Thank you sir. I missed that!
Now the correct Tom2D results for Newcastle:
Abit KV8 PRO

Before wrapping up part 1 it is time for the standings:
First we are going to see how Clawhammer is performing relative to Newcastle on each chipset; for ease of comparison CH gets a 100 % base result:

Abit NF8 PRO/nforce3
CH: 100%
NC: 100.626%

Abit KU8/ULI M1689
CH: 100%
NC: 100.02%

AbitKV8 PRO/K8T800 PRO
CH: 100%
NC: 100% - yes they are a perfect match performance wise!!

So AMD performance rating is spot on! 512kb of L2 cache = 200 MHz in frequency! That also means that the performance rating of the 3700 is inflated probably because it is a halo product as the top CPU for socket 754! The correct performance rating should be 3600!

When comparing the 3 chipsets with AGP we are getting the following (VIA is getting a 100% rating):

The first generation of Athlon 64 - CLAWHAMMER

1. VIA K8T800 PRO - 100%
2. ULI M1689 - 95.43%
3 - NVIDIA NFORCE3 - 95.36%

The second generation of Athlon 64- NEWCASTLE

1. VIA K8T800 PRO - 100%
2. NVIDIA NFORCE3 - 95.90%
3 - ULI M1689 - 95.54%

So all 3 are within 5% of each other which means that from the practical point of view they are equal! VIA gets an enormous advantage in the GUI test however not enough to break past the 5% limit of error!
I made an excel file for anyone interested:

NVIDIA, usually the top performer, retreats to prepare for part 3, while the best performing AGP system is going against its older cousin in part 2!

PART 2: K7 versus K8 at the same clockspeed and amount of Level 2 cache!

Before "entering" part 2 I need to clarify 2 things:
1.There is one more generation of Athlon 64 code named Venice on socket 754. I recently bought off eBay an Athlon 64 3200 Venice and tested it against a 3200 Newcastle that i already had! I am not going to put up the result because they are within 1% of each other! So nobody should worry about a particular revision of Athlon 64 when buying one - they all perform the same!
2. Where is Sempron? There is none and it will not feature in this comparative. Prices on eBay for Athlon 64 3000 - 3200 are about the same as the prices for Semprons, in fact the top of the line Sempron 3400 cost more than Athlon 64 3400 because of its rarity. It is pointless to pay the same amount or more of money for less performance. Very few people are interested in socket 754 so prices for CPU's and motherboards is low, in fact high end socket A chips such as XP 3000 & 3200 cost more than Athlon 64 3400!

nd22 wrote on 2025-11-19, 15:20:

1.There is one more generation of Athlon 64 code named Venice on socket 754. I recently bought off eBay an Athlon 64 3200 Venice and tested it against a 3200 Newcastle that i already had! I am not going to put up the result because they are within 1% of each other! So nobody should worry about a particular revision of Athlon 64 when buying one - they all perform the same!

Performance per clock will be the same but there are other advantages:

• SSE3 support
• 90nm process and potentially better overclocking. AM2 Windsor used the same process and it ran up to 3.2Ghz. This is an older version of the same process and an older die not designed to clock as high. Still, it should be possible to run 2.4Ghz at up to 2.6Ghz without voltage adjustment and 2.2Ghz at 2.4Ghz.

After you are done with testing, perhaps you could do a little experiment - use stock voltage and try to increase CPU clock in 100Mhz increments until it is unable to boot into Windows XP or fails to post. Use DDR333 memory clock not to OC memory. Don't do any stability testing, just a quick test. I wonder what the limits will be for your samples of Venice and Newcastle. It could corrupt Windows XP so it is best to do after all testing is done.

Update from Need for Speed Most Wanted (2005) - it gets pretty hard at levels 3-4 with lot of traffic and opponents being very fast. Winning opponent's green Dodge Viper with maxxed out upgrades ahead of level made it a lot easier. It's a ram fest at this point. You win by opponents making more mistakes than you. The best car in the game seems to be Porsche Carrera GT (unsurprisingly). Lamborghini Murcielago (AWD) drives like a brick compared to the Porsche. A great car for most of the game is Porsche Cayman S. I won it from opponent, but I would recommend buying it if you don't win it. In this game almost all cars can be upgraded to be competitive until the end. You could possibly complete the game with WW Golf. My driving style matched Porsche handling perfectly, they are very responsive to change direction and easy to catch grip after skidding. You make more money in the game by racing the easiest Toll Booth repeatedly. Those are worth racing even with maximum police presence. If you get caught by police in pursuit just ctrl+alt+del and kill the process.

Need for Speed Most Wanted (2005) was very positively received and sold well.

Although many people on Vogons overclock their retro parts I am not one of them. I respect them - the results are many times extraordinary. For me long life is more important - Venice has the lowest TDP and operating voltage out of the all socket 754 processors - that means if you got a Arctic freezer like me its RPM will be even under full loud very low! With Venice the only thing you are going to hear during games is the small fan of the 7800gs!

Now in order to spice things up I am going to use the Venice 3200 when comparing against XP 3200. One thing to remember the whole time: we are going to compare the platforms and not only K7 versus K8! So we got XP 3200 on nforce2 ultra versus Athlon 64 3200 on K8T800 PRO.
In period correct reviews K8 got a significant lead over the K7; in retro reviews it got an even bigger lead because many youtubers used high end AGP graphics cards like the 7800gs.
It is not all doom and gloom for K7: Athlon XP has 2 advantages on its side!
1. dual channel memory available on nforce2 and missing completely on socket 754! And no I am not going to put the RAM in single channel on Abit AN7 just to have the same memory configuration - each platform will use the best available settings. That also means that any unnecessary controllers are disabled (serial ,parallel, ...).
2. sound storm on nforce2 is on another level compared the crappy ALC658 on all socket 754 boards that I got! Again I am not going to use a third party sound card such as Audigy 2; sound storm is part of the nforce2 ultra/MCP-T chipset just as Realtek is part of the VIA K8T800 PRO/VT8237 chipset!
All reviews - period correct or retro - tested in a "vacuum"that is they tried to isolate CPU performance as much as possible and minimize the influence of all other components! Not this time around: although I tried to use automatic testing procedures as much as possible all tests will be done in "realistic"scenarios - that is all games at maximum settings and with sound where possible; if a synthetic benchmark has a sound option than that will be tested - example 3dmark 2003.

The first system is the Athlon 64. I simply dropped in the Venice 3200 and started testing. Here are the system details: