Response from FS to HoCP "real world" tests.

[coldpower27]

Originally posted by: DLeRium

Originally posted by: nyker96
I think real world does have its values, but bottlenecking the system with a GPU isn't benchmarking. But I also have to say I do agree with HCop, we do need some real world benches to put some sense into this "20%" faster, 50% better cliams. As far as I know, at least you need to say 50% faster if xxx.

In real world all that 20%//50% might or might not manifest, so reviewers should make it clear that you need say a X1900+ g-card to see a 20%+ different at xxx resolution.

Bottlenecking a system with a GPU when there are ways to not bottleneck it is just WRONG.

If you're buying a 7900GT, I don't think it's that much of a stretch to get SLI at that point. You might as well bench wtih a freaking 7600 or 6600GT if you're going to GPU limit it (basically for mid-range buyers).

Real world has benefits, but honestly, if we didnt have fanboiism, I swear we wouldn't really be seeing so many real world benchmarks.

Originally posted by: dexvx
LoL

Where were all the AMD fanbois when the Pentium-D 920 could match an FX-60 due to the GPU bottleneck? All you fanbois saying that the P-D 920 = FX-60?

Moreover, HardOCP benchmarks are fuxxed up. The AMD Test systems used a "lighter"configuration (in the config Kyle noted that the AMD system had dynamic shadows turned off). So much for that.

Like I said, if we flipped it the way around and Opterons really killed Conroes the same way Conroes kill Opterons today, what would you say if we showed Opteron benchmarks being GPU limited? You would hear millions scream. Just because Intel sucked yesterday doesn't mean you give them this huge bottleneck so their victory is reduced simply to appease AMD fanboys.

Actually you don't even need to use a theoretical sample like that, if you used an Athlon FX-60 vs a Pentium D 920 in such a manner, you will likely get the same occurance.

 

[deeznuts]

Originally posted by: classy
What a weasle response. When I buy any component, I am buying based on what impact it will have on my system. If I used your thought process, 99% of all people shouldn't read anything from a hardware site then, because 99% of the people can't afford the system configurations used to test hardware with at most of the sites anyway.

Here is the reality in the whole Conroe aftermath, if your building for gaming the video card is the most important piece and any of these processors on the market can easily provide plenty of power to have top notch gaming. Conroe is the best cpu for both gaming and its encoding and rendering power is incredible. So if do a lot of desktop stuff as well then go Conroe. Conroe is the better cpu, but for right now AMD has the better platform. AM2 is boards are cheaper, fast, and solid. Conroe doesn't have that yet. There is really only one board for Conroe and that will set you back $250.

You start off with flame, then go on to post bs? AMD has the better platform for AM2, why? Because it's slower? Cheaper? There are several boards for Conroe. From less then $100 to this:
Gigabyte G965Out of stock right now but is on sale elsewhere, and was on sale at newegg. Either you are an amd zealot or reading weeks old material.

 

[classy]

Originally posted by: deeznuts

Originally posted by: classy
What a weasle response. When I buy any component, I am buying based on what impact it will have on my system. If I used your thought process, 99% of all people shouldn't read anything from a hardware site then, because 99% of the people can't afford the system configurations used to test hardware with at most of the sites anyway.

Here is the reality in the whole Conroe aftermath, if your building for gaming the video card is the most important piece and any of these processors on the market can easily provide plenty of power to have top notch gaming. Conroe is the best cpu for both gaming and its encoding and rendering power is incredible. So if do a lot of desktop stuff as well then go Conroe. Conroe is the better cpu, but for right now AMD has the better platform. AM2 is boards are cheaper, fast, and solid. Conroe doesn't have that yet. There is really only one board for Conroe and that will set you back $250.

You start off with flame, then go on to post bs? AMD has the better platform for AM2, why? Because it's slower? Cheaper? There are several boards for Conroe. From less then $100 to this:
Gigabyte G965Out of stock right now but is on sale elsewhere, and was on sale at newegg. Either you are an amd zealot or reading weeks old material.

Why people like you have to get on the net and act like an @ss. I didn't flame anyone, I stated I liked what Kyle did. And right now AMD is the better platform because you have so many top notch boards available. You can go crossfire or sli and if the cpus are that cheap you can trick out a AMD system bigtime for a decent price. There are no decent cheap boards for Conroe out yet, you idiot. And nvidia boards won't be out for several more weeks. Conroe is a top flight chip, maybe one of the top pieces of hardware that we have seen in a long time. But, the best boards are expensive. And as for a zealot, I have not had an AMD machine ever as my main box and still right now use a P4. We are still in the waiting game till everything settles in with Conroe, so give it a rest, so sit down, eat some pudding, and shut the bleep up.

