My AM2 Motherboard upgrade to Phenom 9600

DesertCat

Gawd
Joined
Jun 14, 2006
Messages
565
This is a post for people, similar to me, that have an AM2 motherboard and wonder about upgrading to a Phenom 9600. Yes, I know about the errata issue. Yes, I would probably recommend an Intel quad for somebody doing a completely new system build. This is the AMD motherboard forum, though, and such conversations belong elsewhere. This post simply provides info regarding the CPU upgrade path for current AM2 owners and what they might expect for results.
----------------------------------------------------------------------------
First the laundry list:

Asus M2N-SLI Deluxe motherboard (Nvidia 570 chipset) -BIOS rev 1302
4 x 1 GB Crucial Ballistix DDR2 800 - BL12864AA804
WD 150 GB Raptor HD (10,000 rpm)
WD 250 GB HD (7200 rpm)
Audigy 2 ZS Platinum soundcard (w/ front controller box)
EVGA 8800 GT Superclocked Video card (650 Mhz variant) - using Zalman VF900 + heatsinks on mofsets
Plextor PX-708A DVD burner
Antec P180 case
Seasonic S12 430W Powersupply
Vista Ultimate 64-bit version

Before: AMD X2 3800+ CPU (89W variant)
After: AMD Phenom 9600 CPU (95W)
----------------------------------------------------------------------------

I had run this system with the X2 3800+ for 1.5 years. About 8 months ago I upgraded from 2 GB up to my current 4 GB. I attempted to purchase the exact same Crucial Ballistix memory at that time and kind of did, though it is the same model number, there was a revision between batches. The system was happy so long as I kept the old memory and new memory on their own channels (i.e. new memory in DIMMs 0 & 3, old memory in DIMMS 1 & 4). Trying to mix the two different batches of memory on the same channel got things angry. With that setup I was able to run the memory at 2.2V and 4-4-4-12-2T timings (the manufacturer's specs).

I had upgraded to BIOS 1302 for approximately a week and experienced no stability issues with that X2 3800+.

With the Phenom 9600 I found that I tended to get some BSOD problems running my memory with the 2.2V 4-4-4-12-2T timings. I should emphasize that the BSOD messages were all over the place of what file or error was occurring. After lots of experimenting I've found that relaxing the timings to 5-5-5-18-2T has resulted in a stable system (5-5-5-15-2T was pretty good but still had the rare BSOD). Asus did add support for DDR2 1066 memory with the 1302 BIOS update, though I have no idea what would be a recommended brand for that.

So... it looks like I now have a stable system, but I'm hoping that Asus will be able to improve the BIOS to allow me to use more aggressive memory timings in the future (though I know the 10-20% hit for the errata fix will be coming as well).

That said, all instability problems I've encountered seem to be related to memory timings and not something I would try to pin on the well-publicized erratum 298.
--------------------------------------------------------------------------------

What kind of performance increases do I see?

3Dmark06 CPU scores on my system:
X2 3800+ CPU score = 1481, CPU1 Red Valley = 0.461 FPS, CPU2 Red Valley = 0.761
Phenom 9600 score = 3150, CPU1 Red Valley = 1.045 FPS, CPU2 Red Valley = 1.519

I didn't go nuts on gaming benchmarks since many do not support more than a core or two, I don't own some of the popular benchmarking games, and most of these games are GPU limited at reasonable gaming resolutions (though my X2 3800+ was probably starving the 8800 GT).

The big question for me was video encoding, which was where I was hoping for the most gains from a quad core. I don't do tons of it but like good performance when I do. As many who have followed the initial benchmarks on Phenom processors may have seen, the Intel processors do much better at DivX than the Phenom. What they usually don't tell you is how an AMD X2 does compared to the Phenom. Well...

DivX 6.8 encode in VirtualDub 1.7.6 32-bit version with VOB plugin. I don't use the 64-bit because there is not a VOB plugin for it yet, I have not been able to find an AC3 to MP3 codec for it, and the 64-bit version has been problematic software for me in the past. The 32-bit version works flawlessly for me.

