the most part. On 1080p it has a much higher usage, which you would say duh to it should. However the Cpu usage is not much different between the two resolutions. The scores that I have been getting speak for themselves.

720p 4536
1080p 4352

I would think I have a bottleneck somewhere, but I'm not too sure of where it is. As the Gpu usage and the cpu usage at the 720p does not go very high. The cpu hits 60@ at its highest, and averages about 30-40%. The Gpu Averages about 40% as well. it goes as low as 25% and as high as 55%. Even when I loaded up fraps to check the fps I was getting, 720p and 1080p are near identical fps wise. I do have the Cata 10.6 drivers. I would have thought the cpu was the issue, but the usage is so low on both it and the gpu. I'm pretty sure that my 720p score should be higher than what it is when I look at some of these other scores, but I'm uncertain what is holding it back.
#370 Jun 20 2010 at 5:05 AM Rating: Good
*
59 posts
ranmas, when 720p and 1080p scores are basically identical, you are being CPU limited. If you weren't being CPU limited, your 5870 should be scoring 4800-5500 on high and 7000 on low, like people with Core i7 are getting. You may want to take a close look at your PC to make sure nothing else is wrong.

I hope this doesn't mean that a score of ~4500 is the CPU limitation of a 3.5Ghz Phenom II... Anybody else with a Phenom II and an ATI 5870 care to share their scores? Similar results? Higher? Lower? What clockspeed?

Edited, Jun 20th 2010 4:27am by Cyberbeing
#371 Jun 20 2010 at 5:29 AM Rating: Decent
5 posts
Ok I did this a bit quick and dirty but here is some screen shots of what my cpu usage is and what the threads are for the benchmark.

http://www.megaupload.com/?d=F00S8THN

its in word 2007 docx format, if you need something different I can probably get another program to take screen shots with. I just used the print screen key and pasted them into word.

I would think that the Phenom II that I have, at the clock settings it is at should at least score in the 6000+ range. Just take a look at what the i7's with the 5800 cards are doing at 1080p. My score of 4384 is not too far from that. So the 720p really should be higher than what it is.

Edited, Jun 20th 2010 7:31am by ranmas
#372 Jun 20 2010 at 5:48 AM Rating: Good
*
59 posts
Kind of odd that all your threads say ntdll.dll. If you try re-opening Process Explorer do they still all show ntdll.dll? If not, can you re-take your screenshots.

In any case, your screenshots do appear to be showing a CPU limitation and seems to look as I'd expect it for 720p on a quad-core. That high-utilization thread hovering near 25% CPU usage is one of the FFXIV benchmark threads. Since you have a quad-core, 25% Total CPU usage = 100% utilization on a single core. It's hard to say what the other threads are though. Some of the low CPU utilization threads should be FFXIV's loading threads, and another should be DirectX.

Can you overclock your CPU any higher to like 3.8Ghz and re-run the benchmark? Well I guess the first thing to check is if your CPU and RAM are even stable at 3.5Ghz. Do you get any errors when running Prime95 in Blend mode? Unstable overclock (CPU and/or RAM errors) could also possibly cause a performance drop.

Edited, Jun 20th 2010 4:56am by Cyberbeing
#373 Jun 20 2010 at 6:22 AM Rating: Decent
5 posts
It doesn't matter what process I check they all say ntdll.dll!RtlUserThreadStart. That might have something to do with UAC, I've never disabled it, or some other permission setting. I'm using Windows 7 Ultimate so there are a lot of options that I haven't really dug into or worried about. I tried to grab screen shots on its highest cpu usage. in that file. The normal usage averages between 16-19%. Still pretty high on a single core, but the 1080p version is the same. I wish this benchmark supported other resolutions to give us a better idea of what it was doing.

3.8ghz was easy to hit, I didn't hardly have to change anything to get it running at that speed. I think I may be able to get it running at 3.9ghz on air. 4.0 is a no go, stupid windows 7 64 and the phenom II's don't get along at that speed or over really. At least amd fixed that bug in the 6 core processors.