 

[imported_seferio]

I think HardOCP provided a different conclusion than most, and are drawing unnecessary flack for it. Conroe is extremely powerful (and quite ahead of its competition), and really much more powerful than the performance any single graphics card can provide. You need at least SLI to really make the processor shine. With Core 2 Duo prices being relatively low (~300 or less as long as you're not going for the XE6800), you can definitely expect a lot of people with C2D systems to only be running a single graphics card. Hence for the majority of C2D users, AM2/K8 will provide similar performance.

However, I agree fully that their elitist attitude is hardly excusable. They approach new hardware differently, but obviously their reviewing methodology is far from flawless.

 

[skooma]

Well, this is alot about nothing.

I'm kinda surprised at Classy though. I thought you'd be one to know that when you're testing different components, you remove the other bottlenecks I'm not an idiot needing { H } to interpret the results or predict their impact on my system.

 

[AyashiKaibutsu]

If he found the CPUs (1 for AMD and 1 for intel) from which improving the CPU didn't improve the games with a wide array of video cards on a wide array of settings(everyone plays at different settings) it'd be an awesome article... still wouldn't be a CPU review though. As it stands it's a pitiful article. Most CPU reviews include high res settings that are bottlenecked, but they give you other settings too. That way if you get a more powerful videocard than the review has you can make decisions based on how each CPU holds up when not bottlenecked.

 

[imported_seferio]

Well I think what Kyle was trying to get at for his review is, if you are going to get a Core 2 Duo, are you going to get a Crossfire X1900XTX / SLI 7900GTX setup or a single high end card like the X1900XTX / 7900GTX? I think the answer to that is "probably not" for a lot of the people buying C2D. If someone can prove that a majority (66% or more) of people buying new computers (either OEM, Dell, Compaq, HP, eMachines, self-built, etc.) are going for SLI systems then that is a clear sign that indeed Kyle was way at fault for testing with such an underpowered system. However, if no one can prove that SLI dominates new systems then obviously his benchmarks do indeed show the performance people can expect with their new C2D machines.

 

[Maethor]

What I find so interesting about the real world testing is that they only did it with 24inch and greater monitor resolutions which is a minority of even the gaming crowd. I mean seriously this is less than 2% of the gaming crowd. Personally I do not like the quality of the larger screens they tested as many have color problems and to get a decent one you are concidering dropping over 800 dollars for a monitor.

Most of us game on either 21 inch or below monitors and in those cases we do see the large performance increases because it becomes more cpu bottlenecked than a GPU bottleneck. Also many games are more CPU bottlenecked than GPU bottlenecked such as many mmorpgs and RTS games.

Added
Also most people buying a high end system would opt for the 7950 GX2 for only an extra 100$ that offers SLI performance on a non SLI board.

Those like me who like the 1600 by 1200 and lower resolutions will see more performance but when you are talking about the difference between 80 and 100 FPS your talking about something thats indistinguishable to the human eye. Personally I do not feel the need to have to have everything in a game maxxed so I mainly concider price vs performance and what I need.

I am not a fanboy and have owned AMD when they performed better and owned intel when they were better. Personally my system is in bad need of an upgrade so I will be choosing C2D as going to a AM2 will cost the same for me so honestly its no comparison but for anyone who has an AM2 system and doesnt plan on upgrading a video card in the next year should not upgrade. When the next gen of video cards though we will see a removal of the bottleneck in games so those waiting on those who dont need the performance should wait.

As a general rule if you dont have a x2 or AM2 platform I would go with the C2D if you already have AM2 stick with AMD for now since unless you are running a high end SLI or crossfire system you wont see a big performance gain and even with SLI the performance is only about 5-10%

 

[lopri]

Originally posted by: wicka
Basically, Kyle shows one good point (that CPU has very little effect on high-end gaming). The sad thing is: we all know this. H doesn't appeal to us. H appeals to idiots who think they're cool because they read it every day (and now I'm NOT refering to people above who said they like H). As long as they get their advertising revenue, they couldn't care less. And by bashing every other review site, they've taken their whole audience of n00bs and made them feel "1337," and thus far more loyal.

Couldn't agree more. I myself haven't been there.. like 2 years. (except occasional visits every other months for their Hot Deals / For Sale forums.) I mean, not that I have anything against the management, but I really don't get much from that site, technology-wise.

 

[dexvx]

HL2 Steam hardware survey:

800 x 600 ------ 3.40 %
1024 x 768 ----- 47.55 %
1152 x 864 ----- 7.84 %
1280 x 1024 ---- 36.11 %
1440 x 900 ----- 0.83 %
1600 x 1200 --- 2.05 %
1680 x 1050 --- 0.79 %
Other ----------- 1.43 %

Look at theat a whopping 4% use 1600x1200 or above.

I guess 1280x1024 isnt real-world enough for 96% of the other people.