Doing a single 1 GB VOB file (480P) with DivX 6.8 in multi-core mode in VirtualDub:
X2 3800+ takes about 26:30 minutes
Phenom 9600 takes about 20:00 minutes

Those results were a bit underwhelming as far as speed improvements. They look more like what I would expect from a fast X2 dual core. So that got me looking. When I would do a DivX encode on the X2 3800+, both of my CPU cores would be pegged at 100%. When looking at the performance meter on the 9600, my overall CPU load bounces between 50-65%. For whatever reason, DivX is NOT able to take full advantage of the Phenom 9600 on my system. This leaves me guessing at two possiblities. 1) My AM2 board/DDR2 800 just can't feed the Phenom 9600 fast enough to peg the processors, or 2) DivX needs to be optimized for Phenom somehow (the notes for DivX 6.8 have comments about decoding improvements for quads but I don't see encoding improvements). Either way, DivX encodes seem more like a fast X2 processor on my system. If it is option #2, that may explain why the Phenom processors seemed to look so horrible on the DivX tests in reviews. I don't recall any of the reviews actually saying they looked at whether any of the Phenom CPU cores were actually running at 100%. If they aren't pegged on an AM2+ system, it would suggest that at least part of the problem lies with DivX.

In contrast, Windows Media Encoder 64-bit edition got up to 90% total CPU usage when converting an AVI file to WMV for me.

All for now, probably too long as it is. Hope this is useful info for some that are considering this upgrade path.
 
Divx certainly favours Intel procs.One reason why I use it.One thing this shows me,is that AMD needs better software dev relations,Its always appreared to me,to be several steps behind Intel.Hopefully that will change in a big way,soonish. :)

Great post and certainly informative.
 
Out of curiosity I went ahead and downloaded PCmark Vantage to run another benchmark. I'll be curious how my scores compare to those using AM2+ boards (with most of the other things similar like an 8800GT).

PCmark Vantage 64-bit edition: 4620 PCmarks
Some CPU relatedsubscores from this bench:
CPU Image Manipulation: 2.61 MB/s
VC-1 to VC-1: 0.56 MB/s
Gaming Data Compression: 698.23 MB/s
CPU Gaming: 9719.5 ops/s
Audio transcode mp3 to WMA: 0.43 MB/s
Wav to WMA Lossless: 3.84 MB/s
PCmark Comm1 Encrypt: 2.64 MB/s
PCmark Comm1 Data Compress: 3.53 MB/s

Since the free edition only gives a serial number for a single benchmark, I could not run this again in the 32-bit version without paying money. Anyway, those scores may give people something to compare against and see how much my AM2 board and DDR2 800 may be bottlenecking the Phenom 9600 (if it is).
 
Phenom has a slightly different memory controller, I think that's why you are having problems with tight timing. Phenom has this ganged and unganged mode.
 
Makes sense to me on the memory controller. I've noticed on the Newegg customer comments that some people gave up and shipped their Phenoms back thinking that it was a lemon due to similar BSOD issues. I think one of the biggest problems with all of the negative press about errata 298 is that people may give up too easily when they run into any instabilities. Instead of going through normal troubleshooting procedures for a system, they think it must be a bad processor.

Still going strong here with those relaxed memory timings. :)
 
Wish I could say the same running a Phenom 9500 and M2N-SLI Deluxe motherboard with 1302 BIOS.

System runs at about half pace and has some issues. :mad:
 
Wish I could say the same running a Phenom 9500 and M2N-SLI Deluxe motherboard with 1302 BIOS.

System runs at about half pace and has some issues. :mad:

I'm not sure if this applies to your situation, but one thing I ran into in all of my tweaking was that using the "Tune System" feature of nTune was very flakey. After one pass it actually boosted my 3DMark06 score to over 11,000, though it didn't seem particularly stable to me (processor was running at 2.45 Ghz and the PCI-E was running at 2850 BTW). Trying to be more conservative, I tried to just save the memory tuning and get rid of the CPU and PCI-E overclocks, but it seemed to hose my CPU speed. My 3DMark06 score after that was down in the 6000's. After trying numerous things, I eventually uninstalled nTune. That seemed to fix the problem (getting me back in the 10,500 range). I have since re-installed nTune just so that I can get some more of the display info, but I've avoided trying to do any tuning tricks. I'm just running stock on everything with better results.
 