Anyway 3.8ghz didn't change the scores by that much, and the only think it made me think is possibly at 1080p we are cpu limited.

720p 4869
1080p 4566

Those numbers are still too close together.

Edited, Jun 20th 2010 9:56am by ranmas
#374 Jun 20 2010 at 6:51 AM Rating: Good
*
59 posts
If it can't be fixed, I wouldn't stress too much over it. A benchmark is only a benchmark and with the frame-rates seen in your screenshots, you should still have a great experience with FFXIV. Once Square-Enix has finished adding the more GPU intensive settings like SSAO, along with you cranking up the anti-aliasing from the Catalyst CC, your CPU limitation will become less and less relevant.
#375 Jun 20 2010 at 12:25 PM Rating: Decent
ranmas wrote:
It doesn't matter what process I check they all say ntdll.dll!RtlUserThreadStart. That might have something to do with UAC, I've never disabled it, or some other permission setting. I'm using Windows 7 Ultimate so there are a lot of options that I haven't really dug into or worried about. I tried to grab screen shots on its highest cpu usage. in that file. The normal usage averages between 16-19%. Still pretty high on a single core, but the 1080p version is the same. I wish this benchmark supported other resolutions to give us a better idea of what it was doing.

3.8ghz was easy to hit, I didn't hardly have to change anything to get it running at that speed. I think I may be able to get it running at 3.9ghz on air. 4.0 is a no go, stupid windows 7 64 and the phenom II's don't get along at that speed or over really. At least amd fixed that bug in the 6 core processors.

Anyway 3.8ghz didn't change the scores by that much, and the only think it made me think is possibly at 1080p we are cpu limited.

720p 4869
1080p 4566

Those numbers are still too close together.


There's a bit of a see-saw shenanigan with the benchmark that largely requires you to mess around with overclocking between CPU and GPU to figure out where your bottleneck is happening. At 1080p, I'm GPU limited with a highest score of just over 5600. When I first ran the benchmark I got just over 4800 and the extra 800 points came from a pretty beefy overclock on my GPU. It's the opposite with 720p. When I first ran the benchmark I got something like 7400 (actual score is in the benchmark thread somewhere on page 1). After some tweaking with the processor overclock I pulled that up to over 7800 and then the same GPU overclock that got me 800 extra points to my score in 1080p only added another 150-200 to my low res score. What makes it tricky is that those limitations and bottlenecks are only descriptive of my rig. Someone with a faster card (ie. GTX480/5970 with better drivers) might experience the bottlenecks at different times from different sources.

Also, for Cyberbeing:

Screenshot
#376 Jun 20 2010 at 12:53 PM Rating: Decent
**
353 posts
I think we all forgot to realize that the benchmark is running in Windowed mode which will lower performance in the benchmark. Fullscreen would have given us better results so don't be sad if your score is low.
#377 Jun 20 2010 at 1:36 PM Rating: Decent
Scholar
*
141 posts
Hey guys,

Ive been trying to use the benchmark and it wont work for me. I kept getting an error so i tried updating my drivers for my graphics card and that didnt work.

The error said that d3dx9_51.dll cant be found. Anyone else get this?

Sorry if this has already been addressed, im in a rush today and didnt really have time to read all the pages.
#378 Jun 20 2010 at 2:29 PM Rating: Decent
5 posts
Try and update directx.
#379 Jun 20 2010 at 7:16 PM Rating: Good
***
1,025 posts
I can't verify the legitimacy of these benchmark videos, but it looks like this person has managed to get to a ~10000 score on High.
http://www.youtube.com/watch?v=15Zw_FX0MDg
http://www.youtube.com/watch?v=o67n8tOHTis



Edited, Jun 21st 2010 1:32am by Keikomyau
____________________________
日本台湾友好!
#380 Jun 20 2010 at 7:44 PM Rating: Good
Keikomyau wrote:
I can't verify the legitimacy of these benchmark videos, but it looks like this person has managed to get to a ~10000 score on High.
http://www.youtube.com/watch?v=15Zw_FX0MDg
http://www.youtube.com/watch?v=o67n8tOHTis


