DH Review: VisionTek & MSI 4850s V NVIDIA 9800 GTX+

craig5320

Read The Review Here
_________________

It is hard to believe that it has been seven months since AMD launched the Radeon HD3870 and over a year since the 2900 XT was their top performing product. The latest AMD product is with us in the form of the Radeon HD 4850 which keeps the ATI product 'refreshes' on a near six month cycle. The 4850 is the first card to use ATI's RV770 core which brings with it significant increases in specification over the 3000 series and is designed to ensure AMD/ATI have a product which is competitive in the $200/£140 price range.

For a few weeks it appeared as if ATI would easily achieve this goal but some aggressive price repositioning of the 9800 GTX by Nvidia as well as a snap announcement of the 9800 GTX+ mean the 4850 is hitting a very crowded but interesting market.

Today we will be running the two new products through a selection of the latest games at resolutions up to 2560x1600 as well as delving into some Blu-Ray playback testing and overclocking in order to establish how well they perform. We will also be looking at how they compare to the more expensive cards currently available to see if the extra cash outlay really does benefit the consumer, or if these new cards provide excellent performance at a much more attractive price.

LukeK

first time ive seen AMD competitive in a long time. great review liked the info and tests on physx, looks like it has some serious potential.

9800 GTX+ seems a good card, im reading issues about 4850s overheating and cases getting hot due to them. would really need a 120mm fan over the card pointing to the rear of the case and a few of the slot covers off the chassis to let the air out. i heard the fan on the 4850 spins at 10% normally? but 90c? someone dropped the ball there I think.

Judas

good review

can't wait to see the 4870 review

wolf2009

awesome review, best looking review on the net. the graph designs could be improved, its a little bit hard to find out which graph is for which card .

i noticed that ati released avivo for HD4800, where is it ?

humonous

i dont really think the graphs are hard to read, just maybe have them in resolution ascending order rather than 2560 then 1920 then 2560 for SLI as it is on some pages. Can follow it easy enough though, however maybe horizontally reading out for all those cards would have worked better rather than vertically.

All very minor though as this review told me more than the other three or four ive read today.

Been a while since I have used an ATI card but isnt Avivo built into the advanced tabs of the newer drivers?

Is it just me or is Physx probably one of the most exciting things in a long time? I really am stoked about it, especially if more games start supporting it and making use of the PPU functions. Nvidia really are hard to beat but that AMD card is actually pretty impressive. The temperatures are very concerning.

Did you see any issues veridian when overclocking the card? i mean 90c seems insanely high to me and im sure you have one of the best cases with great airflow.

brutusmaximus

brilliant review. great detail and info in that. i like the maximum playable settings page that is a fantastic idea.

First AMD card ive seen in a long time which actually looks powerful enough to run the newest games at high resolution and its very competitive.

My concern like a few others in this thread is the sloppy cooler design. 90c? I was reading on the anandtech forums that several people have had them die ! surely they should have cranked the fan speeds a bit. I mean we all like silentoperation but not at the cost of fried hardware.

the 9800 GTX+, is interesting. I am im two minds about it. I think its lazy of Nvidia to "remake" the 9800, however as it is on new 55nm perhaps it is clever thinking. The naming is hideous however, why not call it the 9850GTX or something? Everytime I see a "+" sign I cringe.

This round certainly seems closer but NV have the edge again I think, especially if you consider the quality of the drivers overall.

Veridian3

Install this file: https://a248.e.akamai.net/f/674/9206...code_64791.exe

Then open Catalyst Control Centre in basic view and it should appear in the list of wizards. If it doesn't you may need to wait for Catalyst 8.7.

wolf2009

alrite installed it, tried encoding with it, its not using GPU at all. using 50% of my quad core . GPU load stays at 0% as shown with GPU-z .

Veridian3

That may be a gpuz bug or it may be that ATI need to update the publically available code for your card (what is your card?). Checking CPU usage doesnt really mean much as even the NV badaBOOM application still uses a significant amount of CPU as well as the GPU.

If you look at the image on this page you will see what i mean: http://www.driverheaven.net/reviews....d=579&pageid=5

wolf2009

oh alrite, my card is HD4850 .

Cow_160483

Wow the HD4850 looks impresive but am concerned about CF for some of the games but i have decided that this is the card am getting as soon as it is avaliable in my area.
But i do hope that they fix the heating problems. Otherwise the review was good as usual !!

WaltC

You are probably not going to like some of the things I'm going to say here, but I feel like being honest this morning...