To follow up on my desire to compare results between AM2 and AM2+ boards, a review over at Gamespot includes some results for the 9900 and 9600 on an AM2+ board. That link his here:

http://www.gamespot.com/features/6183736/index.html?tag=topslot;title;1

Major differences I would point out are that they are using an 8800GTX for their graphics card and using DDR2 1066 memory. I'm using an 8800GT OC (650 MHz) and DDR2 800 and relaxed timings (5-5-5-18-2T). The M2N-SLI Deluxe (my AM2 board) now supports DDR2 1066 in the BIOS, but I have do not have that kind of memory to put in my board. Though my Ballistix DDR2 800 is using Micron D9 chips, it isn't the speed-binned stuff that can actually achieve 1066. Trying to run it at those settings results in a no-boot on my system.

But enough jabber:

Gamespot's Phenom 9600 on AM2+ w/ DDR2 1066 on 3Dmark06 (1280x1024): 11,247
My Phenom 9600 on AM2 2/ DDR2 800 on 3Dmark06 (1280x1024): 10,626

3Dmark06 CPU Scores on same systems:
Gamespots 9600 AM2+: 3474
My 9600 AM2: 3150

I would say the second score may give the best indication of performance differences between the 9600 on AM2 and AM2+ because it does not reflect graphics card differences. I assume that memory speed is not an issue in this test? Whether BIOS revisions would make any difference is anyone's guess.
 
is this promoting the 9600 or discounting it or what..???

i mean, lots of useful info here, but what does it all mean?

sounds like a waste of money on an "upgrade" to me honestly

even AMD themselves is openly admitting they screwed the pooch with this new core.....
 
I really have no interest in promoting Phenom 9600. It's not like I get any money put in my pocket if somebody goes and buys one. What I am trying to do is give people some frame of reference for what a Phenom 9600 upgrade will do in an AM2 system.

Virtually every Phenom 9500, 9600, 9X00 article I see out there is about putting it in a new AM2+ motherboard. That's just fine and dandy for somebody considering whether to put together a whole new system (after which they will likely go build a Q6600 or higher). As somebody interested in the AM2 upgrade path, what I don't see is anyone talking about doing it as an upgrade (other than 1 or 2 sentences saying it can be done). What I wanted to know was whether putting a Phenom on an AM2 board would badly cripple it compared to an AM2+ board. I'm guessing that others are interested in that question as well.

I've intentionally tried to avoid editorial comments but let the numbers speak for themselves. The problem is that, given all of the variables in a system (mem speed, gpu, etc.) and the different tasks that people perform (gaming, audio encoding, video encoding, etc.) I can't make a simple judgement. For people strictly interested in gaming, there isn't much to be gained by going quad core over even a mid-grade dual core, regardless of whether it is made by Intel or AMD. That may change as game designers actually start writing for more cores. For tasks that actually use all four cores, though, it is an upgrade over my X2 3800+. Though I may state that a DivX encode does not fully tax my system, it is nice that I can have that going full tilt in the background and still web surf, etc. without any noticeable delays. People can decide for themselves what they need/want.
 
Just tried my Gigabyte 690G board, no go. Running F3a BIOS, no POST progress at all- no monitor signal, no blinking keyboard lights, just fans spinning at full speed. Damn.

Edit, forgot to update. Flashed to F4, works fine now.
 
I for one thank you for this information...I have a post in this board that is asking exactly what you answered --- what is lost using an AM2 motherboard with a phenom rather than an AM2+

My note to you would be this.

Apparently AMD systems earn NOTHING by going to faster RAM clocks, so getting 1033 memory would do you no good whatsoever.

Here's a linky -

http://www.hardforum.com/showthread.php?t=1249653

What this means to me is that I think your slight performance hit as compared to the benchmarked AM2 boards has nothing to do with your memory speeds. I would suggest that you've done pretty well with the upgrade if the newest motherboards are only a slight percentage ahead of what you've got.


Now I have a couple questions as well. Does Crysis run any better? Have you tried overclocking the Phenom?
 
