7900 GTX specs confirmed!

Finally, the specs for G71's 7900 GTX is here:

- Core Clock = 650 MHz
- Memory Clock = 800 (1.6 G)
- Memory Bandwidth = 51.2 GB/s
- Memory Size = 512 MB/1024 MB
- Pixel Pipelines = 24

Definitely a beefed up 7800 GTX, and a not-so beefed up 7800 GTX 512. My point is that this couldn't be more than just another relaunch of 7800 GTX 512 with reduced core size. Well, that makes sense since that helps increase the yield per wafer, directly increasing the availability of the card, where 7800 GTX 512 is lack of. Reduced core size also said to help Nvidia reach the core clock they hope for this card.

Performance wise, I don't know if this card can beat X1900XTX from ATi or score even. Even that is so, I reckon that Nvidia is planning to beat ATi in terms of the availability, thus gaining the share in the high end video card market. Purely an opinion of mine.

Finally.........

Back home..



Back to family.....



back to Nvidia!

Benchmarking UT2004....

It seems that I am under the spell of older games nowadays...not! I was actually thinking of how to enrich the benchmarking experiences and at the same time trying to find the way to justify my X800 GTO performances. What a lame excuse anyway!.

I installed Unreal Tournament 2004 Demo (too lazy to install the full version), with the same settings as usual:

- X800 GTO at 500/550 core/mem
- Max. settings in the graphics option in UT2004 at 1024x768 res.
- Forced 4X AA, 8X AF in the driver setting.
- FRAPS benchmarks at 60 secs, then averaged. Run for 4 times.

I chose to play the Team Deatchmatch. The map is an indoor map just to get the environment with the most triangles, vertices at all times.

Results

Running no./ no. of frames	Min. FPS	Max. FPS	Avg. FPS
1/ 5025 frames	61	86	83.8
2/ 4694 frames	47	86	78.2
3/ 5047 frames	72	86	84.1
4/ 4448 frames	54	86	74.1

Those are what's reported by FRAPS after the benchmark. One obvious thing to note is that the performance at 1024x768 is capped at 86 FPS. However, the number of frames involved in each of four runs are not the same. This is probably due to the fact that for each test of 60 seconds, I didn't follow the same route and environment. I don't know how to benchmark UT2004 properly using the FlyBy demo..bla bla..whatsoever, so this is the only method I can use. Whatever it is, the benchmark gives me some idea that I was running UT2004 for more than 70 FPS at all time. That's more than enough as 60 FPS is considered as the point where human eyes can barely see the differences of FPS above it.

Graphical anomaly?

I did say that I forced the game to run at 8X anisotrophic filtering, right? Take a look at the screenshot below.



Click the pic above for the biggest size available at my Flickr. If you look carefully at the floor above 'You've won the match!' bar (about half an inch), you can see that the floor textures are sharp, but above that, the textures are blurry (for those who are familiar with Image Quality investigation, this is easy to notice!). IMO, that is not showing a sign of 8X AF being applied at all. I have some other screenshots that shows the same property, but enough for me to show just one. I guess this is a bug in the driver, or maybe the game has problem rendering the filtering, I just don't know.

This is another older game that I enjoyed replaying under new graphics accelerator. It is a fast pace FPS game that really demands high framerates to enjoy it. Just like my experience with NFSU 2, I feel that I can finally play this game the way it is meant to be played!

Screenshots:

Rumours of the day: 7600 GT, X1800 GTO

Nvidia Geforce 7600 GT:

- use G73 core
- PCIe native
- 560MHz core frequency
- 128-bit memory interface
- 20.2GB/sec. memory bandwidth
- 7B pixels/sec. fill rate
- 730M vertices/sec.
- 12 pixels per cycle

Picture of 7600 GT (credit goes to DailyTech.com):




ATi X1800 GTO:

- 500 MHz core/ 500 MHz (1 GHz) memory
- 12 pixel pipelines
- MSRP rumours to be around $249
- directly targeting 7600 GT, thus possible launch in March

Nearly finalised G71 specs!!

Wow, we are nearing the release of G71 or 7900 GTX as more teasers about it surface around the Net. I did made an entry regarding this before, and it seems more info has arrived.

