Crank that S#!t Up
Perfect World Beta
Life After DX 11
AMD Contest Group
Final Round FiringSquad-Intel Edito...
Round 3 Editors Challenge Sponsored...
S.T.A.L.K.E.R. Screenshot Contest
Top 10 Challenge Round Sponsored by...
Editors Challenge Sponsored by Inte...
FiringSquad Editors Challenge Round...
Synchronous Failure (1147)
||23 entry(ies) in this category
| SHOWDOWN: Intel, ATi, NVIDIA! Oh, my! (Two) (7 comments )|
by: jacobvandy (1636) | Posted in cluster Final Round FiringSquad-Intel Editors Challenge
Posted 73 months ago ( edited 73 months ago ) in category DEFAULT
It's been a hectic few days. Two computers, one mouse, one keyboard, and one monitor. You do the math. Remote Desktop Connection helped a little, except you can't access a BIOS remotely. Believe me when I say that this competition has bestowed upon me a whole new respect for those that do this kind of thing all the time. But I digress. We both know why you're here. I hope you enjoy.
|» MEDIA (13)|
Figure 1: 3DMARK06 System Comparison
Figure 2: e-GeForce 7900 GS KO Game Test Results
Figure 3: Company of Heroes Test Settings
Figure 4: Counter-Strike: Source Test Settings
Figure 5: Game Tests GPU Comparison
Figure 6: 3DMARK06 GPU Comparison
Figure 7: 3DMARK06 CPU Comparison
Figure 8: Counter-Strike: Source CPU Comparison
Figure 9: Supreme Commander CPU Comparison
Figure 10: SuperPi CPU Comparison
Figure 11: SANDRA Arithmetic Operations Results
Figure 12: SANDRA Multi-media Operations Results
Figure 13: CPU Temperature
With the entire system up and running, the technical specifications consisted of the following:
• Intel Core 2 Extreme X6800 2.93GHz Processor
• EVGA NVIDIA nForce 680i LT SLI Motherboard
• 2GB Corsair Dominator DDR2-1066 Memory
• EVGA e-GeForce 7900 GS KO 256MB Video Card
• Maxtor 6L300S0 300GB SATA Hard Drive
• Microsoft Windows XP Professional w/ SP2
Throughout testing, parts were mixed and matched with my previous rig:
• Intel Core 2 Duo E6600 3.0GHz Processor
• Gigabyte GA-965P-DS3 Rev. 2.0 Motherboard
• 2GB OCZ Platinum DDR2-800 Memory
• ATi Radeon X1900 XT 512MB Video Card
• Maxtor 6L300S0 300GB SATA Hard Drive
• Microsoft Windows Vista Ultimate 32-bit
Yes, they have names. But since they were two complete systems, what else could I do but pit them against each other? For kicks and giggles, or perhaps a bit of foreshadowing, preliminary 3DMARK06 runs were recorded. See Figure 1. Now, without further ado, let the games begin!
• 3DMARK06 Basic
• Call of Juarez
• Counter-strike: Source
• Supreme Commander
• Company of Heroes
• SiSoftware SANDRA Lite
• SuperPi Mod
• EVEREST Ultimate Edition
• Orthos Prime
» AN OLD DOG PERFORMS NEW TRICKS
The GeForce 7900 series of graphics cards was released last year to replace the 7800 series. Based on the G71 chip, a revision of the G70, it offered higher performance, lower power consumption, and more competitive pricing. How does EVGA's version of this last-generation midrange card perform? Read on!
• Intel Core 2 Duo E6600 (3.0GHz)
• Gigabyte GA-965P-DS3 Rev. 2.0 (BIOS version F7)
• 2GB OCZ Platinum PC2-6400 (OCZ2P8002GK)
• Maxtor 6L300S0 (7200 RPM, SATA, 16MB Cache)
• Windows Vista Ultimate 32-bit
• EVGA e-GeForce 7900 GS KO 256MB (ForceWare 158.18)
• ATi Radeon X1900 XT 512MB (Catalyst 7.4)
See Figure 2.
Call of Juarez is a first-person shooter from Ubisoft and Techland that takes place in the Old West. Several advanced visual effects like HDR lighting, depth of field, and high-resolution dynamic shadows serve to simultaneously please and frustrate you. Of course, the degree to which it does these things depends upon what sort of hardware you're packing and the graphics settings you play on. For the purposes of this article, the three settings presets were used, along with a resolution of 1280x1024. FRAPS recorded the average FPS of a rambunctious romp through a heavily wooded area and subsequent leap into a ravine. The 7900 GS KO performed surprisingly well, though not even the preset entitled "Best Quality" included the highest caliber of shadowing, with which I am used to playing.
