GPU Reviews: What do you want to see

FarFarAway

New member
Hi guys

Well I've been merrily getting on doing GPU reviews for some time now but I wanted to pause and ask you lot what I could do to make them better.

Now what I'm looking for is something along the lines of suggestions on what to add like game benchmarks etc. Remember I can't do everything ever but I can try to add some bits and pieces here and there

One thing that's in the pipeline is measuring power consumption, just awaiting some parts.

Soooo...over to you...
 
In the recent light of the 8800 cards that generate a huge amount of heat, would it be feasible to touch on how effective the stock cooler is at actually expelling heat out of a case?

My 8800gtx has a good quiet cooler, but it just doesn't get the heat out of the case effectively
 
Hmm let me think how I could do that.

I could place a thermal proce inside the case and one directly on the card exhaust?? I need to get a laser thermometer tbh
 
When I take a look at the OC3D reviews, GPU-wize, I`m not often left thinking "what about this..". There are always the odd mundane questions that crop up now and again, but they pretty much always get covered in the follow-up forum thread.

What I feel is important to OC3D and it`s members, and future members it may target are:

  • Packaging

    The site recognizes that part of the experience of ordering and receiving all goods for pcs is that moment u have the `box` infront of u, and what greets u as u open it. Personal pc builders, love this and from a practical pov it shows u how well the item is cared for during it`s transit and how it appears from it`s on-shelf appeal.
  • Contents

    It`s a notorious craving that the packaging, once opened and observed, doesn`t merely contain the item plus flimsy book or merely a cd. An item of varying quality, in this case the new GPU, is arguably the 3rd most important item a pc enthuse will purchase (arguably it can be the 1st for some) and it`s that much more gratifying that u don`t simply stick the card in the pc, sling the cd (cos u download drivers) and the balance is an empty box. That`s almost an OEM delivery from my pov. Important things for a GPU box in addition to an OEM package should be - a) a driver cd, doesn`t have to and won`t be the latest, but if u`r stuct in a build u can`t download. b) a benchmark cd to show they`re keen for u to test it. c) a game - doesn`t matter what it is in terms of genre, but something old or new to test it further. d) (optional) an app. creative or entertaining. e) if there`s a requirement for a cable - it has to be there. Matters not that the tester has a box of 10 of them. f) a book of the non-flimsy variety with basic quick-start and follow up item critical info. g) bonus stuff.

    All of this costs the manuf next to nothing and hence there`s no excuse, and it all lengthens the `new-item` feeling.
  • Dimensions

    Although widely regarded as close to standard sizes, it`s getting increasingly important when we consider htpc cases, the height, will my drive bay get in the way of the gtx - and is this heat-sink going to negate the use of one of my pci slots.
  • Application

    A basic trip from the box to the slot. Simple as that. It has been known for manufs to badly fit back-slot panels that make them a pita to put in without persuation. Again, did we have to hunt down a cable they never supplied us.
  • Asthetics

    All of the members of OC3D are proud of their builds. Cable-tidying aside (to me this is both a skill and an art), the member demands the gpu board, sink, back slot panel, to add to the overall picture of creation made inside their lovely case. A gpu card that`s put in and has a less than plain pcb and overall doesn`t advertize itself can help or hinder the overall effect. I say hinder as some manufs can go over the top, maybe with gawdyness or retro looks. I personally feel not enough attention is made to this by the manufs - should we care if it breaks the 3dmark06 score at the end of the day ? I think just looking at the number of posts in the rig gallery will answer that. An additional petty consideration, are the caps all bent out of shape or raised to dangerous heights.
  • Advertized facts

    What the manuf says the card will do, and what the background of the pcb is. Technically this can get long winded. Although a member can be linked to the manufs representation of this as a whole, the basic clock speeds, mem sizes, pipes and other specific boasts are on-paper arguements for buying one card over another. Then of course we`ll find out if it lives up to it. Additionally anything made as a claim to buy this gpu over another.
  • Under-the-hood

    Water coolers take great interest in this. But increasingly perhaps passive cooling members would care to see how easy it was to remove the sink, and of course lets those who aren`t interested in that, what the thermal goo is like and if it`s applied well or something flimsy and cheap is being used. An insight as to a manuf using a stoopid screw hole arrangement, and have they made sure the caps aren`t causing immediate problems by their location. There are as well the use of top-plates, do they get in the way, will they remove easily. Positioning of ram and does the stock cooler actually cool them.
  • Simple oc`ing

    How much of a % increase u can get from stock. This both challenges the stock cooler`s performance and gives the custom cooler an idea of the potential to work from. I like %ages here. Temps and stablilty measured against the clock are important. These are consuming tasks tbh, but are very intertwined with the results of benchmarking and the contraversial views of power consumption.
  • Power consumption

    Just to get the contraversial bit out of the way. Power consumption of the gpu, interms of `green issues`, matters very little to pc enthusiasts. I only say this because u offer a G92 cutting edge gpu to a member and tack on the fact it uses alot of power - despite their recycling and outside `green stuff` - they won`t care. It is bleeding edge tech, I have it in my pc and I can play css @ 3 billion fps more than my last card. That`s just how people are in general. It is of course a basis to compare boasts of a manuf that might make claims to be `greener`, but as far as the member is concerned, they`d like to care, but I don`t think they do. Me included tbh.

    However, they do care about if the gpu consumes enough power to warrant a purchasing of a new psu. That is wholey important to them.

Aside from benchmarking, these are the things that are important from an OC3D`ers pov, imo of course.