 

[wilki24]

Wow, Kyle is a real moron, isn't he?

He keeps going on about "real world testing" as if the underlying point (games are GPU bound at higher resolutions) isn't something that the vast majority of people who read hardware websites already know. Thanks, Captain Obvious, without your wise words we sheep would be blinded by "canned benchmarks" and we might have to use multiple sources to make our decision!

The amount of condescention in both the original article and the rebuttal pretty much demonstrate that they were written by a total blowhard. The amount of ass-kissing on their forum is amazing.

 

[Maethor]

Originally posted by: dexvx
HL2 Steam hardware survey:

800 x 600 ------ 3.40 %
1024 x 768 ----- 47.55 %
1152 x 864 ----- 7.84 %
1280 x 1024 ---- 36.11 %
1440 x 900 ----- 0.83 %
1600 x 1200 --- 2.05 %
1680 x 1050 --- 0.79 %
Other ----------- 1.43 %

Look at theat a whopping 4% use 1600x1200 or above.

I guess 1280x1024 isnt real-world enough for 96% of the other people.

See this proves my point exactly H is perhaps one of the biggest biased when it comes to doing Reviews. Now those of us using something bigger than 1600 by 1200 know that games are by and large GPU bound in the most recent releases. Furthermore anyone who is using a monitor with a resolution above 1600 by 1200 has likely got the latest and greatest SLI or crossfire system and would see improvement from a new processor

 

[imported_aashburns]

I believe each website has their own unique perspective to reviewing computer products. Obviously, no one method is perfect and this why there are so many sites out there and why I read several sources to get a complete picture. With this in mind, I do not think it is HardOCP's approach that is insulting, but rather their professionalism. When a review is conducted, the author must show that the article is not biased and credible. HardOCP has personally insulted other websites and professionals in their industry by questioning the professionalism of others in their article. Furthermore, he insults the efforts by people who create benchmarks to provide an accurate representation to compare hardware. HardOCP (Kyle) needs to wake up and understand that one gains the respect of the community by the quality of what they do not by simply being controversial. By this I do not imply that Kyle lacks quality in his work, but rather that the unnecessary additional of controversial content to his article takes away from the quality of his review and damages his credibility. Kyle, why do you ruin the actual purpose of what you are trying to accomplish by throwing insults? When you do this I consider what you write an editorial and not a review. Therefore, it is hard for people to take you seriously as a professional. You have been in this business too long to act so immature. People tend to respect quality websites better and hold higher standards for them. I believe many readers expect that here at Anandtech, as well as at many other sites including HardOCP.

Kyle,

YOU MUST EXPLAIN to everyone IN DETAIL about your review methodology FOR EACH REVIEW. The concept of benchmarks is to represent the performance of a certain aspect of a computer accurately as possible and allow results of subsequent runs to be comparable within a strict margin of error. In reality, benchmarks are not perfect and their inadequacy is unacceptable for you. When you do your review, you are simply giving your wise two-cents about particular product. But what is lacking is the consistency. You do not list your margin of error for your testing in each case. You do not list how each benchmark is conducted so that users can try to replicate them. You do not encourage enough user feedback about your testing parameters to revise them over time to better represent others. The review style which you prefer requires that you explain yourself more, because you cannot assume everyone thinks you are God and are perfect. Therefore, you must give more detail to convince the person reading your review that your review methodology is trustworthy. Benchmarks do this by their wide-adoption among reviewers and wide use among enthusiasts, but you do not have this advantage. Furthermore, this will clear up the confusion of people thinking your opinion and experience with the product is a benchmark simply because you list numbers in your testing. I believe that just because you list numbers does not automatically make what you say credible ? I expect a better review than that from you. AND I DON?T WANT TO HEAR THE ATTITUDE THAT WELL I?M SORRY YOU DON?T AGREE BECAUSE MANY OTHER PEOPLE DO (I have seen you write this to several people). That is not the attitude of a website trying to be the best computer review enthusiast on the web that is a website that only does enough to get by and is contrary to the reason why you originally decided to change your approach to testing. If you had done this in the first place, there wouldn?t be any threads about your article being crappy or unfair and you would spend your time on follow-up articles and not explaining yourself for past publications.

 

[Maethor]

Something I would like to point out is that in most of the benchmarks kyle purposly only using a single non SLI system to cap out the GPU. I would say that even at the resolutions testing SLI would have made a large difference

 

[imported_seferio]

Originally posted by: Maethor
As a general rule if you dont have a x2 or AM2 platform I would go with the C2D if you already have AM2 stick with AMD for now since unless you are running a high end SLI or crossfire system you wont see a big performance gain and even with SLI the performance is only about 5-10%

I think Maethor put it the best. X2, AM2, fast/overclocked Pentium D will all be fine for the gaming crowd now as long as you aren't running SLI.