Back for more system crushing, F.E.A.R. was benchmarked through the proprietary performance test using the medium, high, and maximum presets. Note that, on the results graph, those three correlate to low, medium, and high, respectively. With triple-digit scores on all but the highest settings, either the 7900 GS KO is more formidable than predicted or this shooter is losing it's chilling touch.
No set of benchmarks is complete without a game that involves fighting Nazis. Enter Company of Heroes, Relic's latest real-time strategy smash. You can zoom in on individual soldiers and it still looks better than a lot of the third-person action games out there. The in-game performance test was run at 1280x1024; Figure 3 illustrates the setting combinations used. Like a lot of other games, the framerate is largely independent of the graphics card until more sophisticated visual effects are enabled. The 7900 GS KO pulled a very respectable average on the highest settings, though an RTS is more than playable at half of that.
With its grand scale of futuristic warfare, Supreme Commander is quite the system taxer. The low, medium, and high fidelity presets were tested at 1280x1024. An integrated performance test assigns a composite score, but I decided to report the average FPS this time around. Who knows how many "SupCom Marks" is a good amount, anyway? The 7900 GS KO struggles in the teens on high settings, but on the bright side, it's only a smidgen slower than medium.
Counter-Strike: Source is the graphically enhanced little brother of the most popular online FPS of all time. As the name implies, it uses Valve's Source engine, one that is well known for its performance scalability. Supporting older versions of DirectX rendering, even lower-end desktops and laptops can run Source engine games with little trouble. CS:S includes a video stress test, which was run at 1280x1024 with the settings outlined in Figure 4. As long as you are comfortable playing with any less than 100 FPS, you can partake in a generally smooth experience on the absolute maximum level of graphical detail with the 7900 GS KO.
Without an entire back catalog of test results for several video cards, I can provide at least one thing for comparison: the Radeon X1900 XT. See Figure 5. These results are from each of the five games run on high settings. Although the 7900 GS KO is obviously not in the same league, it is rather decent, especially considering it costs about three times less than I paid for the X1900 XT a little over a year ago.
For this last bit of testing, I decided to flex the 7900 GS KO's overclocking muscle. Using an application called RivaTuner, the GPU core and memory clocks were increased to 550MHz and 775MHz, respectively. Admittedly, not much time was spent determining the absolute highest stable frequencies, but these results are nothing to scoff at. Increasing the core clock by 10% and the memory clock by 12% yielded a very decent performance boost. See Figure 6 to find out how the 7900 GS KO and its overclocked counterpart fared against the X1900 XT in the "Gamer's Benchmark."
» SIBLING RIVALRY: E6600 VS X6800
In my previous review of the Core 2 Duo E6600, I hypothesized that overclocking it would yield performance comparable to the much more expensive Core 2 Extreme X6800. I am pleased to announce that this competition has provided me an opportunity to comprehensively test that theory.
• EVGA NVIDIA nForce 680i LT SLI (BIOS version
• 2GB Corsair Dominator PC2-8500 (TWIN2X2048-8500C5D)
• ATi Radeon X1900 XT 512MB (Catalyst 7.4)
• Maxtor 6L300S0 (7200 RPM, SATA, 16MB Cache)
• Windows XP Professional SP2
• Intel Core 2 Duo E6600 (2.4GHz and 3.0GHz)
• Intel Core 2 Extreme X6800 (2.93GHz)
With a sizable portion of the test suite dedicated to putting your CPU through the proverbial grinder, 3DMARK06 seemed a logical place to begin comparing these Core 2 processors. The main focus of these tests is to compare an overclocked E6600 to the X6800, but I've included numbers for the stock E6600 in all of my results to reiterate the performance gained with a simple front side bus tweak. See Figure 7. The E6600 running at 3GHz outperformed the X6800, though the difference is minimal in every category. 3DMARK bragging rights are all well and good, but let's take a look at performance in two popular games and find out whether or not the extra money spent on a Core 2 Extreme translates to extra performance.
On any given afternoon, Counter-Strike: Source has tens of thousands of players online. I don't think it's unreasonable to assume that at least some of those people are contemplating a CPU upgrade, so I'll help them decide. The video stress test was run using the same settings as before. Figure 8 illustrates the benefits of a faster processor at low levels of detail. However, with settings maxed out, the graphics card determines the frame rate. Once again, the overclocked E6600 is neck-and-neck with the X6800.
