My old non-mobile Barton 2500+ (Week 8/03: AQUCA 0308TPEW) and Hynix PC400 2x512MB dual channel (SPD: 3-3-3-8 at 200MHz, HY5DU56822BT-D43) would never do above ~180 FSB stably on my Asus A7N8X Deluxe Rev 2.0 board, BIOS 1.07. This seemed likely to be the memory's fault, even though that particular DIMM set has been validated to run on the newer A7N8X-E and -X and -VM boards. However, I couldn't rule out the Barton, since some early week 7-8 were known to have problems >185 FSB.
Well, I took the plunge and upgraded to a mobile Barton 2500+ (Week 8/04: IQYHA 0408XPMW) and top-of-the-line, low latency, Corsair 2x512MB TWINX1024-3200LL (SPD: 2-3-2-6 at 200MHz, tested on the A7N8X platform). The old parts were passed along down the family
Overall, system runs great ... I can get 2.38 GHz stable at CPU 1.65V (set), and 2.48 Ghz at 1.675V (the Asus mobo overvolts about ~.5V). Peak temps under full load only 45-48C (depending on ambient) with Vantec Aeroflow with AS5 after several week "burn-in". Sandra give this a PR rating of almost 3600, which isn't bad at all!!!
My one small problem: I can't take it further since I'm unable to run Prime95 stable above 198 FSB.
At 198 or lower, I'm consistently Prime95 stable at least >5 hours, repeatedly (as long as I've tried it). But >198, I fail Prime95 usually within 1-2 minutes, and always within 1 hour for sure. It's defintely not the Barton, since I can run it stably at 200 FSB if I run the mem async at 166. And I get the same type of Prime95 errors if I run the mem at 200 and the Barton async at 166.
However, I suspect it's actually my Asus mobo that's to blame.
Although it is a Rev 2.0 board, it is one of the first (I bought it before we knew about the 2.0 boards, and was pleasantly surprised to discovered it was one afterwards). I've seen comments on various discussion boards that some of the early 2.0 boards are not Nvidia Ultra 400 chipsets (i.e. still have the original 400 chipset of the v1.06 boards), and have problems running at 200 FSB. It seems 198 is a magic number for certain other holders of these early boards as well.
Another reason I suspect it's the board and not the memory is that reducing mem timings (i.e. 7-3-3-2.5, 8-3-3-3, etc.) and raising DIMM voltage (e.g 2.8V) has absolutely no effect on Prime95 stability. Also, the difference is dramatic: 198 FSB, no errors at all, 200 FSB immediate errors. My old memory got increasingly worse >180, as you would expect ... this rapid shift >198 leads to me to suspect a chipset problem with my board. I welcome your thoughts, though.
Anyway, despite that little hiccup, I can live with 198x12.5 pretty easily
. Here's my system specs and some benchies:
Mobile Barton 2500+ (IQYHA 0408XPMW) @ 192x12.5=2.48GHz, 1.675V (set, actual ~1.73V)
Corsair 2x512MB TWINX1024-3200LL @ 2-3-2-6, DIMM 2.7V
Asus A7N8X Deluxe Rev 2.0, BIOS 1.07
Vantec Aeroflow CPU cooler, Artic Silver 5
ATI RADEON 9500 PRO (stock), AGP 1.6V
WD 8M 80GB, SB Audigy 2, DVD-burner, etc, etc.
Windows XP Home SP1
Antex SX-1030 tower, with cooling fan on side panel
Antec 480W True Blue PSU (controlling 2/3 case fans)
---------------------
Scores are as follows (average of 2-3 readings):
"2500+" 166x11 (2-3-2-5) ..... "3200+" 198x11 (2-3-2-6) ..... "3500+" 198x12.5 (2-3-2-6)
PCMark CPU: 5693 ..... 6767 .... 7628
PCMark Mem: 5288 ..... 6145 .... 6427
PCMark HD: 868 ....... 869 ...... 867
Sandra04 CPU-A: 6929/2886 ......... 8212/3426 ......... 9360/3894
Sandra04 CPU-M: 17293/18472 .... 20524/21816 .... 23330/ 23914
Sandra04 Mem: 2555/2414 ......... 3036/2861 ......... 2968/2855
Sandra04 HD: 28MB/s ............... 28MB/s ............... 28MB/s
3DMark 2001SE: 12800 ...... 13650 ....... 14045
UT2K3 1024 (flyby/bot): 180.1/74.6 ..... 193.7/87.3 .... 199.4/94.1
UT2K4 1024HQ (min/avg/max): 27.4/79.6/101.6 ... 32.3/86.9/110.9 .... 35.8/94.2/119.5
(using custom UMark benchie)
------------------
So, an expected 30-35% boost in synthetic CPU scores, and 20-25% boost in real-world gaming FPS benchies. Not bad for virtually stock voltage!