Excerpt from DailyTech:

"Quick specifications for GeForce 7900 GTX:

• PCIe native
• 655MHz core frequency
• 256-bit memory interface
• 52GB/sec. memory bandwidth
• 15B pixels/sec. fill rate
• 1450M vertices/sec.
• 24 pixels per cycle
• Built in twin dual-link DVI support for twin 2560x1600 resolution displays"

While I'm fine and dandy with the specs, one thing strikes me most: 24 pixels per cycle? Does that mean 24-pixel pipelines? If this is true, it truly contradicts the rumours stating a 32-pixel pipelines G71 before. Strike one.

Strike two: Just about 3 hours before DailyTech posted the news above, it posted the latest news regarding another G71 variant, dubbed 7900 GT. Here's the excerpt:

"Quick specifications for GeForce 7900GT from NVIDIA documents:

• PCIe native
• 450MHz core frequency
• 256-bit memory interface
• 40GB/sec. memory bandwidth
• 11B pixels/sec. fill rate
• ~900M vertices/sec.
• 24 pixels per cycle
• Built in dual-link DVI support for 2560x1600 resolution

More to come. "

That's right, another '24-pixels per cycle' product under G71's hood. Obviously something is wrong here and they did say 'more to come'. I hope it's just a typo error in 7900 GTX specs. I don't know how much I can rely on DailyTech as they are the news provider for AnandTech (maybe they're the same guys from AT or just totally different entity), but I hope it doesn't end up being critised by readers just like what happened to The Inquirer (readers complaints regarding BS news provided by them, I don't know if this is true, correct me if I'm wrong!). What I know, those specs above stay uncorrected at the time of writing. My point is that I love Anandtech and I hope it stays that way.

Ok, back to the cards, I tend to believe that 7900 GT is a 24-pipeline graphics card as it makes a good sense of a 'refreshed' graphics card w.r.t 7800 GT. The same feeling goes to 7900 GTX; that's why I still believe it's not gonna be a 24-pipeline card. Overall specs boast a whopping increment in bandwidth department, as NVIDIA claims a doubled performance compared to G70.

Hmm, I wonder if I should wait for 7900 GT now....

edited on 23rd Feb : it seems that 24-pixel pipelines 7900 GTX is true! What a sad day!

Benchmarking Battlefield 2

I must say that this is one tough (visually) game. No, let me put it this way: this is one visually beautiful war scenario game. As usual, using my own standard setup as I've done with previous game benchmarks, I took a deep breath as my X800 GTO endured a 60-seconds benchmarking as monitored by FRAPS. The map that I used is the first single player map, the one where...emm I can't remember, but it's the first of 16 missions available.

This is the BF2 in-game video settings:



Looks scary to me, but bare in mind that this is played at 1024x768, a minimum resolution that any 'above average' gamers would play at nowadays (I bet many of them would laugh when looking at this!). Once again, at the GTO clocks of 500/550, here's the result:

Max. FPS = 80, Min. FPS = 53, Avg. FPS = 64.5

I realize that this result does not apply to all maps in the game. What I did was benchmarking it during quite heavy close combating scenes with a few tanks around, just to get the most taxing scene possible.

Anyway, the overall experience is just okay, as I still notice frequent stuttering moments throughout normal gameplay. I definitely need a more powerful graphics card to play this baby.

Screenshot:

Benchmark on NFSU 2

Although it's a bit old, I tend to see how my system performs on Need For Speed Underground 2. I played it roughly a year back, just beforeI got married (I never completed the game till today, but I keep the savegame just in case I want to continue playing it later!). In this benchmark, I just use the NFSU 2 Demo, as I'm lazy to install the full version, and actually not feeling to complete the game yet.

The experience however was different than before. When I played this game a year back, I was using ATi 9800 SE unlocked to 9800 Pro. The result was okay, but I didn't switch on the full graphics setting (I think I turn off motion blur and only use the lowest AA setting and run at 800x600). Now I turn on every settings in the graphics department and run it at 1024x768 res (I have a decent 17" CRT). The game played smoothly and I enjoyed every second of it. I guess now I realize how important it is that we gamers must play to to the experience intended by the game developer (The way it is meant to be played?).