I almost wish you hadn't posted that because now I want to be all smug and "I told you so" with all the people who cried and screamed that SE dropped the ball for not supporting SLI/Crossfire. Here's what the person did in that video to breach the 10000 mark on high res:

1) i7 980X processor.
2) Two Radeon 5970 GPUs in Crossfire (aka four 5870 cores all working happily together)
3) Renamed benchmark executable to 3DMark06.exe to trick the GPU drivers into supporting Crossfire.
4) Windows running in basic mode (ie. no Aero)
5) All CCC enhancement options disabled (ie. AA)
6) ???
7) Profit.

So yes, you're looking at a video of someone with a $4-5000 computer overclocking their processor (and I believe their GPUs) to produce the kinds of results that make us all drool, but the point is that they wouldn't have been able to score anywhere near that high unless Crossfire was working.

Edit for clarity: For those with SLI/Crossfire rigs who want to try this, the executable you want to rename is not the one in the main extract directory. It's in the \FFXIVBenchmark\Data directory called FFXivWinBenchmark.exe.

Edited, Jun 20th 2010 6:51pm by Aurelius
#381 Jun 20 2010 at 8:15 PM Rating: Decent
Scholar
44 posts
Keikomyau wrote:
I can't verify the legitimacy of these benchmark videos, but it looks like this person has managed to get to a ~10000 score on High.
http://www.youtube.com/watch?v=15Zw_FX0MDg
http://www.youtube.com/watch?v=o67n8tOHTis

Edited, Jun 21st 2010 1:32am by Keikomyau


Awesome! Does anyone have a screenshot of what 2 5850s would score using the method Aurelius posted above? I'm about to either order another 5850 or just swap my current one for a GTX 480. Id like to see how it scales in this specic bench first.
#382 Jun 20 2010 at 8:26 PM Rating: Decent
malcolmtn wrote:
Keikomyau wrote:
I can't verify the legitimacy of these benchmark videos, but it looks like this person has managed to get to a ~10000 score on High.
http://www.youtube.com/watch?v=15Zw_FX0MDg
http://www.youtube.com/watch?v=o67n8tOHTis

Edited, Jun 21st 2010 1:32am by Keikomyau


Awesome! Does anyone have a screenshot of what 2 5850s would score using the method Aurelius posted above? I'm about to either order another 5850 or just swap my current one for a GTX 480. Id like to see how it scales in this specic bench first.


It's too early to tell. I posted the same information on BG so between the two forums, hopefully we get some quality multi-GPU scores in soon.
#383 Jun 20 2010 at 11:03 PM Rating: Good
Thief's Knife
*****
15,053 posts
I said earlier that the reason SLI/Crossfire wasn't working was because it didn't have a profile yet.
____________________________
Final Fantasy XI 12-14-11 Update wrote:
Adjust the resolution of menus.
The main screen resolution for "FINAL FANTASY XI" is dependent on the "Overlay Graphics Resolution" setting.
If the Overlay Graphics Resolution is set higher than the Menu Resolution, menus will be automatically resized.


I thought of it first:

http://ffxi.allakhazam.com/forum.html?forum=10&mid=130073657654872218#20
#384 Jun 21 2010 at 9:03 AM Rating: Default
*****
12,824 posts
Well, I don't trust this benchmark tool all that much. My old gaming notebook scored a 780, then a 794, then 840, then 804. So I decided to flip on the overclocking for the CPU... and it score a 734.

Half the time the frames were running smooth with the overclocking but the meter was flatlined. My other computer tested through high enough that if my wife wants to play, I know at least one computer is guaranteed to run it. I'll just install it on here then (or compare the specs when they are released), and go from there.

