VOGONS


Battle of the platforms: socket 939!

Topic actions

Reply 600 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

Again there is no difference between 4200 and 4400 in the CPU scores - cache size does not matter - but we got a nice progression in the memory scores even if below the PR of these processors! Graphics score have nothing in common with the ratings of the X2 again!

31. TOM2D - the only GUI test in the entire suite. Unfortunately nforce4 desktop acceleration is broken regardless of the nforce and forceware version and for the first time we got negative scaling as 4200 performs just below 3800. It is the only test so far where cache size heavily influences the results: on one side we got the 2 models with 2*512kb of L2 cache and on the other the 2 models with 2*1mb of L2 cache.
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 601 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

32 & 33 & 34. Finally we reached the games and the first one is FARCRY. It is only single threaded so the dual core X2 can not be put to good use, still it is interesting for testing and comparing with the scores obtained with the single cores Athlon 64's.
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 602 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

We got an excellent progression as we go up in frequency an cache size! 4200 performs like a 4100, 4400 like a 4300 and 4800 like a 4500. Again the difference in frequency dos not justify the difference in the performance rating!

35. Half life 2 is the second game. I used the steam version that is far more difficult to run on period correct systems than the original release. Still the Athlon 64 and GTX 285 combo have no difficulty running this game at maximum details, including 6X MSAA.
1024*768:
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 603 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

36. Half life 2 - 1280*1024
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 604 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

37. Half life 2 - 1600*1200:
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 605 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

We are completely CPU-limited across all resolutions which is a good thing as we want to see how the different Athlon 64 X2 models stack up! There is almost perfect linearity - each model scales according to its frequency and cache in Half life 2! The problem remains: 200 MHz frequency increase translates to only ~250 PR points. The exception is 4800 which resembles a 4600! That suggests the frequency bump of the missing 4600 model is offset by the smaller cache and its performance should be around that of the 4400 model!

Reply 606 of 616, by SlowA

User metadata
Rank Newbie
Rank
Newbie

Thank you for posting this. Am I reading it correctly? OCed 3800 to 2.4/2.5 can be good enough if I cannot find 4800?

Reply 607 of 616, by AlexZ

User metadata
Rank Oldbie
Rank
Oldbie

Get a CPU with 1MB L2 cache, ideally s939 Athlon 64 X2 4400+, OC to 2.4Ghz if needed. It will be very hard to find though and you may have to wait for years.

If you cannot find it, get an AM2 board with Windsor 5200+ (2.6Ghz 1MB L2, ADA5200CSBOX, ADA5200IAA6CS). It will be replacement for s939 Opteron 185 (2.6Ghz X2). 2x 2GB DDR2 800 for 4GB RAM and CR 1T. This is a super cheap alternative as nobody wants AM2. Pair it with GTX 285.

Pentium III 900E,ECS P6BXT-A+,384MB,GeForce FX 5600, Voodoo 2,Yamaha SM718
Turion 64 MT-40@2.4Ghz,Gigabyte GA-K8NE,2GB,GeForce GTX 275,Audigy 2ZS
Phenom II X4 965,Gigabyte GA-MA770-UD3,8GB,GeForce GTX 780
Vishera FX-8370,Asus 990FX,16GB,GeForce GTX 980 Ti

Reply 608 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

Actually socket 939 X2 are readily available.
The problem is price: https://www.ebay.com/itm/800792226332?_skw=am … ABk9SR6DP5Nm_aA

I got all the models you see in this screenshots ~10 years ago where they were in the range of 30-40 euros. Just now I have searched my purchase history on EBay and I payed 60 euros including shipping for Athlon 64 X2 4800 back in 2018.
I have not posted all the screenshots but the general idea is: 200 MHz = 200 PR points; 512KB of cache = 200 PR points exactly the same as it is with the single cores.

Reply 609 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

38 & 39 & 40. Doom 3 is the first OPENGL game to be tested. All OPENGL games are based on IDTECH4 engine! Single threaded, at high quality with advanced settings all enabled is playable on a geforce 7000 series such as 7800gs. The major performance hit comes courtesy of anti-aliasing. It is the first game that allows a 16X settings for AA. In order to not get a video card bottleneck I selected only 8X AA as the maximum settings in all games tested - that results in the CPU being the limiting factor:
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 610 of 616, by AlexZ

User metadata
Rank Oldbie
Rank
Oldbie

To get a good price, one needs to settle for a lower model like AMD Athlon 64 X2 4200+ (https://www.ebay.com/itm/198794799542 ).

