ATI Radeon HD 4850 Versus NVIDIA GeForce 9800 GTX+

[Apoptosis]

ATI Radeon HD 4850 Versus NVIDIA GeForce 9800 GTX+

The battle between ATI and NVIDIA reached a tense moment today when ATI removed the gag order on the ATI Radeon HD 4850 graphics card by more than a week. NVIDIA was trying to spoil the launch of the ATI Radeon HD 4850 by launching an ultra secret graphics card called the GeForce 9800 GTX+! Read on as we benchmark the Radeon HD 4850 versus the GeForce 9800 GTX+!

The ATI Radeon HD 4850 and GeForce 9800 GTX+ rasie the bar for graphics cards and perform better than one would expect for a $199-$229 product. Enthusiasts and gamers better start saving some money up because you'll want one of these before summer is over.

Article Title: ATI Radeon HD 4850 Versus NVIDIA GeForce 9800 GTX+
Article URL: http://www.legitreviews.com/article/731/1/

PLEASE DIGG IT IF YOU LIKED IT: http://digg.com/hardware/ATI_Radeon_HD_ ... e_9800_GTX

[Bio-Hazard]

Some good info there Nate, thanks for the quick turn around on the reviews................

[Apoptosis]

Some good info there Nate, thanks for the quick turn around on the reviews................

Thanks a ton! Its been a long day... I'm still working on the max board power on the GeForce 9800 GTX+ as no one I talk to knows the answer. I spoke with someone in the know and the replied with this:

I asked our engineering team on the exact figure for the max TDP power consumption of the board and waiting to hear back from them.

Given the hour of the night, it may take till tomorrow to find out.

To give you a ballpark figure, the GeForce 9800 GTX+ TDP is within 10-15W of the standard GeForce 9800 GTX (it’s certainly below 180W)

I will send you the official figure as soon as I get it.

I'm curious what this number is as I am sure many of our readers are... I'm working on it and will update the article as soon as I get it.

[stopthekilling77]

Nate and crew, seriously - you are troopers! I had a long crappy day and I sign on now to find this. You do your jobs very well!
Onto the cards though, I would wonder if you would show some glimpses into the world of gaming at 1680x1050 as a few of us cant quite get to those higher resolutions, and trying to imagine the numbers is bleh knowing it won't necessarily be the median between the two given ones. I'm thinking this 7800GT of mine will be getting replaced soon!

[skier]

good thing i remembered this wasn't the high end card and whats this, ATI's new cards aren't handicapped on lightsmark anymore

[DX]

Bah I was hoping that ATI would win this round. But at least they are putting up a good fight.

[gwolfman]

Thanks! Are they going to enable PhysX on the 8800 GTX?

[Apoptosis]

Thanks! Are they going to enable PhysX on the 8800 GTX?

Very soon... This is what NVIDIA says, so next month.

In the next few weeks, we will also bring this functionality to all GeForce 8 and 9-series GPUs.

[gwolfman]

thanks!

[Apoptosis]

I spent another couple hours overclocking the card and managed to get the NVIDIA GeForce 9800 GTX+ higher with full stability.

When it came to the GeForce 9800 GTX+ we overclocked it manually using the EVGA Precision overclocking utility. We were able to reach 840MHz with full stability and had the shaders locked in at 2084MHz. The memory was able to be overclocked from 2200MHz to 2430MHz, which was also a nice increase.

I manged to get it up to 855MHz on the core, 2200MHz on the shaders and 2550MHz on the memory ICs. This is over a 100MHz boost on the shaders and memory, which is great.

max_overclock.jpg (93.9 KiB) Viewed 16710 times

If you recall from our 3DMark Vantage test page the stock clocks scored 7888 with Physics enabled. After the overclock the score increased nearly 1,000 points up to 8,862 3DMarks (a 12.3% increase in performance)! After doing these last few increases the overall score jumped up to 9,211 3DMarks, which ends up being a 16.7% performance increase!

[DX]

So is there any reason now to purchase an HD 4850? Other than Quadfire and keeping AMD in business?

[gwolfman]

I spent another couple hours overclocking the card and managed to get the NVIDIA GeForce 9800 GTX+ higher with full stability.

When it came to the GeForce 9800 GTX+ we overclocked it manually using the EVGA Precision overclocking utility. We were able to reach 840MHz with full stability and had the shaders locked in at 2084MHz. The memory was able to be overclocked from 2200MHz to 2430MHz, which was also a nice increase.

I manged to get it up to 855MHz on the core, 2200MHz on the shaders and 2550MHz on the memory ICs. This is over a 100MHz boost on the shaders and memory, which is great.

