Sweet Website!

[vacuumtube]

Alright so I'm new to Techgage but I'm very familiar with technology/hardware reviews. I have been reading articles after articles especially related to processor and their affect on Gaming performance. I'm regular reader on various other websites (anandtech, hardocp, tomshardware etc) and I finally found what I was looking for and TechGage's recent Skulltrail review did excellent job in comparing Intel 9775 series processors to midrange CPUs especially in games like Crysis.

Thus, I'm highly impressed by the quality of articles. I have a Core 2 Duo E4500 and I have been monitoring my processor usage while running Crysis on Very High at 1024 and figured that my processor usage never goes beyond 80% at most (with work load shifting from 1 core to another from time to time) therefore I wondered how extreme edition processors were able to give 10 to 15 fps more as claimed by other websites. But thanks to TechGage, I now have my answer that Crysis is not CPU depending whereas other website claim the same but their results do not concur with what they say (look below for references for such cases).

However I believe Techgage is correct on this regards and my results concur to techgage's findings. I'm hoping to see/support more quality reviews like this from Techgage. Good job Rob W.

References for Contradiction and ambigious conclusions from other sites:
http://www.anandtech.com/showdoc.aspx?i=3271&p=4
and search for this : "This indicates that the higher the graphical quality, the MORE CPU bound we are." <-- huh? ("graphical quality" = poor grammar and it should be more GPU bound we are) -- correct me if i'm wrong.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3272&p=12 (scroll down to Crysis section, look at the benchmark results and look at the paragraph right after the chart...Contradiction. At those frames, difference between 42 to 50 fps is quite noticeable. AMD Phennom giving 42.4 fps compared with Core Duo E8200 giving 63.2 fps. Umm is that not a big difference? Very strange. In addition, look at Test Setup, why two different memory modules being used? 1066 vs 800...The difference in fps could be due to memory as well).

 

[Merlin]

Welcome to Techgage.......
I also run Crysis on very high using 2 8800GT in SLI with an E8400 Core2 Duo

:techgage::techgage: Merlin :techgage::techgage:

 

[Kougar]

As with most games, they tend to vary between being CPU dependant and GPU dependant according to the level... it fluctuates.

For the second link, Anandtech used the CPU dependant benchmark for Crysis... not the GPU dependant benchmark. The Crysis benching utility offers both options.

As far as the first link goes, it helps to take into account their explanation:

This indicates that the higher the graphical quality, the MORE CPU bound we are. Crazy isn’t it? It's counter-intuitive, but pure fact. In speaking with NVIDIA about this (they have helped us a lot in understanding some of our issues here), the belief is that more accurate and higher quality physics at higher graphical quality settings is what causes this overhead. Also, keep in mind that we are testing in a timedemo with AI disabled.

Basically I took from it that Crysis can be a "platform dependant" game... just about any component that isn't as fast as the rest of the system hardware can turn into a bottleneck. Their Skulltrail verses 790i platform results pretty well illustrated this, at least to me. It's just a bloody mess though, until Nehalem+GT200 get here in Q4.

To answer your memory question, keep in mind Anandtech is testing multiple platforms... Skulltrail uses FB-DIMM memory, and 800MHz is the fastest FB-DIMMs available. 1067MHz is standard DDR2 memory. Again what I said about being platform dependant... FB-DIMMs are high latency memory, designed for capacity at the expense of performance. So you would be correct, it contributed to the difference in results between Skulltrail and 780/790 Nvidia hardware.

 

[vacuumtube]

so let me get this straight. As far as CPUs go (if the rest of the hardware remains the same), I'm going to get same fps in Crysis if I have a Q6600 versus a Extreme Edition Q9770 correct? Afterall, anandtech NEVER did a skulltrail review vs a Q6600 to give us a better understanding that a Q6600 will perform just as good as a skulltrail which costs 5 times as much.

 

[Rob Williams]

Hi vacuumtube,

Welcome to the forums and thanks for the nice comments!