First, this sort of pissed me off:

So...first of all, the entire comparison is built around looking at an off-the-shelf card which has been shipping for weeks (I have 2x 4850 CFX installed at home) with a reference product sent directly to you by nVidia which is not yet shipping anywhere at the moment and which no one can buy.

Yet, you decide that the unavailable reference card is "best in class" over the off-the-shelf shipping HD 4850. It goes without saying that the pertinent and fair comparison would have been to hold off on making this comparison until the GTX + was actually on store shelves and available for purchase. But to start handing out awards before the product even ships? Suppose the sky fell in and for some mysterious reason nV isn't able to ship too many GTX +'s until much later in the year? Did this occur to you at all? It seems to me that "not available for purchase" merits a big minus on points tallied towards any kind of award, regardless of who makes the gpu.

Next, the title of this comparative review reads:

"HD 4850s V 9800 GTX +"

Yet, from the very first comparison it is obvious that the title should have been entitled "HD 4850s V 9800 GTX + V 9800 GTX OC V 3870x2 V GTX 260 V GTX 280"...man, talk about bait and switch...

Adding insult to this injury, when I examine the "test system" data you published, I find that the only 3d cards tested for *this comparison* were as follows:

So I'm wondering from which test system it was, and which test, that you decided to take the HD 3870 numbers, since apparently the HD 3870 was not included in the test system used to run the tests upon which this article was based.

Also, on a much smaller scale of criticism, while you listed the GTX 260 you used here as also being a *reference card* no doubt supplied to you by nVidia for this test (hard to know where it came from as you don't say), you don't call the 9800 GTX + a reference card in the test system specs you published, although you do reveal it much later on in the review as I've quoted you above, buried in the text.

Next, I have to say that I found the presentation of the bar chart organization very haphazard, confused, and inconsistent. Some examples:

Grid. As it is obvious that neither CFX or SLI drivers support GRID as of the date of your test, what's the point in publishing CFX/SLI frame-rate bar charts for the game at all? Better to say simply that neither IHV driver set supports GRID presently, imo, and leave it at that.

Tiger Woods '08. You say:

Why mention only the HD 3870 x2 here?

Couldn't you have also said that people running CFX or SLI systems with TW '08 would also be treated to single-gpu performance in this game? I don't know first hand if that's the case, but since you didn't attempt an SLI/CFX frame-rate chart for this game at all, unlike you did with Grid, the only explanation would seem to be that neither IHV supports TW'08 in its CFX/SLI drivers at present.

Lost planet. Aside from the fact that I'm getting really tired of seeing this terrible game constantly benchmarked all over the 'net, I don't know what you meant here:

The numbers you presented for the HD 4850 and the 9800 GTX + present darn near identical "scaling" in this game between single and CFX/SLI operation. Both products scaled to nearly 100% improvement when tested under CFX/SLI conditions.

I'd also like to know why when testing SLI/CFX for this game in particular you decided suddenly that 2560x1600 wasn't worth doing. If you couldn't do it for some other reason aside from the fact that you weren't happy with the way any of the cards performed at that res even under SLI/CFX (although that so far hasn't stopped you before), it would have been nice if you'd explained jumping around here.

When people start playing shell games with testing conditions in a comparative product review it always makes me feel as if the author is trying to manipulate me somehow--but that's just me , and maybe other people don't feel that way about it.

HL2: E2.

It would be really nice at this point if you might go into a little detail about just how much more money a 260 or a 280 costs above the price of an HD 4850--never mind the fact that this review was *supposed* to be comparing a non-shipping 9800 GTX + with a shipping HD 4850. In fact, the price disparity is so much that there's really no place for those products in a product comparison including an HD 4850. I mean, when you entitled this article you had a good idea--why didn't you stick to it?

I mean, I can get 2x 4850s for significantly less than I can buy 1 GTX 280. And GTX 280 SLI? Aside from the fact I could darn near buy SIX 4850s for the price of TWO GTX 280's, it appears from your presented numbers that GTX 280 SLI would not be a worthwhile purchase at any price...

It's this kind of talk that really sort of pisses me off. OK, so what's the "average frame-rate" number in your testing actually mean? I mean, clearly, it does *not* mean an average taken between the highest and lowest frame-rate numbers you recorded, so then it *must* mean something like this:

Average Frame Rate = the frame-rate at which the game most often runs.

Now, in this case you have said that even though the 4850s have a higher average frame rate than the GTX + that it is, and I quote, "irrelevant when we experience stutters from the twenty frames per second minimum."

