Zardon

Read The Review here
______________

Despite Nvidia having a strong product catalogue rumours have been rife in the last week or so about the future of their mid and high range products. It is business as usual in the budget range today however as we see the launch of a new card in the sub $79 price range. This isnt just a cut down version of a different card though; for once we have a budget model which brings something a little different to Nvidias product range.

The Captain

Interesting (little) card. I wish we could have seen some Physx benches, though, comparing it to some other inexpensive Physx-capable cards like the 9800 GT or 9600 GT. I'm looking for an inexpensive Nvidia card solely for Physx purposes.

Zardon

Yeah, would have been nice, however we have been testing, writing and designing three graphics card reviews in the last 5 days,. another one being published tomorrow. It is rather difficult to cover all angles in these time frames.

comp_ali

Thanks for the review , I think the performance will be so close to my Ati 4670HD and in the same price rang as well. Yeah , I noticed there are a lot of reviews published lately besides forum upgrades, No rest for you

__________________

blibbax

Nice review.

On techpowerup they had the GT220s performing at a similar elvel to a 9500GT. Is this right? I can't really see how it could be, they're better in every way.

These aren't really gaming cards though, and it's nice to see Nvidia soughting out their audio over HDMI. Plus, the GPU compute is very respectable. Would be a sensible choice for a media centre.

__________________

Alex

Yeah I wouldnt really rate the quality of the reviews very highly on techpowerup. They are a great site for little applications and other snippets, but the last few reviews ive read over there have been appalling. I have a GT220 ordered so I will report back shortly, thought id get one for a media center.

EDIT: read the techpowerup GT220 review. why the hell would I want to see how this low end board performs at 2560x1600 with 4aa and 16af? is that for real? Drivers are out of date too, quite shocking really.

Veridian3

OK, so since I started reviewing on this site (so long ago I can't even remember!) I have done what I can to stay away from commenting on other peoples reviews/testing. At Driver Heaven we have a tried and trusted test ethic/plan which has served us well. The main points of it can be summed up as follows:

1) Always test with the latest drivers, OS, OS patches and software patches (inc game updates).
2) Always test each comparison product at the time of the new review.
3) Use a system which is appropriate to the products being tested (e.g. Don't allow bottlenecks where possible).
4) Always use real world gaming, not time-demos.
5) Repeat each test 3 times to ensure there are no anomalies in the testing.
6) Provide full system information and details of each test for the readers.
7) Provide full results, do not exclude min/max from the graphs.

Through the above steps we are able to produce repeatable, correct and more importantly relevant results.

The problem is (and this is the reason for this post), other reviewers do not have a structured, thorough approach to testing. Whether it be through laziness, boredom or just lack of time the quality of work I see on other sites is often abysmal and I fail to see how they can produce correct comparisons. This then means their results do not match ours.

Using the Techpowerup review as an example because it was mentioned above...In that test the reviewer (who is actually a talented coder originally) has tested some Nvidia cards on one driver, then the GT220 on another. In addition to this his combination of hardware does not match the software he is using (6gb on a 32bit OS?). The use of older drivers for one card but not another generally means the person is taking older results and adding the new card to them. E.g. they reviewed/tested a 9500 6months ago and then tested the 220 at the weekend. Then added the new figures to the old graph. There is also no note of what DX level each card is using.

What are the problems with this? Well, as we all know performance changes from driver revision to driver revision. Systems also change in performance over time as the OS matures, or gets clogged up. Games can change greatly over time too. Using TPU's BattleForge test as an example, they have already stated that the 9500 uses a 186.xx driver and the GT220 is on 191.07. We can assume from this that the Battleforge patch level for 9500 results was that available at the time of the 186 series drivers (after all, if you were re-testing the 9500 for that review, why wouldn't you test it with 191?). So not only do we have changes in the OS and Driver, we also have a game which has undergone significant updates over the past few months. Including massive changes to the use of SSAO/HDAO. Finally, they are testing with the inbuilt timedemo which lasts about 30 seconds and is absolutely not representative of the actual game.

There is absolutely no way from that method of testing that the results are comparable. On top of that the person testing has not listed the exact settings used (changing SSAO setting actually changes DX level/ComputeShader options in the game) where as we always provide a screenshot of the settings we used. There is also no mention of what the FPS figure they use is. Is it average, maximum or minimum? Who knows, and knowing this makes a huge difference. If a card is listed as 25fps... if that is the minimum we are ok, if that is the maximum then we wouldn't want to consider that card for a monitor which used the same resolution. It is also worth taking into account that a card which wins on Avg FPS could actually loose out to another on min FPS... thats information you need to know which is not being provided.

So, in summary... for that review we have essentially two (or more) completely different builds masquerading as a comparable set of results. Which then prompts people to question why ours do not tell the same story?