Benchmarking is a difficult beast. Only because time moves on if u care about how 1 generation compares to another.

  • The likes of 3dmark05/6 are standards and will remain so afaic, so will 07, 08.
  • OC3D also has a strong relationship with fps games, one in particular, so the benching of it and it`s updated standard are a must. Think we all know by now that they aren`t the absolute in benchmarking, but the test do give u a standard to base in comparison to the next card as well as 100s of cards b4 it.
  • Next u have a mixed bag. I think it`s wholey important to bench a game that is heralded as "built for nvidia" or "built for ati" when it starts up. We`ve all seen these flashes at the start of games now. (well I haven`t seen, or don`t remember an ati one) but if not ingame maybe if it`s boasted somewhere else, during it`s dev or on the retail box or whatever. This can demonstrate that a card performs well if a game is written to take advantage of it. Kemp was very clever, imo, in sticking with the benching of Oblivion. I`ve trashed this game at every opportunity, cos imo it seemed very badly written - how much more mistaken can u be when u see it run on an Ati platform... I think this is a good example of this.
  • A top of the chart game. A non fps but non-the-less graphically challenging. C&C3 has been used recently and is an excellent choice. Perhaps this will move to World in Conflict soon. Problem is here, how to compare when u make a move.. I think to be fair u just have to make the move and bench it on it`s own, if this is the case. Next time around u can make a comparison. Same can be said for changing mobos in between ranges of testing.
  • A 1-off benchmark. A randomly well known game plucked out of the blue as a long standing favorite. If it`s been benched b4 maybe a comparison, but this would be mainly a speculative feature.

All these will require a stock test and the gpu under stress. Max/min/avg I`m not so sure. I think with so many stats being listed in this section, an average may be just enough. For sure the other 2 are very important stats, but with lots of benching, from my pov only, if it`s possible to keep the averages on one page as opposed to taking a whole page for one game just because of the inclusion of the min/max. It`s debateable I know. It could be just as a benchmarking conclusion.

Saying that, if it were left to me I`d have 1 graph, each card using a color, and each game fps represented with a point that is lined-up. This is only to have a pictoral of the whole benchmarking on one sheet. With the 3dmark tho, maybe a % muliplication of the score would have to take place to show it in fair standing with the fps scores of games. Maybe easier to the writer too, in that if they want to make a card comparison, u can just introduce it to the graph as a layer.

Taking this to a bigger scale, all OC3D reviews made from this point onwards could then be accessible from a drop-down list. That would allow the reader that maybe has no interest in one card, to maybe select 4 cards of their own from the list and they`d appear as an overlapping layer on the graph. Be almost like flipping over a transparancy irl. Of course a default selection would have to be there to reflect the writers` content.

It`s at this point that I can come to realize how much work actually goes into each review. U can take the best part of a day doing most of this, then of course u have to write-up and graphically arrange u`r posting. Ugh. There has to be an enjoyable aspect to it tho.

Being honest, I think the majority of this post already gets done, but I have included the odd thing that from my pov are important that may fine tune it for me perhaps. I haven`t mentioned photos, cos tbh I`d take photos all the way through the process, including the part where the test rig is described, for me this adds to the asthetics point.
 
Lost planet? See now I haven't included this as it's a badly written console port that doesn't fairly represent a PC game, that's just my opinion

Cheers for all that rast. I've read it and think I do most of it. Maybe I should take a pic or two of the setup used?

as for min/max average thing I include those as they are WHOLLY important when gaming. Hey a game could be zooming along at 600 FPS but then you get to an inside part and it drops to 10 FPS and you're buggered.

So hmmm not sure what else is missed. If you want to give me a list of what I miss off (power consumption aside, I am taking care of that)

Oh yeah and I'm getting a bit fed up of using CNC3, all the top end cards pwn it, mefinks I need something else
 
name='Kempez' said:
Oh yeah and I'm getting a bit fed up of using CNC3, all the top end cards pwn it, mefinks I need something else

Company of Heros? I swear you have used it before.. but now it has DX10.. you could compare between DX9 and DX10? ;)
 
name='Toxcity' said:
Company of Heros? I swear you have used it before.. but now it has DX10.. you could compare between DX9 and DX10? ;)

Hmm good thinking batman

I'll see if I can get my hands on it
 
Yeah I can see the min/max are imporant in that sense. Tbh I think a better form of averaging should take that into account. Then tho, maybe we`re getting into a mathematics too much when I guess displaying the two could be easier.

Perhaps a compremize would be something like a bar as u use atm, but basing it on a single bar that has a lighter shade from 0 to the min, then darker to the average, then darkest to the max.

A mathematical OC3D stat based on fps numbers would be clever.

Think I`m trying to make it simpler to keep more stats on less pages, I dunno.

C&C3 could be u`r speculative game, hehe, just pick another :) It has to be said, the Conflict game has a good benchmark thing on it - but I think u`ll find the high end cards will boss this also. I`d like u to use total war as a bencher at one point, purely for a selfish reason. The game is well written like C&C, and tbf the highend cards will pwn it too. Maybe u can take requests in a thread :p

I like the idea of being able to select graphs taken from each review for a member to make a comparison themselves, maybe outside of a review tho. U`d have to base this graph on something that would have to be constant for a very very long time tho.

EDIT: be a bit weary of patches that make games work or show features of dx10, they aren`t wholey representitive of the dx10 platform. I think people fall into the trap of thinking that these patches make the game a dx10 dev`d game and get too disappointed.
 
Back
Top