Pentium III 900E,ECS P6BXT-A+,384MB,GeForce FX 5600, Voodoo 2,Yamaha SM718
Turion 64 MT-40@2.4Ghz,Gigabyte GA-K8NE,2GB,GeForce GTX 275,Audigy 2ZS
Phenom II X4 965,Gigabyte GA-MA770-UD3,8GB,GeForce GTX 780
Vishera FX-8370,Asus 990FX,16GB,GeForce GTX 980 Ti

Reply 611 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

That offer is exceptionally good! I have not seen socket 939 X2 for sale at such low price in a long time!

Doom 3 loves cache! The difference between the 2*512kb L2 of the 4200 and the 2*11n L2 of the 4400 shows us that they are separated by 4.71% which translates in exactly 200PR points! The frequency scaling is exactly half of AMD performance rating: 4200 is 5.66% better than 3800 which fits to a hypothetical Athlon 64 X2 4000.

41& 42 & 43. Splinter cell chaos theory is the first game from 2005 to undergo testing. The game only allows you to use AA if you also use shader model 1.1! I tested SM 3.0; the game is very easy to run for both the CPU and the video card.
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 612 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

There is roughly 6-7% separating the 4 Athlon 64 X2 models tested. That corresponds to about 250 PR points; again the difference between 4400 and 4800 is the same as the difference between 3800 and 4200 - that means the missing 4600 performs about the same as the 4400!

44 & 45 & 46. Serious Sam 2. Released in October 2005 it is one of the first game to use 2 threads/cores. I used HardwareOC SS2 becnhmark tool for accurate and repeatable results. The game is very easy on the GTX 285 - we are CPU limited again.
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 613 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

Again the Athlon 64 X2 scales far below the official AMD PR!

47 & 48 & 49. Quake 4 is the toughest game to come out in 2005. Extremely unforgiving on processor and the video card, it eats RAM and VRAM for breakfast! Based on an improved and enhanced version of IDTECH4 engine it is the only game where GTX 285 is actually limiting performance besides the Athlon 64 X2! So we are both CPU and GPU limited! Only with the 4800 we can see some difference in the average FPS across all 3 resolutions.
There is a common belief that using video cards with more than 1gb of VRAM will result in an image with low quality textures. That happened to me when I tried to see what card would be required in order to play Quake 4 at maximum details - that means ultra quality, advanced settings all ON, AA 16X. So I put in the system the newest card I got: GTX 560 TI. Because it requires newer drivers than GTX 285 I installed the latest driver - from what I remember. Of course all I got was low textures that required a fix found online.
So I tried different cards and among them was radeon 4890. Because I got Saphire 4890 2gb I installed and used catalyst 10.2 driver from 2010. I specifically used 10.2 because I know this is the best driver for radeon 9000, X and X1K series. And I got "normal"textures, nothing looking horrible. Of course the poor radeon was crying for mercy the entire time as it could not run the game above 30 FPS at max details.
So the problem is also driver related not only VRAM related.
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 614 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

We need a higher clocked Athlon 64 X2 to really reach an average of 60 FPS! As for GTX 285 - it can run the game but only just! 8x AA is already a bit too much in this game even for this card!

50. Next is FEAR. Released in October 2005 just like Serious Sam 2 and Quake 4 it is a highly atmospheric FPS that I replay once every couple of years. It benefits a lot from a good sound card such as Audigy 2ZS. Here I tested it using the in-game benchmark with the AC97 codec that is part of nforce4.
1024*768:
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 615 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

51. FEAR - 1280*1024
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800

Reply 616 of 616, by nd22

User metadata
Rank l33t
Rank
l33t

52. FEAR - 1600*1200:
A. Athlon 64 X2 3800
B. Athlon 64 X2 4200
C. Athlon 64 X2 4400
D. Athlon 64 X2 4800