The Setup

To simplify, my system is:

AMD 64 3000+ Venice (stepping E6) at 2.25 GHz (250 HTT*9)
2X512 MB DDR Kingston Value RAM
40 GB IDE Maxtor
Sapphire ATi X800 GTO at 500/550 (1.1 G)
Win XP SP2

I ran the first benchmark using High Quality setting, with 'Application Preference' selection at Display Properties of the video card. I used FRAPS to benchmark the game. The benchmark was run for 60 seconds and then averaged.

I actually wonder what is the level of Anti-aliasing and anisotropic filtering applied in the game when every setting is maxed out. Therefore, I decided to make another run but this time, I forced the AA and AF level at the driver to 4X AA and 8X AF respectively.

Finally, just for the fun of it, I made another run with forced AA and AF, but this time, I switch on the SMARTSHADER on the driver using 'Classic' mode.

All benchmarks above are performed at 1024x768.

Results

Setting	Min. FPS	Max FPS	Avg. FPS
High Quality	25	65	45.5
Forced 4X AA, 8XAF	33	46	41.3
Forced 4X AA, 8X AF with SMARTSHADER turned ON	34	46	42.1

There it goes. Obviously, High Quality setting impliments lower AA and AF quality and 4X AA and 8X AF. The SMARSHADER also affected the FPS obtained by a little margin, considering the max fps is the same with that in no-SMARTSHADER benchmark.

Overall, I am satisfied with this result, considering the game is more than a year's old. I'm not gonna be too much infatuated by it, however I learn something useful in gaming: play the game the way it is intended to be played!

Screenshot:

I'm in dilemma...7800 GT

Yesterday, I saw a very interesting video card when I dropped by at Low Yat Plaza. Can you believe that Forsa 7800 GT was being sold at RM1199 and XFX 7800 GT at RM1249?? This is really making me nuts right now. A few months back, heck..just a month back they were sold at RM1400++ at least. Since 7800 GT was launch in August 2005 with MSRP USD449, it's been 6 months now, and I just couldn't stand it anymore. By this time of writing, many 7800 GT can be had for less than USD300.

THE PROGRESSION

There's nothing wrong with my Sapphire ATi X800 GTO. It serves me well for 4 months since I bought it. 4 months..just imagine. And now I'm dreaming of getting another graphics card that
is supposed to be mine 4 months ago. Why? Well, actually I was planning to get 6800 GS a few months back while it was about RM950. The 7800 family never really captivated my interest in POSSESSING them, although performance wise, I really wish I could have them. However, yesterday was really an eye opener for me.

You see, X800 GTO is my 4th GPU so far. I have made some observation on the history of my GPU, and I see a trend: pixel pipelines trending. I'm somehow very much affected by the number of pixel pipelines as I believe that it determines how 'powerful' a graphics card can be (I know there are many other factors). But whatever it is, I make it as a main point of progression, together with some other specs such as memory bandwidth, geometry rate etc.

Video Card	GPU Code	ROPs	Video RAM	Vertex shader
GeForce 256 DDR	NV10	4	32 MB	1
GF4 Ti4400	NV25	4	128 MB	2
Radeon 9800 SE @ Pro	R360	8	128 MB	4
X800 GTO	R480	12	256 MB	6
7800 GT ??	G70	16 (20)	256 MB	7

Okay, I use ROPs instead of pixel pipelines, because I believe ROPs affects the final performance of the card before the image is displayed. Yeah, the concept is totally different when it enters 7800 /X1900 era now. However, the number of ROPs of the graphics cards I've owned follows a nice incrementing manner until X800 GTO.

Imagine this, if I didn't buy X800 GTO 4 months ago, instead I opted for any 16-pipelines graphics card around, say 6800 GT, X800 XL or even X800 GTO 2 (16 pipes unlockable), what can you imagine about it? I'm making a 'multiple of 2' progression of number of pipelines or ROPs! That is, 4 then 8 (4 times 2) then 16 (8 times 2). That brings me to this question; what graphics card should I buy now?? Definitely a 32-ROPs one!!

Where can I get one of those now? The answer is possibly next month, during the launch of G71. But wait, nobody gives any confirmation that it's gonna be a 32-ROPs video card, they just said 32 pixel pipelines! In fact, we are nearing a new era in PC graphics that a 32-ROPs card might never see the light of day (my concern is stated below).