However, I do notice now that in the HardOCP benchmarks, the apples to apples comparison runs all the benchmarks at 1280x1024 (17" 19" LCDs or CRTs) which is the second most popular resolution according to the steam survey. 1024x768 would most likely show an even greater advantage for C2D, but if you're running resolutions that low even with a 7900GTX or X1900XTX you would most likely turn on AA/AF and the performance gap will shrink.

Edit: Also the point of using a single card is that most people who will be buying C2D machines won't run SLI (Dell will probably sell 1000x more C2D machines than people will be building on their own, and I doubt more than 25% of those will be SLI even). If they do, then definitely ignore his review (which should be obvious considering the test system isn't remotely specced like your machine) and read someone else's (i.e. Anandtech) review.

 

[will889]

Originally posted by: dexvx
HL2 Steam hardware survey:

800 x 600 ------ 3.40 %
1024 x 768 ----- 47.55 %
1152 x 864 ----- 7.84 %
1280 x 1024 ---- 36.11 %
1440 x 900 ----- 0.83 %
1600 x 1200 --- 2.05 %
1680 x 1050 --- 0.79 %
Other ----------- 1.43 %

Look at theat a whopping 4% use 1600x1200 or above.

I guess 1280x1024 isnt real-world enough for 96% of the other people.

Thats about right. Most gamers still run CRT's but that gap is closing fast. You can't buy CRT's easily anymore - LCD's are what's out anymore and thay have a fixed res. People with budget GPU's (up to 7600GT for nVIDA and X1800GTO for ATI) really benefit from being 'able' to scale resolutions down in game like Oblivion with CRT to 1024x768 or 1152x864-1280x960 to get better frame rates. - VS a gamer that has a fixed 1280x1024-1600x1200 res @ 4:3, or a 20" WS res that nearly has to use at least a 7900T and/or above to get decent playable frame rates in games like Obvivion-FEAR. I guarantee you this - you hear all the time people make the blanket statement, especially @ hardocp that no one games at 1024x768 anymore (which is true only if you have a 19" or above LCD) - where they can't scale resolutions - but if they could they WOULD scale back to 1024x768. If LCD techlonogy actually allowed scaling we undoubtably would see average resolution averages lower. Why? Because -

1. it would aloow said user to not have to upgrade to such an expensive GPU

2. it would allow said gamer to also possibly keep using his/her current GPU and go down to a more playable resolution.

 

[thilanliyan]

Originally posted by: will889
If LCD techlonogy actually allowed scaling we undoubtably would see average resolution averages lower. Why? Because -

LCDs don't allow scaling or do you just mean scaling up?? I can scale my LG 17inch LCD DOWN, but not up.

 

[imported_RedStar]

"H appeals to idiots who think they're cool because they read it every day (and now I'm NOT refering to people above who said they like H)..."

Now you are making blanket statements like .
If you had actually bothered to read the forum response to Kyle's conroe gaming performance and squid pieces...you'd know that the majority of posters have taken issue...

 

[dexvx]

Even for people that run 1600x1200 or whatever, CPU performance is important. Why?

When people buy the newest and best graphics card, they dont want their CPU to be the bottleneck. Right now a Pentium-D 920 may offer about 90% of the FPS of a E6600 at 1600x1200 FSAA, but once you upgrade to the next gen of graphics cards, that amount may drop to 60% as its less of a CPU bottleneck. I would sure be pissed (if I didnt know) and based my decision on HardOCP and bought an FX-62, only to find out 6 months down the line that an upgrade to R600 would only give 3/4 the performance of a X6800. Not to mention a lack of non-FPS games (I think WoW was their only way, but even then, the "real-life" performance of WoW using their currrent test methodology was off - yes lets run around aimlessly in the world instead of doing a live 40-man raid).


Moreover, lets get into the economics of things. Anyone who serious games on the PC knows that GFX is the way to dump money (as opposed to an extra 200Mhz on your CPU). People would drop $600+ for an SLI/Xfire setup faster than dropping $600+ for a CPU. IMO, its assumed that anyone with a $1000 CPU should have a GFX setup equivalent to $1000.


IMO, the review was either incredibly naive or skewed. No one with a straight face can recommend going with a high-end AM2 over a Conroe setup. Not only is performance/price worse, but also performance/watt, overclocking AND longevity.

 

[jones377]

Am I the only one thinking that Kyle wrote his review in this manner EXACTLY because he would get a response like this? I mean if he showed Conroe beating A64X2 by 10-30% like all other dozens of websites did, no one would have paid much attention to his review more than anyone else posted. This way he gets LOTS of attention.

An angry Intel zealot clicking on his page to pick apart his review is just as much ad revenue as someone looking for genuine info....