nd22 wrote on 2025-11-11, 16:51:

Ozzuneoj wrote on 2025-11-11, 16:49:

I'm guessing you meant nforce3. […]
Show full quote

nd22 wrote on 2025-11-11, 16:20:

Nforce2 is off to a strong start in Quake 4 - relatively speaking - however all 3 platforms are strongly bottlenecked by the weak 7800gs!

I'm guessing you meant nforce3.

The 7800GS is definitely not a top performing GPU for a game like Quake 4, but the 3400+ is also pretty weak for a game like this:
https://www.tomshardware.com/reviews/cpu-char … 07,1644-17.html

The fact that you're only dropping from 36fps at 1024x768 to 23fps at 1600x1200 shows that there is also a substantial CPU limitation preventing the frame rates from ever getting much over 40.

You are correct sir! Sorry for the typo!

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance.
https://www.hexus.net/tech/reviews/graphics/4 … preview/?page=5

Though, obviously their benchmark situation could be different from yours too.

Unlike Quake 4 when running even a simple monitoring program like fraps would get you an out of memory error you could actually play FEAR with fraps and not go over 2gb of RAM. However the video card is simply too weak to handle this game - the minimums are just too low to be of comfort.

Ozzuneoj wrote on 2025-11-11, 16:55:

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance. https://ww […]
Show full quote

nd22 wrote on 2025-11-11, 16:51:

Ozzuneoj wrote on 2025-11-11, 16:49:

I'm guessing you meant nforce3. […]
Show full quote

I'm guessing you meant nforce3.

The 7800GS is definitely not a top performing GPU for a game like Quake 4, but the 3400+ is also pretty weak for a game like this:
https://www.tomshardware.com/reviews/cpu-char … 07,1644-17.html

The fact that you're only dropping from 36fps at 1024x768 to 23fps at 1600x1200 shows that there is also a substantial CPU limitation preventing the frame rates from ever getting much over 40.

You are correct sir! Sorry for the typo!

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance.
https://www.hexus.net/tech/reviews/graphics/4 … preview/?page=5

Though, obviously their benchmark situation could be different from yours too.

I have read that review in December last year just as I was starting testing socket 754. They do not mention the video settings used. I higly doubt that any geforce 7800gs is capable of 75 FPS at 1280*1024 at ultra quality with advanced settings all ON, VSYNC OFF, AA OFF.

nd22 wrote on 2025-11-11, 17:00:

Ozzuneoj wrote on 2025-11-11, 16:55:

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance. https://ww […]
Show full quote

nd22 wrote on 2025-11-11, 16:51:

You are correct sir! Sorry for the typo!

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance.
https://www.hexus.net/tech/reviews/graphics/4 … preview/?page=5

Though, obviously their benchmark situation could be different from yours too.

I have read that review in December last year just as I was starting testing socket 754. They do not mention the video settings used. I higly doubt that any geforce 7800gs is capable of 75 FPS at 1280*1024 at ultra quality with advanced settings all ON, VSYNC OFF, AA OFF.

Is that the settings you're benchmarking FEAR at? The game is from the year before the 7800GS was released as an upper mid range card, so there is no way that it is too weak to run the game, but at maximum settings nearly any PC was struggling back then. That game had some notoriously performance destroying settings if I remember correctly... I believe the soft shadows (or something else, its' been 20 years...) were barely usable until the 8000 series was released, and did not work with anti-aliasing.

This benchmark mentions that too:
https://hothardware.com/reviews/nvidia-geforc … -kicking?page=5

And there you have it! The first generation of Athlon 64 - the clawhammer - went through 57 tests on 3 AGP platforms!
There is just one question remaining: should Tom2D be included in the final score or no? Should we take into account the colossal victory and enormous difference between VIA and NVIDIA/ULI or simply drop it from the test suite?

nd22 wrote on 2025-11-11, 16:46:

We reached the final game: striking FEAR in even the most powerful video cards of 2005 the last 3 tests will be at the same 1024/1280/1600 resolutions with all the details maxed out of one of the best games ever made!

Did you run the benchmark with FSAA disabled just like in Quake 4 and Doom 3?

Tom2D should be included, but given lower weight.

Ozzuneoj wrote on 2025-11-11, 17:09:

Is that the settings you're benchmarking FEAR at? The game is from the year before the 7800GS was released as an upper mid range card, so there is no way that it is too weak to run the game, but at maximum settings nearly any PC was struggling back then. That game had some notoriously performance destroying settings if I remember correctly... I believe the soft shadows (or something else, its' been 20 years...) were barely usable until the 8000 series was released, and did not work with anti-aliasing.

FEAR and Quake 4 can be made playable by reducing the resolution to perhaps 1280x1024 and reducing quality level. But rather than doing it, you would play these games on a more powerful PC. These benchmarks serve to show limits of Athlon 64 with AGP.

Ozzuneoj wrote on 2025-11-11, 17:09:

Is that the settings you're benchmarking FEAR at? The game is from the year before the 7800GS was released as an upper mid range […]
Show full quote

nd22 wrote on 2025-11-11, 17:00:

Ozzuneoj wrote on 2025-11-11, 16:55:

With a faster single core CPU, like the FX-53, it looks like the 7800GS is capable of significantly more performance.
https://www.hexus.net/tech/reviews/graphics/4 … preview/?page=5

Though, obviously their benchmark situation could be different from yours too.

I have read that review in December last year just as I was starting testing socket 754. They do not mention the video settings used. I higly doubt that any geforce 7800gs is capable of 75 FPS at 1280*1024 at ultra quality with advanced settings all ON, VSYNC OFF, AA OFF.