*snip*

If you recall from our 3DMark Vantage test page the stock clocks scored 7888 with Physics enabled. After the overclock the score increased nearly 1,000 points up to 8,862 3DMarks (a 12.3% increase in performance)! After doing these last few increases the overall score jumped up to 9,211 3DMarks, which ends up being a 16.7% performance increase!

Does only PhysX/CUDA enabled program benefit from this? If so, how much longer before the drivers can intercept calls on the OS/C language level and redirect them to the GPU?

[LuxZg]

So is there any reason now to purchase an HD 4850? Other than Quadfire and keeping AMD in business?

Ofcourse there is! Here:
- it's already out
- it's already available for 170$
- by the time GTX+ comes in mass it will drop even lower probably
- DX10.1
- HDMI audio throughput "the way it's ment to be done"
- single slot (btw, heared that just re-mounting the cooler with different paste drops temps A LOT!)
- less power drain and single power connector (no need for big and/or advanced power supplies)
- CrossfireX works on Intel too!

And as for 9800 GTX/GTX+ I see only 2 real advantages
- dual slot for OCers as it will cool it better for sure
- limited use of PhysX driver, but since it's VERY LIMITED so far, it will take a lot of time to have some real advantages I'm affraid

I still think it's a tie, and you'll have to wait 4870 to be sure. I think ATI will under-price nVidia and keep the upper hand on price/performance, and I wouldn't be surprised if 4870 gets price close to GTX+ 230$. Since it too will have dual slot cooler, and better clocks&overclocking potential, that leaves only PhysX for nVidia as an advantage. And where do you use it? In Age of empires? Or UT3 that runs hundereds of fps anyway? Blah!

Waiting far ALL these pricedrops to kick in, and both 9800GT/GTX+ and HD4870 to come out, and custom cooled 4850's as well, so only around the end of July will we have better view at all of this.

P.S: Oh, and I think that ATI has more space to improve in drivers for 7xx chips than nVidia with G92 series, as it's pretty much tweaked as it is.

P.P.S. Can anyone confirm Accelero S1 fitting HD4850? The word is that mounting holes are the same as in HD3xxx but I'd like to have at least 2 sources for this :D

[Apoptosis]

I've had some people e-mail me and tell me I made up my testing numbers... What a joke... My numbers are all legit and correctly tested. I've been doing reviews for nearly seven years now, so I hope I have this down by now.

I had someone on DIGG say this:

The results of the 4850 differ from most other reviews and user results in various forums. Since there is no reason presented for this difference I cannot digg it.

I don't think my test results are significantly different than other site. I spent a few minutes putting all the numbers into a new excel table and came up with this:

chart.jpg (65.8 KiB) Viewed 16316 times

I dropped Tomb Raider since ATI was tested at 4x AA and NVIDIA was at 16x AA since it wasn't apples to apples. (ATI tops out at 4x AA on this game). The average difference was 10.4% in favor of the GeForce 9800 GTX+. I then looked around the web..

FiringSquad Said This:

This battle is a neck-and-neck race. In games like Quake Wars, F.E.A.R., and Call of Duty 4, our testing gave the edge to the 4850. But in Lost Planet DX10 and BioShock DX10 the GeForce 9800 GTX ran faster than the 4850 (in Company of Heroes, the 4850 and 9800 GTX traded wins).

I had the 4850 doing better in COD4 also... The 9800 GTX+ did better in Bioshock... In CoH we tested at the highest quality settings and the 9800 GTX+ was faster hands down. FiringSquad tested on a really old game version and didn't list all the settings the used, so who knows on why they had mixed results. I think this confirms our numbers are spot on... Not too many other sites have benchmarks to go off of and some like PC Perspective only have four tests completed... We have 10... I stand behind our numbers 100% and they are right.

[DMB2000uk]

There should be a mandatory senders age stamp on all internet correspondence. I'm sure you'd find the majority of moaners to be the pre-pubescent type.

Some people just think they know everything, if they even knew the work involved in reviewing I think they'd shut their traps and be a little more appreciative.

From people in the know, you did good Nate. It must be upsetting when people point blank question the validity of your work, but try not to let it get to you. They're not the ones with the reviewing experience, and besides the vast majority of people are thankful for what you do (Its a shame that its only the moaners that are the vocal ones).

Dan

[DMB2000uk]

Just read a thread over at overclock.net and it bugged me how many were stupid enough to cry fake because of the differences in scores between reveiws, I've signed up and posted the same thing over there, but if anyone is reading this and wondering about discrepancies, I think they come from the driver differentiations between the sites. If you notice LR are the only ones to be using Catalyst 8.6 (and the 9800GTX+ is using a newer forceware version).