Being one of the few games known to take significant advantage of multi-core processors, I decided to run a few extra tests in Supreme Commander to find out what sort of advantage a dual-core processor offers. The 680i LT BIOS allows you to disable extra cores, though I cannot fathom why you would want to. That is, unless you were testing the performance difference. The built-in performance test was run using the low, medium, and high graphics presets at 1280x1024. As you can see in Figure 9, each of the three CPU configurations performed almost equally. With such slight variations between them, the difference is unnoticeable in-game. On top of that, the manner in which the average FPS is calculated seems to be limited by the video card. Nonetheless, a second core provides a substantial advantage of about 15 percent on high settings.
SuperPi and SANDRA
Synthetic benchmarks that stress the number-crunching power of a processor are a quick and painless form of measuring performance. While the results obtained may not be representative of real-world performance, they are easily compared and thus popular with the enthusiast crowd.
SuperPi times how long it takes for your CPU to calculate pi to the specified number of decimal points. Common comparison figures range in the millions, so that is what was tested. See Figure 10 for the results. The discrepancy between the overclocked E6600 and the X6800 is never much more than a second.
Developed by SiSoftware, the System ANalyzer, Diagnostic and Reporting Assitant not only tells you everything you ever wanted to know about your computer's hardware, but features several benchmarks. Figure 11 showcases the results of two arithmetic tests called Dhrystone and Whetstone, which measure in Million Instructions Per Second and Mega Floating-point Operations Per Second, respectively. The multi-media test measures how many iterations of instructions, such as MMX and SSE, the CPU can perform per second. See Figure 12. Much like the SuperPi results, these seem to almost scale linearly to the frequency of the processor.
» AN ALTERNATE COOLING METHOD
The retail heatsink and fan for the X6800 is certainly adequate for normal use. In fact, it is most probably more than enough for modest overclocking. However, it can get loud! If you want to reduce the noise level and/or the operating temperature, you will need something better. To find out if the Thermaltake Blue Orb II fits the bill, idle and load temperatures of the X6800 at stock and slightly overclocked speeds were measured with EVEREST. Idle is easy; The temperature was taken after the computer was left alone for about fifteen minutes. "Heavy load" meant the Orthos Prime stress test had been running for at least twenty minutes. Check Figure 13 for the results.
The temperature wasn't lowered by a whole lot, at least in this particular setup. It's hard to complain when the fan is whisper quiet, though. On full blast, the retail cooler sounds like one of those leaf blowers they put on video cards these days.
Now that we've seen how the parts perform, I'm sure you're dying to know what I think! In the exciting conclusion of this masterpiece, find out if the new setup would be as good if it weren't free. Don't miss it!
|7 User Comment(s) • 5 root comment(s)|
| Trogdor (39) May 05, 2007 - 12:56 am | Edited on May 05, 2007 - 01:00 pm|
|» Back with more comments...|
Okay, just going to randomly comment as I read this time. Then I'll figure out a score. LOL
Call of Juarez: is it any good? I realize this isn't really part of the article, but I have to ask as I haven't played the game. Readers always seem to want benchmarks with more games, but it's impossible to benchmark everything out there so most sites stick to a selection of the more popular games/engines. Does Call of Juarez use a proprietary game engine? My own personal opinion is that if the game isn't very fun to play there's no point in benchmarking it. A good example of this is Dark Messiah of Might and Magic -- FiringSquad uses that game, and while I actually did beat it I don't think I would give the game more than a 60% score (three stars out of five), so you might as well stick with the more reasonable Half-Life 2: Episode One. /tangent
The choice of resolution obviously has a large impact on how well most games run. Since I tend to use very large monitors and very high resolutions, I was actually a bit surprised at how well the 7900 GS performed at 1280x1024. It tends to be completely inadequate for 1680x1050 and above in most recent games (unless you turn down the detail settings). If you have access to higher resolutions that might be another area to examine.
Counter-Strike: Source: does GeForce 7 hardware actually support 6xAA? I thought that was specific to ATI hardware and GeForce 8 series cards. I also think it tends to be overkill and running at 4xAA is probably more than sufficient. 6xAA might also account for the poor performance of the 7900 GS at the high detail settings... although the card may simply be running out of memory bandwidth.
Given the cards you're testing, I almost think the comparison of different detail settings is pointless. At the very least, I don't think the low detail setting is necessary. I think the information in figure 5 is far more useful. It's too bad that you didn't have the 256MB aversion of the X1900 XT, or the X1950 Pro, as both of those cards compete better with the 7900 GS on price.