Sorry for the long post, but you're looking at about 3 weeks of testing/tweaking on my part - needed to get it off my chest.
Well, I took the plunge and upgraded to a mobile Barton 2500+ (Week 8/04: IQYHA 0408XPMW) and top-of-the-line, low latency, Corsair 2x512MB TWINX1024-3200LL (SPD: 2-3-2-6 at 200MHz, tested on the A7N8X platform). The old parts were passed along down the family
Overall, system runs great ... I can get 2.38 GHz stable at CPU 1.65V (set), and 2.48 Ghz at 1.675V (the Asus mobo overvolts about ~.5V). Peak temps under full load only 45-48C (depending on ambient) with Vantec Aeroflow with AS5 after several week "burn-in". Sandra give this a PR rating of almost 3600, which isn't bad at all!!!
My one small problem: I can't take it further since I'm unable to run Prime95 stable above 198 FSB.
At 198 or lower, I'm consistently Prime95 stable at least >5 hours, repeatedly (as long as I've tried it). But >198, I fail Prime95 usually within 1-2 minutes, and always within 1 hour for sure. It's defintely not the Barton, since I can run it stably at 200 FSB if I run the mem async at 166. And I get the same type of Prime95 errors if I run the mem at 200 and the Barton async at 166.
However, I suspect it's actually my Asus mobo that's to blame.
Although it is a Rev 2.0 board, it is one of the first (I bought it before we knew about the 2.0 boards, and was pleasantly surprised to discovered it was one afterwards). I've seen comments on various discussion boards that some of the early 2.0 boards are not Nvidia Ultra 400 chipsets (i.e. still have the original 400 chipset of the v1.06 boards), and have problems running at 200 FSB. It seems 198 is a magic number for certain other holders of these early boards as well.
Another reason I suspect it's the board and not the memory is that reducing mem timings (i.e. 7-3-3-2.5, 8-3-3-3, etc.) and raising DIMM voltage (e.g 2.8V) has absolutely no effect on Prime95 stability. Also, the difference is dramatic: 198 FSB, no errors at all, 200 FSB immediate errors. My old memory got increasingly worse >180, as you would expect ... this rapid shift >198 leads to me to suspect a chipset problem with my board. I welcome your thoughts, though.
Anyway, despite that little hiccup, I can live with 198x12.5 pretty easily
Mobile Barton 2500+ (IQYHA 0408XPMW) @ 192x12.5=2.48GHz, 1.675V (set, actual ~1.73V)
Corsair 2x512MB TWINX1024-3200LL @ 2-3-2-6, DIMM 2.7V
Asus A7N8X Deluxe Rev 2.0, BIOS 1.07
Vantec Aeroflow CPU cooler, Artic Silver 5
ATI RADEON 9500 PRO (stock), AGP 1.6V
WD 8M 80GB, SB Audigy 2, DVD-burner, etc, etc.
Windows XP Home SP1
Antex SX-1030 tower, with cooling fan on side panel
Antec 480W True Blue PSU (controlling 2/3 case fans)
---------------------
Scores are as follows (average of 2-3 readings):
"2500+" 166x11 (2-3-2-5) ..... "3200+" 198x11 (2-3-2-6) ..... "3500+" 198x12.5 (2-3-2-6)
PCMark CPU: 5693 ..... 6767 .... 7628
PCMark Mem: 5288 ..... 6145 .... 6427
PCMark HD: 868 ....... 869 ...... 867
Sandra04 CPU-A: 6929/2886 ......... 8212/3426 ......... 9360/3894
Sandra04 CPU-M: 17293/18472 .... 20524/21816 .... 23330/ 23914
Sandra04 Mem: 2555/2414 ......... 3036/2861 ......... 2968/2855
Sandra04 HD: 28MB/s ............... 28MB/s ............... 28MB/s
3DMark 2001SE: 12800 ...... 13650 ....... 14045
UT2K3 1024 (flyby/bot): 180.1/74.6 ..... 193.7/87.3 .... 199.4/94.1
UT2K4 1024HQ (min/avg/max): 27.4/79.6/101.6 ... 32.3/86.9/110.9 .... 35.8/94.2/119.5
(using custom UMark benchie)
------------------
So, an expected 30-35% boost in synthetic CPU scores, and 20-25% boost in real-world gaming FPS benchies. Not bad for virtually stock voltage!
Sorry for the long post, but you're looking at about 3 weeks of testing/tweaking on my part - needed to get it off my chest.