Is that the settings you're benchmarking FEAR at? The game is from the year before the 7800GS was released as an upper mid range card, so there is no way that it is too weak to run the game, but at maximum settings nearly any PC was struggling back then. That game had some notoriously performance destroying settings if I remember correctly... I believe the soft shadows (or something else, its' been 20 years...) were barely usable until the 8000 series was released, and did not work with anti-aliasing.

This benchmark mentions that too:
https://hothardware.com/reviews/nvidia-geforc … -kicking?page=5

Actually you CAN play FEAR with both soft shadows and AA on geforce 7 series just fine and I suppose on geforce6 series also however performance wise you will get a slideshow.
Here how do it:
1. launch FEAR. set everything on max but disable AA and SS.
2. benchmark to get a baseline data with the built in test
3. enter settings and enable AA 4X
4. test again!
5.enter settings and disable AA but enable SS
6. test!
7. for the last time go to settings and enable AA and leave SS on
8. test and be amazed at the performance drop!
Be aware that in order to enable every single change you have to go BACK to the options menu every single time!

Case in point: geforce 7800gs on NF8 PRO - LOOK AT THE TIMESTAMPS at the screenshots - irfan attaches the date and hour at all times:
A. 1024*768 NO AA, NO SS
B. 1024-768 AA 4X, NO SS
C. 1024*768 NO AA, SS ON
D. 1024*768 AA 4X, SS ON

The screenshots taken for the tests have been done with all settings enabled - both AA and SS from the start. This will disable SS but leave AA on. Testing with SS on geforce 7800gs would have been pointless! This game was launched in October 2005 while geforce 7800gs in February 2006 - SS shadows is the killer feature that will slow down your performance tremendously. You need something along the line of gtx 260/280 to run FEAR with both AA and SS above 40 FPS at all times.

Performance recap with all the tests including Tom2D:
VIA gets a historical win; overturning every single period correct review and established knowledge K8t800 pro is the best chipset with AGP for socket 754 clearly outperforming both nforce3 and uli M1689. So the chipset choice does matter!
All tests get equal "share"in the final score; just like on socket A where nforce2 left VIA in the dust all results are equal, there are no tests with lesser weight than others!

VIA probably benefits from memory controller being moved to CPU as nforce2 had probably lower latency and small advantage of dual channel. These advantages for nvidia are gone in s754 so VIA can be victorious with better AGP implementation.

Benchmarks show it is possible to play year 2005 games with AGP but not with maximum settings - AA and performance killing features have to be off.

GTX 275 scores 76 average fps in FEAR with everything on in 1600x1200 so using better GPU does help.

Chapter 1 is done - Clawhammer has been tested. No time for VIA to celebrate as the glass of champagne is only half full! There are more generations of Athlon 64 to be tested!

AlexZ wrote on Yesterday, 07:41:

VIA probably benefits from memory controller being moved to CPU as nforce2 had probably lower latency and small advantage of dual channel. These advantages for nvidia are gone in s754 so VIA can be victorious with better AGP implementation.

Benchmarks show it is possible to play year 2005 games with AGP but not with maximum settings - AA and performance killing features have to be off.

GTX 275 scores 76 average fps in FEAR with everything on in 1600x1200 so using better GPU does help.

You are correct sir. Yes, it is absolutely possible to play 2005 games on socket 754 IF you use a PCI-express card far better than anything available in AGP.
In the far future, if time permits, I will test the younger brother of socket 754 with more pins - if you know what I mean - and I am already thinking of splitting that comparative in 2: one with AGP only and one with PCI-express only so I can use overpowered GPU to get over the video card bottleneck.

Chapter 2: Newcastle
The next generation of Athlon 64 is where AMD tried to downsized the chip by cutting the L2 cache to 512kb. In order to compensate for the loss of cache the frequency has been increased by 200 MHz for all models when compared to the Clawhammer ones. So not only we are going to test the 3 AGP platforms again but we again to test if the PR rating is correct: does a Clawhammer running at 2200 MHz with 1024kb of L2 cache is equal to a Newcastle running at 2400 MHz with 512kb or no? Does 512kb of cache = 200 MHz or no?
20 years later we already found that the old axiom: integrated memory controller makes all chipsets equal already false in the case of Clawhammer! So what about Newcastle?

All 3 systems stay the same, only the CPU is changed so only a short recap:
Abit NF8 PRO/nforce3

Unlike Clawhammer that clocks down to 800 MHz, Newcastle will only lower the multiplier to 5 so 1000 MHZ! This is important for people who want to build a all in one retro system! Regardless both CH and NC run cool and quiet - AMD stock aluminum cooler is all you need!
Those with keen eyes will immediately see the Kingston ram used: if all voltages are set on AUTO in the BIOS the board will happily take whatever modules you have. If you want 2.5V than you got to use Corsair!

Abit KU8/ ULI M 1689:

Abit KV8 PRO/k8t800 pro:

Sorry for asking.
One of us wrote that on VIA with SSD, loading only works on the second attempt.
I don't really want to scroll through 20 pages.
Did this happen with an auto-reboot or just with an error?

It happened to me on nForce 3 and Intel SSD. Windows XP froze on boot screen very early and after manual reboot offered safe mode. I selected standard boot and it booted normally. Everything worked fine in Windows once it booted. But after turning the PC off and on it repeated. There may be SSDs that will work fine with nForce 3, but one should be prepared for such issues.

I use Samsung 860 EVO on nForce 4 and it works in SATA-1 mode, not SATA-2 the chipset supports. So this is also an issue but at least it is reliable. This SSD failed to work reliably on AMD SATA controller.