On your overclocked 7900 GS performance, you probably could have shown these results along with the stock and X1900 cards in games as well as 3DMark06. (3DMark = garbage, even if it is extremely easy to run. LOL)
Perhaps more telling about your overclocking results with the processor is that the 25% processor speed increase does virtually nothing for your 3D performance. Obviously, you're totally bottlenecked by the GPU right now, which as I mentioned on the previous article was pretty much a foregone conclusion. In fact, even at low detail settings the 25% overclock only resulted in about 10% faster performance in Counter-Strike: Source, and at higher detail settings you could probably run an E4300 and still be GPU limited.
I guess the game benchmarks really just point out the problem I mentioned in your first article: the selection of components is really haphazard and not generally something useful for showing true performance potential, unless you are simply looking at the performance potential of the graphics card. I would have liked to see your single core versus dual core Supreme Commander results with a faster GPU. Right now, your overclocked E6600 is only 7% faster in dual core mode, so it seems likely that with a bit more overclocking dual core wouldn't help at all in that game... unless you add in a faster GPU that is. This is more or less FiringSquad's fault and not yours; they should have sent most of you guys a much better GPU in my opinion. But I know how it goes -- it took a hell of a lot of work for me to get and 8800 GTX for my own use! :-)
Have I ever mentioned that I hate synthetic benchmarks? SuperPI and SiSoft Sandra are almost completely useless in my opinion. You state: "While the results obtained may not be representative of real-world performance, they are easily compared and thus popular with the enthusiast crowd." I could have phrased this differently: "Enthusiasts that compare SuperPI and SiSoft Sandra scores are a bunch of retards!" (Well, maybe not *quite* that harsh.) It's all just a silly pissing contest, just like most of the 3DMark results. Not that it's terrible to include them, but I still think they're pointless except as stress tests.
The results of your Thermaltake Blue Orb are hard to put into context without actual noise values. However, I'm a bit surprised that the results under overclocked settings were so close between the retail and Blue Orb coolers. Considering the price, I would say there's a good chance that the Thermaltake Blue Orb isn't even worth considering unless getting a quieter and cheaper CPU cooler is your primary concern. For overclocking, I think the Thermalright Ultra-120 (eXtreme) is probably the way to go in nearly all cases. Sure, it costs twice as much, but you have to look at it in terms of total system cost as well as CPU temperatures. If I'm going to spend $30 on a better HSF, I might as well go all the way and spend $60 for significantly better cooling performance.
Finally, I realize this is part two of three, but the article just sort of runs into a dead-end. I guess the hard part about all of this is that you guys only have a certain amount of hardware and you don't have a backlog of test results, but as it currently stands I find most of the information provided here is a bit outdated or at least not particularly useful. I can already imagine the conclusion: a lot of this hardware, while fast, doesn't make sense for the "hardcore gamer". Beam us a faster GPU, Scotty!
The organization and flow of this article was also somewhat lacking in my opinion. There's useful information provided, but I think you could put together a better introduction and conclusion for part two. Spend less time talking about the benchmarks and more time analyzing the results and helping the readers to understand whether or not the hardware you're testing is worth purchasing. Or maybe that's supposed to be part 3 -- I'm not quite sure what exactly is supposed to be in part to. If part two is supposed to focus on performance of the various parts, I think you've mostly covered the CPU, GPU, and cooler at this point. There still hasn't been much discussion of the motherboard or power supply. As a performance article, I think you could have trimmed down some of the extra information and added a couple more benchmarks looking at the other components.
Still, good job for the most part. I'm being very stingy with my scores, so don't take it too hard that you didn't get a perfect score from me. After all, I have to leave room for someone to come along and top your score. :-)
» Login to reply to this
| jacobvandy (1636) May 05, 2007 - 02:52 am|
|call of juarez is a pretty decent game. i haven't played it in a long while, but i chose it to bench mostly because it's really demanding. it uses the chrome engine, made by techland.|
about the 6xAA with the 7900, you're right, it doesn't support higher than 4x. i took those screens on the machine with the x1900 and it totally slipped my mind. i'll amend the image ASAP.
the ultra 120 is what i would buy, but if i ever get those replacement water-cooling parts for the kandalf, i may not need one. ;P
but yes, part 3 will tie everything together. i've got 45 hours... let's hope it's good, yeah?
» Login to reply to this
| Trogdor (39) May 05, 2007 - 01:02 pm | Edited on May 05, 2007 - 01:20 pm|
|» 45 Hours?|
Whoa... you have to have the last part up in 45 hours? If you don't have the water cooling stuff (I forgot that was part of the case), that's going to make it difficult. Good luck! You should visit the forums here as well. :p
» Login to reply to this
» Note: You need to be logged in to write a comment!Login here, or if you don't have an account with FiringSquad, register here, it's FREE!
My Media-Blog categories
No categories created yet.