Therefore, it's obvious that X800 GTO is the breaking point of the progression I'm planning to achieve. However, such progression might not be able to achieve anymore due to the end of usual understanding of pixel pipelines scenario used to reflect the performance. This whole new approach to 3D architecture requires new understanding, and I'm here to understand it.

So, do you think that I should get 7800 GT? Or should I wait for the next refreshment from G71? Or should I take a peek at ATi current and future offerings?

THINGS TO COME

I have some phrase or terms that really boggle my mind when thinking about future graphics card upgrade:

- Shader Model 3.0
- HDR with AA
- WGF 1.0
- DirectX9.L
- WGF 2.0
- DirectX10
- Shader Model 4.0
- Unified Shader Architecture
- Supporting games
- dual-core CPU/multithreading supported games
- 64-bit gaming
- PPU

Allright, some of those terms are actually the same thing, e.g. Windows Graphics Foundation 1.0 equals DirectXD9.L, the last DX9 iteration before we enter DX10, or WGF 2.0 which comes as standard in Windows Vista. SM3.0 is now, but still many people question about its rate of adoption in games. While HDR is an appealing eye-candy, some say 'let's hoping it would also allow AA at the same time". Well, it's here under the hood of X1900 series from ATi(maybe X1800 also, I'm not sure!).

Up till today, I reckon that we won't have problems enduring WGF 1.0, but things will change when NVIDIA G80 arrives (possibly June) , and ATi R600 (possibly in October), supporting DX10 which means........UNIFIED SHADER ARCHITECTURE! We are not gonna have scenarios we see today, in fact, in less than a year, every graphics card around can almost become obsolete, to some extent. I don't know the details of DX10 implementation (can I install it under XP??...highly doubt it), but it surely marks an era of 64-bit computing, and gaming possibly.

I can almost describe my very concern about what the coming 6 months hold for graphics card purchasing, but to describe it is also almost difficult, especially because I'm a gamer with budget in mind. Take a look again at those phrases above and tell me, is now a good time to get a graphics card, one last time before everything becomes different??

Possible NVIDIA G71 specs...even G80

This is one of them that hold me back regarding the next graphics card upgrade, NVIDIA G71:

• basically a 'faster' G70......
• based on 90 nm process
• 32 pipelines, 32 ROPs
• possible name: 7900 GTX
• possible clock: 750 core/900 (1.8 GB) memory
  
• possibility of 7900 GT with 24 pipes, 24 ROPs
• might be launched during CeBIT around mid March 2006.

Not enough? Enter G80:

• based on 80 nm process
• first card to embrace DX10, thus Shader Model 4.0, Unified Shader architecture.
• use GDDR4 memory
  
• possible launch : June 2006.

rumours...rumours...rumours...

Poor X800 GTO in 3DMark06!!

By far, this is the worst experience of benchmarking I've ever had. Only 1678 score in 3DMark06!!
I pity myself all the way during the benchmark...



The benchmark was run with the default setting (at 1280 x 960 resolution). Although I'm not a gamer playing at this res, still it's a piss poor performance at best. Less than 10 fps in SM2.0 test and virtually zero fps in CPU test.

Although it's a synthetic test, 3DMark06 represents what the future graphics requirements will be. I already have 7800 GT in my mind, but something else..........

100% Completion in NFS:MW!

At last, a 100% completion in NFS:MW. I gotta tell ya, what a feeling it was blistering the M3 GTR through Supercharged SUVs and all. It was all good in the end for the unnamed racer.
And Josie, you're really a b**ch but nice!!

It took a much tougher side of me to complete the Challenge mode, that I felt it even sicker than the Career Mode. Sorry, cheaters and trainers are not allowed here! (although I did apply some sort of 'cheating' : waiting on top of the bus station to complete the 30 minutes pursuit survival, while the sick-*ss cops just couldn't find the way to get me on the top).

86.2 FPS in Quake 4 benchmark!

Whoaaa! I was impressed myself, how my X800 GTO pull itself in the stressful benchmark under Quake 4. I used a demo called 'HWSpirit' (I don't remember where I got this demo!) but it is recorded in the very early scenario in Quake 4 (immediately after your ship crash in Strogg planet).