FailBenchmark is Fail.
____________________________
Twitter: http://www.twitter.com/pawkeshup
YouTube: http://www.youtube.com/pawkeshup
Twitch: http://www.twitch.tv/pawkeshup
Blog: http://pawkeshup.blogspot.com
Olorinus the Ludicrous wrote:
The idea of old school is way more interesting than the reality
#385 Jun 21 2010 at 9:07 AM Rating: Excellent
Quote:
Half the time the frames were running smooth with the overclocking but the meter was flatlined. My other computer tested through high enough that if my wife wants to play, I know at least one computer is guaranteed to run it. I'll just install it on here then (or compare the specs when they are released), and go from there.


On an old laptop I imagine you are bottlenecking at your GPU not your CPU.

I have a relatively new video card and older processor, and have seen increased benchmark scores from upping my overclocking.
#386 Jun 21 2010 at 9:22 AM Rating: Good
Pawkeshup Delivers on Time wrote:
Well, I don't trust this benchmark tool all that much. My old gaming notebook scored a 780, then a 794, then 840, then 804. So I decided to flip on the overclocking for the CPU... and it score a 734.

Half the time the frames were running smooth with the overclocking but the meter was flatlined. My other computer tested through high enough that if my wife wants to play, I know at least one computer is guaranteed to run it. I'll just install it on here then (or compare the specs when they are released), and go from there.

FailBenchmark is Fail.


Hard to help point out whether your abrupt criticism of the benchmark is on point or not when you give us next to no information. Be reasonable...the scale we were given says that any score lower than 1500 means your computer isn't going to be able to manage XIV at all...a 5% variance on a benchmark score for a computer working so hard just to chug through the benchmark doesn't mean much. It's like saying your lawn tractor doesn't manage the hills in San Fransisco with any consistency. The benchmark tool is fine with abundant scores in now from dozens of machines that indicate that not only is the benchmark a reasonable estimate but that the hardware demands for XIV are on point with what we were told they would be.
#387 Jun 21 2010 at 2:10 PM Rating: Decent
Guru
***
1,673 posts
Aurelius, I sent you a PM.
#388 Jun 21 2010 at 3:49 PM Rating: Decent
Scholar
**
630 posts
Been browsing around on newegg today and I think I may have figured out why the ATI cards are running better scores on the benchmark...stream processors. Even the higher end GTX 480 maxes out at 448 cores, whereas the 5850s have 1440 processing units. Of course they don't quite rank these all the same, but that seems to be the only major difference (most everything else is similar, except with the 470/480s you are paying extra for a bigger memmory size, interface, and effective clock).
#389 Jun 21 2010 at 5:34 PM Rating: Decent
*****
12,824 posts
The One and Only Aurelius wrote:
Pawkeshup Delivers on Time wrote:
Well, I don't trust this benchmark tool all that much. My old gaming notebook scored a 780, then a 794, then 840, then 804. So I decided to flip on the overclocking for the CPU... and it score a 734.

Half the time the frames were running smooth with the overclocking but the meter was flatlined. My other computer tested through high enough that if my wife wants to play, I know at least one computer is guaranteed to run it. I'll just install it on here then (or compare the specs when they are released), and go from there.

FailBenchmark is Fail.


Hard to help point out whether your abrupt criticism of the benchmark is on point or not when you give us next to no information. Be reasonable...the scale we were given says that any score lower than 1500 means your computer isn't going to be able to manage XIV at all...a 5% variance on a benchmark score for a computer working so hard just to chug through the benchmark doesn't mean much. It's like saying your lawn tractor doesn't manage the hills in San Fransisco with any consistency. The benchmark tool is fine with abundant scores in now from dozens of machines that indicate that not only is the benchmark a reasonable estimate but that the hardware demands for XIV are on point with what we were told they would be.
Sure thing. My gaming notebook is an XPS 1730 running Windows XP with an Intel Core 2 Duo X7900 running at 2.8Ghz (Overclockable to 3.0 Ghz, 4GB of RAM, SLI 8700M GT's with 512MB of RAM onboard (if you need more info I'll just screenshot it), and RAID 0 configured HDDs whose specs elude me at the moment.

Now, I'm not going to be as egotistical as to say that this system should run this game. Why I call bullsh*t on this particular benchmark is because increasing the clockspeed would have a major difference (dropping 100 points is not 5% loss, by the way) in a negative way makes no sense. At worst, it would be just the same as before. And since the thread on BG is having some random good and bad results with similar system builds, the test is flawed.