Just to clarify, here is our test system for the GT220 review: http://www.hardwareheaven.com/reviews....d=854&pageid=4

All of the results were obtained on the same hardware, with the same OS/Software patch level and the latest drivers. For this particular review each card was tested over the 3 days before publishing. This includes the 4650 which I had actually tested a week earlier for a different article... no results were re-used.

I could go on through other pages in that article to point out issues but you get the idea. I could also go to a number of other sites (including a number of the larger ones) and tell you why their testing methods don't tell you what you really need to know, but thats not really relevant here.

Hopefully you all have a better understanding of what to watch out for when looking at reviews on other sites... so let me know, which results do you trust, us or them?



EDIT: Just wanted to add something... there is one piece of info we don't give you about our testing. That is the level used on each game. The reason for this is that we do not publish that info as it would allow GPU driver engineers to target a specific level for optimisations (and gain an unfair advantage), this has happened in the past and so we feel it appropriate to withold this information.

Alex

I gotta say veridian3, I think the more intelligent among us notice the testing standards here, its actually why I base my purchases on DH reviews. they have often been right when other publications I used to be active on have been wrong.

**Claps**

humonous

Yeah, totally agree with Alex. I was along time member on tomshardware and I noticed some serious lapses in their testing in the last few years. I think various sites 'suit' different people and some are well known to have a certain amount of bias. I think that DH was a little 'pro ati' year ago, but its very balanced in the last 24 months, I trust what you guys say here, and I have based both my last two graphics card purchases and CPU purchases on your reviews. Glad I did too.

Thanks for the info.

mkk

It's always easy to throw up numbers and neglect to mention that they should be taken with a tablespoon of salt. Good work is hard work.

IvanV

One thing that I remember whenever I'm about to write something like: "I wish :that_particular_card: was tested together with the new ones." is that for each game test Stuart has to replay the same largish portion of the game three times for every single test participant and I think again (though I occasionally do write it). That really is the hard way to test a card and just thinking of replaying the same level nine times in a row for each graph (3 cards * 3 passes) makes me a little sick of gaming (and he got off cheaply for this review, for the 5700 series review there were 10 configurations * 3 passes * 2 resolutions ). I really appreciate the effort. Although, maybe you could do 4 passes per test and discard the one that is furthest from the average for increased accuracy.

__________________
If anyone has Portal 2 and hasn't played the co-op and wants to do me a favour, let me know (PM me or whatever).

ET3D

Every site is having it against other cards. AnandTech had it against the 4670 and it lost by quite a bit. Which makes sense when compared to the 4650 here.

Veridian3

That part of a review is entirely down to the website choice I guess, based on the pricing information they get. We opted for cards which were appropriately priced at the time of testing so that people had a feel for what they could get in that price range.

As things stand, in the UK the 220 is available for quite a wide price range, roughly 20 ($30ish) between the top model and the bottom which is a lot for a budget card. The 4650 is right in that range and some 4670s are up to 20 ($30) more expensive than the top 220. For us that makes the 4650 the more valid comparison.

Bolletje

Sorry V3, but I still miss some aspects:

- The 40 nm is quite new right? Why don't look at the GPU? I see no photo's with the real core visible.

- You don't give a room temperature with the temperature testing. I also can't see which case you're using (considering the airflow). Or are you using a benchtable?

You don't have tested the 9600GT, yet you make a statement. As far as I know this isn't a real improvement. The 9600GT has 64 shaders and all clocks (mem, gpu and shaders) are higher comparing to the GT220.

Since when can you relate noise with a serie? In this case the 9600.


Not that bad review, just some things I noticed.

Veridian3

We were running to a very tight deadline for this review (and the 5700s the next day). Given the target market I took a call on what to take the time to include and felt that on this occassion other aspects were more important than a pic of the GPU itself.

28c room on a test bench, this will be noted more clearly in future articles.

I have experience from using 9600's on a regular basis and for noise i felt the 9600 wording was a good description which would have some relavence to people reading. If i was to say "its XXdb" there is a large portion of the cards target market who would not have a clue what that actually meant for them.

As for "its better than the 9600". This was an overall view of the card, taking everything into consideration I would rather have a budget model with DX10.1 and improved HDMI etc, the GT 220 is a more rounded/complete proudct. (Better doesn't always mean faster)

DJ BIG T

not a bad card at all i like it..

__________________

Bolletje

Aah, too bad, since I (maybe the only one :P?) was interested in the new GPU. Better luck next time then

Ok, clear. You also have a picture of you setup? Just curious...

Uhu, a lot of them are sold in OEM PC's. However, a lot of people also don't know what a '9600' is So is it more quiet than a stock CPU-cooler (Intel C2D - AMD X2) for example? Since most of those cards are sold to people that just want to play a simple game. Those PC's aren't equiped with after-market coolers I think.

True, it isn't so bad, but the performance (ie framerates) is lower than the 9600 serie right?