According to my previous findings, your assumptions are correct - there is no discernible difference in Crysis regardless of what CPU you are using. That game is hardcore GPU-bound, and I think we laid it out pretty clear that it doesn't matter which CPU you use... it's all about the graphics card.

Many sites, like AnandTech, use timedemos for everything (which stresses the CPU more than the GPU), while we manually perform our benchmarks. We actually sit in front of the PC and play the same level over and over in order to deliver real-world results. This process takes a lot longer, but in my opinion, it delivers more realistic results to our readers.

I don't quite understand sites that fore go real-world gameplay in lieu of timedemos. Like you mentioned, our Crysis results show identical performance between all of our processors, while sites that run timedemos show results far different. To me, that's a disservice to their readers, because even though a low-end CPU can run the game fine, a timedemo says otherwise. If these sites ran Crysis through manual benchmarks like we did, I am confident their findings would be identical to ours.

I believe timedemos have a use, but not in GPU reviews. In CPU reviews, they make sense, since they are CPU bound, and essentially show how much better a high-end CPU would be over a low-end, if the game took full advantage.

That all said, I need to actually re-benchmark all of our CPUs (gah!) due to a motherboard issue with the board we've been using up to now. I am going to begin testing on the ASUS P5E3 Premium (X48) and will have the opportunity to re-run all of our benchmarks, including Crysis. Since the Skulltrail review, I've changed the level that we use, so I am looking forward to seeing if this new level delivers us differing results from what we saw in the first level of the game (which we used prior).

 

[vacuumtube]

Since the Skulltrail review, I've changed the level that we use, so I am looking forward to seeing if this new level delivers us differing results from what we saw in the first level of the game (which we used prior).

Oh that'd be sweet. Wish I could help ya write or edit those reviews.

And thanks for elaborating on the timedemos. My purpose for this post is to prevent users from falling into the traps of over-hyped marketing of high end systems and being confident in their purchase of midrange systems without a notion of thought that high-end would have given them twice the frames in crysis (when in reality it doesn't). Thus I could only do so by critiquing anandtech's recent articles. However I'm not alone in this, there have been numerous negative comments at their websites where users are not satisfied with the quality of the article/setup used as well as their testing methodologies.

"The Test" page at anandtech only mentions the system configuration unlike your "Test Methodology" page which includes more detailed explanation and as well as the screenshots of settings used. I think the difference in quality and effort put into the articles is crystal clear. I've been reading anandtech for over 2 years now and will continue to read but I'm dissatisfied with their recent articles and more than satisfied with techgage.

Thanks.

 

[Rob Williams]

Where abouts in Canada do you live? We don't get enough Canucks in here.

As much as I love AnandTech (I've read the site for years), I have to also disgree with their methodologies. For a professional website, they just don't seem that fair, or realistic. For Call of Duty 4, their "benchmark" consists of recording the avg FPS from the opening cutscene... that just doesn't make much sense, in my opinion.

In all fairness, though, AnandTech is far from being the only site to have similar methodologies. HardOCP and PC Perspective are the only other sites I know of that perform all manual benchmarking like we do.

We take some pride in how we lay out our reviews, so I'm glad someone notices ;-) Glad to hear we are doing a few things right.

 

[vacuumtube]

Central Canada - Manitoba

 

[Kougar]

I will confess as much as I like Anandtech, some of their articles are confusing and others I just don't agree with, such as their desire to have Skulltrail used for all their GPU testing because it offers SLI + Crossfire.

Gotta say that I really enjoy their in-depth and detailed coverage of BIOS options and what they do, along with much of their work on Intel processors and overclocking. Much of the info just can't be found anywhere else, or is considered "near rumor" such as the truth about vdroop.

 

[Rob Williams]

Gotta say that I really enjoy their in-depth and detailed coverage of BIOS options and what they do, along with much of their work on Intel processors and overclocking. Much of the info just can't be found anywhere else, or is considered "near rumor" such as the truth about vdroop.

I can agree with that.