So, I will wait for the next build or for the actual game and give it a spin. As an FYI, the other computer is definitely more powerful in several ways over this one, though not cutting edge so, and it scored above 2400 on low res (Monitor is the only old school part, so no reason to bother testing high res as it couldn't fit the screen).

Oh, and to repeat what has been said on BG: Running the test in Windowed mode will also produce worse results as well.

Edited, Jun 21st 2010 7:38pm by Pawkeshup
____________________________
Twitter: http://www.twitter.com/pawkeshup
YouTube: http://www.youtube.com/pawkeshup
Twitch: http://www.twitch.tv/pawkeshup
Blog: http://pawkeshup.blogspot.com
Olorinus the Ludicrous wrote:
The idea of old school is way more interesting than the reality
#390 Jun 21 2010 at 6:51 PM Rating: Decent
5 posts
could someone provide a link to this BG forum. I can't say that I know what the BG stands for, but having another thread to look at is always a good thing.
#391 Jun 21 2010 at 6:56 PM Rating: Good
Pawkeshup Delivers on Time wrote:

Now, I'm not going to be as egotistical as to say that this system should run this game. Why I call bullsh*t on this particular benchmark is because increasing the clockspeed would have a major difference (dropping 100 points is not 5% loss, by the way) in a negative way makes no sense. At worst, it would be just the same as before. And since the thread on BG is having some random good and bad results with similar system builds, the test is flawed.

So, I will wait for the next build or for the actual game and give it a spin. As an FYI, the other computer is definitely more powerful in several ways over this one, though not cutting edge so, and it scored above 2400 on low res (Monitor is the only old school part, so no reason to bother testing high res as it couldn't fit the screen).

Oh, and to repeat what has been said on BG: Running the test in Windowed mode will also produce worse results as well.


So in other words, it's easier to blame SE than to come directly to terms with the reality that your laptop is in its golden years. Maybe you don't quite grasp how benchmarks work. It's really not all that complex. The benchmark score is a arrived at from a calculation based on frames rendered. That's it. So if your scores are wonky, it's a PC issue, not a benchmark issue. If your scores aren't as high as you think they should be, it's a PC issue, not a benchmark issue.

Did you read the part on BG where they quoted Tanaka's statements from a year ago with regards to what XIV would require for hardware at release?

Edited, Jun 21st 2010 6:43pm by Aurelius
#392 Jun 21 2010 at 7:21 PM Rating: Decent
5 posts
My laptop is failing at the benchmark. Have i7 QM720 processor and nvidia GTX 260M for graphics card, and getting about 430. Is that graphics card really that bad, or is something jacked up? I know video card driver is up to date at least.
#393 Jun 21 2010 at 7:45 PM Rating: Decent
Malitoto wrote:
My laptop is failing at the benchmark. Have i7 QM720 processor and nvidia GTX 260M for graphics card, and getting about 430. Is that graphics card really that bad, or is something jacked up? I know video card driver is up to date at least.


Sounds low for a low rez bench score but you didn't tell use whether you ran low res or high res and there's nothing I can do for ya until you do :P
#394 Jun 21 2010 at 7:57 PM Rating: Decent
5 posts
Yeah, it was low res. Also, I just remembered, though it seems like something that would be hard to forget, the motherboard was just replaced. I don't know if that would affect anything though.
#395 Jun 21 2010 at 9:15 PM Rating: Decent
Malitoto wrote:
Yeah, it was low res. Also, I just remembered, though it seems like something that would be hard to forget, the motherboard was just replaced. I don't know if that would affect anything though.


It's hard to say for sure but I've seen higher scores with comparable hardware. Low res scores are typically bottlenecked by CPU on desktop but laptops seem a little different. How much RAM does your laptop have and what OS are you using?
#396 Jun 21 2010 at 9:20 PM Rating: Decent
5 posts
I'm on an Alienware M15X, only about 5 months old, running Windows 7 home premium and 4 GB DDR3 RAM.
#397 Jun 21 2010 at 9:34 PM Rating: Decent
Malitoto wrote:
I'm on an Alienware M15X, only about 5 months old, running Windows 7 home premium and 4 GB DDR3 RAM.