Here are the snapshots of the in-game setting I used in Quake 4:

Ah-haa, you might say! Yeah, I did compromise the realism of the game as I turned off the shadows. I just think that it's too taxing that it doesn't fit me as a framerate freak.

Once again, the clocks on my X800 GTO is 500/550.

A few screenshots of the game:

You can notice that the environment looks significantly bright. I maxed out the brightness as I don't like fragging in the dark that much. IMO, it's best to use an LCD monitor to play that kind of environment in Quake 4 (better brightness and contrast!).

NVIDIA 7800 GS AGP at Anandtech

Good review, I'll make it a sticky for myself!

edited 2 minutes later: another good review of 7800 GS AGP at FiringSquad.

edited on 21st February:

Basically, GeForce 7800 GS is a 302M transistor GPU based on 110 nm process and is targeted at AGP platform only (remember it was meant for PCI-Express version, in fact a few cards were produced, consider them to be rare ones!). Below are the specs of the card:

Specification	GeForce 7800 GS AGP
Core Clock	375 MHz
Memory Clock	600 Mhz (1.2 GHz)
Frame Buffer Size	256 MB on 256-bit interface
Memory Bandwidth	38.4 GB/sec
Vertex Shading Units	6
Pixel Pipelines	16
ROPs	8
Peak Power Consumption	75 W
Texel Fillrate	6 Gigatexel/sec (375 Mhz X 16 pipes)
Peak Pixel Fillrate	3 Gigapixel/sec (375 Mhz X 8 ROPs)
Geometrical Rate (Vertices/sec)	562.5 Mtriangles/sec (6 pipes X 375 Mhz) / 4 clocks per vertex

Fourth video card: Sapphire 9800 SE

Hmmm...I miss this card. I sold it already to someone else just recently. Bought in May 2004 due to its unlocking capability. Anyway, this card helped me a lot when enduring NFS Underground and Underground 2, emm...Half-life 2 (no doubt about it, it's ATi anyway!!), and a few other games that I can't remember. I just stop its contribution to my gaming world when it was downright sucked at playing Doom 3. Fair or not, many will have to agree, that game is one SOB.

Allright, on to the card itself. It is among the most sought after card around middle 2004. Why? Apart from 9500 Non-pro can be unlocked to Pro 'softmodding' festival, this card also drew attention because it's unlockability to provide further 4 pixel pipelines,

I got to admit, the Powercolor 9800 SE was more popular. I guess the reason is it's almost quaranteed to get another good quad pipelines when softmodding it. Furthermore, there were speculations saying that not all 9800 SE in the market that time sported 256-bit memory interface. Got the catch??

Sapphire that time, released 2 models of 9800 SE (from my observations) : 128-bit and 256-bit memory to mention their exact achitectural difference. Maybe there were more, but this point is important. Why? Because the aim is to softmod this guy to become 9800 Pro! Get 4 more pipes, 256-bit memory ready, and overclock the standard 370/340(680) clock to beat the 9800 Pro (some were more motivated to reach the 9800 XT performance, wow!)

I was lucky that I got this card. I added more juice to it by slapping Coolviva cooler to it (at All IT, Low Yat, it costs RM60 now!) , replacing the stock heatsink.

And Sapphire magically ties me still, with his X800 GTO...

Another note : I'm not an ATi fanboy anyway..

5337 in 3DMark05 benchmark!

Did another benchmark on 3DMark05. The experience was better compared to when I used Sapphire 9800 SE to benchmark on this.

The setting:





here's the result:





The Feature Test should indicate 5 of 5 (as this shot was taken after the benchmark!). One thing to note: 3DMark05 reported a more accurate value of fillrate in multitexturing for my X800 GTO (nearing theoretical 6 Gtexels/sec) compared to the same result in 3DMark03 (see previous 3DMark03 benchmark).

11206 in 3DMark03 benchmark!

Here's my 3DMark03 benchmarks on my system.

The setting:



and here is the result:



One thing to note: the multitexturing fillrate, 5.346 Gtexels/sec seems lower than the theoretical fillrate (12x500) = 6 Gtexels/sec.