Well, gee, you know, that's the thing about a "minimum" frame-rate number--you have no idea of how *often* it occurs. In fact, it is entirely possible that a minimum frame rate occurs only *once* during testing, and hence is recorded as the absolute minimum frame rate the game ever reached when tested. Clearly, then, the most important number of them all is neither the minimum or the maximum, but it is the "average" frame rate at which the game plays. But you've said it was "irrelevant," which is dead wrong, imo.

Look, let's suppose that the 4850 during the test often hit the minimum frame rate number you've recorded, and let's suppose that this induced constant stutter as you've implied. What do you think constantly hitting the minimum would do to the *average frame rate* you recorded?

Well, gee, it would drop it, wouldn't it? Yet, since the 4850 has a higher average frame rate than the 9800 GTX +, then doesn't that clearly tell us that although the GTX+'s minimum frame rate is higher than that of the HD 4850, that the +'s average frame rate is *lower* than the 4850's simply because the GTX + is hitting its minimum frame rate number a lot more often than the 4850 is hitting its minimum?

I think it does, and I also think it tells us that if we are going to see "stuttering" in either system, then it is more likely that we'll see it running the GTX + because the GTX +'s average frame rate is lower than that of the 4850.

One last little peeve here, even though there were several other things I wanted to mention but won't because I'm tired of sitting here typing this at the moment:

Aside from the fact that you managed to shoehorn lots of other, far more expensive products into a product comparison that was titled as though it was going to review only two products from a comparative standpoint: why didn't you include 9800 GTX OC SLI and GTX 260 SLI results here? What's the point of including these products singly but not in pairs when you've gone to so much trouble to pair up everything else, and even thrown in 3870 x2 results that you picked up from somewhere?

Suffice it to say that I think this product comparison would have been far better if you had restricted it to three products only, the 9800 GTX OC and the +, and the 4850. And of course it goes without saying that a non-shipping reference product should never score higher than a shipping one, for the simple reason that you can't be sure if the reference design will exactly equal the product that finally ships and that people will buy, and you can't even be sure of exactly *when* it will ship, either. That seems like nothing but common sense.

If you can't tell I was disappointed that this comparison wasn't what it should have been. It read far more like a "Let's try and put the best face on nVidia's product line that we can" puff piece as opposed to a sincerely undertaken product comparison between the 9800 GTX and the 4850. I mean, the way the article was written and approached I think the only correct title for it would have to be:

NVIDIA's ENTIRE MID-RANGE AND HIGH-END PRODUCT LINE V AMD's HD 4850

mkk

Good points as always Walt, and don't worry about being too honest because this is only helpful for future articles.

Syndicate2083

I can see your point on the reference card, but disagree that it should automatically 'lose' because it isn't in a store yet. In any event, I really appreciate seeing the multi-tiered reviews, it gives me a good picture of performance to price, and how a card stacks up against a variety of other cards. That can be a very difficult thing to discern on your own with the deluge of cards now.

There are a plethora of sites doing a 9800 gtx+ to 4850 review at the moment, so I don't really see why this is such a huge problem, except that the + is coming out the winner here. I havent really seen a specific bias from DH toward/for nVidia at all. Recently ATi has been up the creek without a paddle, they are trying, but still lacking.

Alex

I am not really so sure they are good points. Sure Waltc can write small books to address things he doesnt like, but ill be honest, ive read his posts on many sites (Ive an account on beyond3d for instance under another name), and the guy is really extremely biased to AMD/ATI. Certainly he can word his posting very well and at times appear to be neutral but at the end of the day he is extremely biased towards the guys in red. I have nothing against ATI/AMD or whatever they call themselves, but I find it rather irritating that he frequently complains about the articles on DriverHeaven being biased when he is in fact the biased person. You can rest assured when an article goes up which ATI do well if there is something which could be classed as even slightly "unfair" he won't mention it. DriverHeaven was seen for years to be an ATI "fansite" and I was one of the people who thought that, however I was quite happy to be proved wrong about 18 months ago when they started giving more positive reviews for Nvidia products, clearly I was wrong and they are just telling it like it is and I value this site for fair and balanced editorials, DH is actually the most important site I view for most hardware review purchases, they have pissed off many companies over the years for just being honest.

It only swings one way for Walt, its pretty common knowledge. No offense to the guy. he is quite intelligent, however he does tend to rant on a little about the most inconsquential things especially if his favourite company is not shown to be the top dog in the reviews.

humonous

You know ive read your posts here many times, for a guy who likes to verbally masturbate this is hardly the ideal way to lead into a discussion, is it?


What about all the ATI paper launches over the years? I don't see you posting a single thing about those. Should DH just ignore the product with the immediate outlook that because "its not available at the time of writing it shouldnt be published"? That is a ridiculous concept, there are many products leading hardware review sites publish that aren't yet available.