ATI CATALYST 8.5 drivers were used on the Radeon HD 3870 X2 graphics card and ATI CATALYST 8.6 drivers were used on the new Radeon HD 4850 graphics card. NVIDIA Forceware 175.16 WHQL drivers were used on all GeForce graphics cards except for the GeForce GTX 280 series cards as they used Forceware 177.34 drivers and the GeForce 9800 GTX+ that used Forceware 177.39 graphics drivers.

Catalyst 8.5
ForceWare 177.34 (for GT200)
ForceWare 175.16 (everything else)

Radeon HD 2900 XT 512MB PCIe: with Catalyst 8.5 drivers
Asus Radeon HD 3870 512MB PCIe: with Catalyst 8.5 drivers
Radeon HD 3870 X2 1GB PCIe: with Catalyst 8.5 drivers
Radeon HD 4850 512MB PCIe: with Catalyst 8.501.1-080612a-064906E-ATI drivers
MSI GeForce 8800 GTX 768MB PCIe: with ForceWare 175.16 drivers
XFX GeForce 9800 GTX 512MB PCIe: with ForceWare 175.16 drivers
XFX GeForce 9800 GX2 1GB PCIe: with ForceWare 175.16 drivers
GeForce GTX 260 896MB PCIe: with ForceWare 177.34 drivers
GeForce GTX 280 1GB PCIe: with ForceWare 177.34 drivers

Given the short notice timescale that these reviews were put under I think it's unreasonable to think that all of the cards should have been retested with the latest driver versions.

Dan

[bubba]

...It must be upsetting when people point blank question the validity of your work...

Yes it is

[Apoptosis]

minor update -

UPDATE (6-21-2008 at 2PM CST): NVIDIA just informed us that the official launch date for the GeForce 9800 GTX+ is July 16th. They also wanted to share that overclocked 9800 GTX cards are currently available on the market today that deliver virtually identical performance as the 9800 GTX+. So if any of our readers want one today, they can pick one up from video card companies like EVGA and XFX have overclocked versions. Pricing on the 9800 GTX has dropped has dropped the past few days and those cards can now be found for as little as $199.99. Based on our new overclocking results it might be worth the wait though.

Here is a chart to show a couple current 9800 GTX cards clocked a tad higher than the 9800 GTX+

[Apoptosis]

Another Update - (geez I think everyone at NVIDIA is working on a saturday)

The max board number I received from our engineering team yesterday was 140W.

Note that this number is lower than our max TDP figure for standard GeForce 9800 GTX.

Our engineering team also went back and tested the standard 9800 GTX with our latest GPU’s and measured 142W. This is due to the improvements made in the process since we first manufactured the 9800 GTX GPU’s.

Essentially, the GeForce 9800 GTX+ and GeForce 9800 GTX have roughly the same TDP board power consumption.

It seems our power measurements on the original cards are not accurate as they have improved the manufacturing process... Great there goes the whole point of keeping 'legit' power consumption charts.

[JohnnyVic]

I don't have a problem with the numbers but I don't like the fact that you reviewed a card that is not out yet and is by all accounts at least $30 - $60 more expensive than the one you were comparing it against. It seems like you were careful to let everyone know that they could buy NVIDIA cards today that can match the GTX+ performance but a quick check on Newegg shows those cards are anywhere from $240 to over $300.

Going by current prices, the 4850 is a $170-$200 part while the GTX+ is a $240-$300 part. Yet you referenced both products when you mentioned "$199-$229" at the very end of your article. Seems too carefully worded to make the NVIDIA card seem to be just as good or a better value.

The ace is that NVIDIA has PhysX support right now

Exactly how does that proprietary "ace" benefit anyone more so than say support for DirectX 10.1 which was not mentioned? PhysX is currently a gimmick by all accounts. It may have lots of supported titles but at the end of the day no one is mentioning the great benefits of PhysX in game reviews.

Another issue is that the overclocking numbers are not necessarily going to be reproducible. I believe you are comparing an engineering sample to a shipping product in this respect.

Bottom line is that nVidia did a vaporware launch of a higher priced part to spoil ATI's party. There are no power savings by all accounts so from the buyers perspective it is no different than an overclocked GTX. Just like nVidia's shenanigans with 10.1 and Assassin's Creed, this NVIDIA maneuvering has stink all over it.

Big for NVIDIA and a little to Legit Reviews