I'd get it checked out. Something isn't working. No reason for an i7 (mobile or not) to be struggling like that on a low res benchmark. I know laptop GPUs tend to not perform anywhere near the level of their desktop counterparts, but it seems to me like your CPU isn't even letting your GPU get to the point where it's an issue.
#398 Jun 21 2010 at 9:39 PM Rating: Decent
5 posts
Alright, thanks a bunch for your help. You're the man, Aurelius!
#399 Jun 21 2010 at 10:04 PM Rating: Good
Sage
**
743 posts
AngusX wrote:
For the people keeping score, I got a low res score of about 650 and a high res score of about 321 (I'll update the numbers when I get home).
My current rig:

Case Used HP case free
Power Supply OCZ 600W ModXStream Pro $50 (on sale + rebate)
MB Gigabyte ga-h55m-ud2h $100
Mem OCZ 3P1333LV4GK 2x2GB Platinum DDR3 $115
HD Seagate Barracuda 7200.12 500 GB $55
Optical drive Asus dvd $25
CPU Intel i3-530 (OC'ed to 3.6 GHz) $100
OS Win7 Ult. 64 $50
CPU cooler CoolerMaster 212+ $30
GPU PowerColor HD4650 $35 (after rebate)


Update: just installed an Asus CuCore HD5770. Score w/o GPU OC:
Low res - 3900
High res - 2557
____________________________
I think you've been smoking the Moko...
http://na.finalfantasyxiv.com/lodestone/character/350413/
http://na.finalfantasyxiv.com/lodestone/character/1628942/
http://www.nerdist.com/
Angus of Cerberus (retired)
#400 Jun 22 2010 at 1:44 AM Rating: Decent
Sage
***
1,675 posts
Final low score = 3705

Windows 7 Ultimate 64-bit
Intel Q9550 @ 3.4ghz
Nvidia 8800 GT 512MB (latest Nvidia drivers/overclocked w/rivatuner)
6GB DDR2 800
7200 500GB HD

Notes: I turned off Aero, and disabled all background programs.

I'm pretty happy with this. I'm sure things will be more optimized as the game progresses.

Also when the game comes out we'll all have a better idea if we should upgrade or not.







Edited, Jun 27th 2010 7:59am by Kierk
#401 Jun 22 2010 at 2:17 AM Rating: Good
*
59 posts
Quote:
Update: just installed an Asus CuCore HD5770. Score w/o GPU OC:
Low res - 3900
High res - 2557


Your high resolution score seems about right, but I would have expected the low resolution score to be a bit higher. Below is what I'm getting from a $670 ATI 5750, Core i5-750 @3.2Ghz (200x16), 4GB DDR3 1600Mhz 7-8-7-20-1T computer I put together for a family member (non-gamer). Since a Core i3-530 @3.6Ghz should outperform the Core i5-750 @3.2Ghz in this benchmark (with the exception of loading), either FFXIV benefits greatly from extra bandwidth of low latency 1600Mhz DDR3 at 720p (quite possible), or something else is holding your score back. Though as long as your low resolution score continues to go up when you overclock the ATI 5770, there should be no problem. If it doesn't go up, you also may want to make sure you raised your BCLK adequately when you did your overclock:

5750 Stock Clocks 700Mhz core, 1150Mhz memory:
Low - 3939
High - 2166

5750 Overclocked 850Mhz core (same as stock 5770), 1385Mhz memory (higher than stock 5770):
Low - 4654
High - 2575

5750 Overclock 900Mhz core, 1450Mhz memory:
Low - 4909
High - 2725

Edited, Jun 22nd 2010 1:24am by Cyberbeing
This forum is read only
This Forum is Read Only!
Recent Visitors: 24 All times are in CST
Anonymous Guests (24)