Actually they didn't "decide" this, the testing showed it was generally better, not always but generally.

Oh don't be so patronising. Of course they are going to review a product that nvidia say is to compete. Why wouldn't they? I enjoyed reading it, and clearly if ATI had won by a large margin so would you. They didnt, hence the bitching. I am glad DH accept the hardware to run previews of products, its interesting and always presented well. What exactly do you do that gives you the right to say "did this occur to you at all?" I am sure Veridian 3 is quite aware of this, after all they have reviewed about 30 ATI boards over the years that weren't released at time of publication.

Ok, lets remember this ATI gave DH a product that isn't released yet, but kicks the ass of the Nvidia product its being compared against. I am sure you will be first in the thread to complain about that, right?


Man, talk about a problem! You ever hear of "additional comparison products?". Thank god you arent handling reviews for a leading site, or there would be no additional hardware in the mix, which is useful to me because a: I can judge the hardware against hardware I might already have, and b: have a wider view of the placement of the performance.


Clearly you aren't quite as smart as you think then, cause it was easy for me to understand


Then skip the page? I personally think its a really good looking title with a class leading engine, has DX10 support with fur, and a load of options most games don't have. Its also a great stress test for high end cards. Talk about being smallminded?


Well you should be glad I am not in charge here because you are basically saying the review is "faked". Anyway I wouldnt worry about you being manipulated because I think you are being paid much more than DH could ever change in a review.


Its not a direct comparison, the 260 and 280 aren't even in the review name, they are ADDITIONAL hardware comparisons for reader interest. DH have done the reverse in many reviews, including high end ATI hardware in a comparison with mid range Nvidia hardware.

Because unlike you, most of us find it interesting to have more comparison hardware in the mix. Guess you will never understand that concept even though DH has been doing it for years and everyone else loves it.


Well as you are being so "honest" and are "pissed off", let me just say, reading your snide, rude and nonsense statements on these forums just pissed me off. Why you are still here is beyond me. This is not an ATI fan site, its not an nvidia fan site. DH tests all hardware they get and will compare it to a wide range of other hardware to educate and inform the readers, i.e, me. They dont play politicial games like many sites I visit. I will never forget buying a graphics card on another sites views and getting it home and finding out it was performing just like DH said, and no one else clearly had the balls to say it.

You my man, need to get another hobby (like maybe using your crossfire system?)because you are neither intelligent, coherant or open minded, you are petty and rather rude.

I know they wont do this but you should be removed, and might I add not for sharing your views, but in the manner you did.

Veridian3

Nope, I like feedback of any sort.

We review the products which are sent to us in a way we think is useful to the consumer. In the case of the GTX+ having it included along with the 4850 allowed us to better inform the readers about how well they could spend their hard earned cash. Should they buy now or wait a month etc.

This doesn't just happen to NV cards, in a few hours we will be publishing an ATI article featuring a product of theirs which will not be available for some time.


We had no reason to believe there would be any issue in delivering the GTX+. If this did happen then the consumer still has the option to go and buy an alternative, they can also do so feeling better educated as they know exactly where the GTX will place when released.



The purpose of the article was the 4850 vs the 9800 GTX+. There were other cards in to let readers see where both products placed in the current marketplace (assisting the consumer/reader yet again) but the new cards which we were reviewing for the first time deserve top billing.


This was an error on my part, as with all launches we are working to a tight deadline and as a result i missed the 3870 from the test system list. Thanks for pointing it out. I can assure everyone reading that all cards tested in the review were benchmarked on identical builds with the only difference being motherboards required for SLI/CF.


Another bit of the test system page which could have been a bit clearer but it doesn't matter too much in my opinion as page 6 of the review clearly states "reference" and begins with "Our 9800 GTX+ review sample is a reference design card direct from Nvidia". This is hardly hidden away.



This is not entirely correct, ATI and Nvidia both provided us with drivers which were designed to work with CF/SLI and GRID (and as promised by Nvidia the drivers have since been made public). On the Nvidia side SLI worked, but not as well as we had expected. On the ATI side there were numerous bugs such as corruption and drops in performance, something we mentioned in the review. If you go back and look at the charts you will see that there is a difference in performance between single 4850/9800 and CF/SLI.



When writing the review I took the decision to include the 3870X2 in the "single card" results. The primary reason for this is that it gave readers a very easy way to see where the 4850 compares to that particular card, and to a certain extent how it would compare to 3870 CF. As a result it was the only multi-gpu product in the graph for tiger woods and the performance which was out of line compared to other tests required some cl