Have you tried overclocking the Phenom?

perhaps you dont know the current state of Phenom overclocking....it's not really something that is done much because for one, if you DO get any overclock it wont be much at all, and secondly you can't just go increasing vcore to get more because they have a tendency to pop when you do
 
perhaps you dont know the current state of Phenom overclocking....it's not really something that is done much because for one, if you DO get any overclock it wont be much at all, and secondly you can't just go increasing vcore to get more because they have a tendency to pop when you do

I haven't followed o/cing on the phenom at all....

But my BE-2400 gets from 2.3 to 2.8 on stock voltage... That's 500mhz for free...
 
I haven't followed o/cing on the phenom at all....

But my BE-2400 gets from 2.3 to 2.8 on stock voltage... That's 500mhz for free...

yeah, there are a few out there that do okay...nothing spectacular, but many are frying from overclocking attempts
 
I'm interested in this thread mostly because I've just been curious as to how the compatibility path works in reality. I've got an AM2 FX-62 that I have no intention of upgrading, but the concept of a "side-grade" rather than a full-blown upgrade was interesting to me.
 
I haven't kept up much with the Phenom overclocking either, but what? Frying with more vcore? I haven't heard of CPU's actually frying for quite a few years now. Is this over at XS or what? Where have Phenom's been getting destroyed?
 
I appreciate the comments and links regarding tests with higher memory speeds.

I have tried some "lazy" overclocking on my Phenom but have really not tried to push it. Using ntune's coarse tuning I was able to get my Phenom 9600 up to 2.45 GHz, but that's basically just from raising the FSB from to 213. I haven't tried any voltage bumping or anything like that.

Going with the Phenom at 2.45 GHz and my PCI-e bus clocked up to 2850, my 3DMark06 score went from about 10,600 (stock) to 11,200.

With my 4 sticks of DDR2 already operating at 2.2V, I found that things were getting pretty warm in that area of the motherboard (41-43 degrees C, never noticed this much heat there before). That might explain some of my stability issues with faster timings. I "Jimmy rigged" an 80mm Nexus fan to blow over the memory and it knocked my "Motherboard Temp" down to 33 degrees under load. I may have to fiddle around with things some more in the overclocking area now.
 
Gamespot's Phenom 9600 on AM2+ w/ DDR2 1066 on 3Dmark06 (1280x1024): 11,247
My Phenom 9600 on AM2 2/ DDR2 800 on 3Dmark06 (1280x1024): 10,626

3Dmark06 CPU Scores on same systems:
Gamespots 9600 AM2+: 3474
My 9600 AM2: 3150

I would say the second score may give the best indication of performance differences between the 9600 on AM2 and AM2+ because it does not reflect graphics card differences. I assume that memory speed is not an issue in this test? Whether BIOS revisions would make any difference is anyone's guess.

In my experience both the bios revisions and ram speed would have an impact.

how big an impact is anybodies guess.
 
perhaps you dont know the current state of Phenom overclocking....it's not really something that is done much because for one, if you DO get any overclock it wont be much at all, and secondly you can't just go increasing vcore to get more because they have a tendency to pop when you do

Still waiting for information about phenom's going pop.
 
To follow up on my desire to compare results between AM2 and AM2+ boards, a review over at Gamespot includes some results for the 9900 and 9600 on an AM2+ board. That link his here:

http://www.gamespot.com/features/6183736/index.html?tag=topslot;title;1

Major differences I would point out are that they are using an 8800GTX for their graphics card and using DDR2 1066 memory. I'm using an 8800GT OC (650 MHz) and DDR2 800 and relaxed timings (5-5-5-18-2T). The M2N-SLI Deluxe (my AM2 board) now supports DDR2 1066 in the BIOS, but I have do not have that kind of memory to put in my board. Though my Ballistix DDR2 800 is using Micron D9 chips, it isn't the speed-binned stuff that can actually achieve 1066. Trying to run it at those settings results in a no-boot on my system.

But enough jabber:

Gamespot's Phenom 9600 on AM2+ w/ DDR2 1066 on 3Dmark06 (1280x1024): 11,247
My Phenom 9600 on AM2 2/ DDR2 800 on 3Dmark06 (1280x1024): 10,626

3Dmark06 CPU Scores on same systems:
Gamespots 9600 AM2+: 3474
My 9600 AM2: 3150

I would say the second score may give the best indication of performance differences between the 9600 on AM2 and AM2+ because it does not reflect graphics card differences. I assume that memory speed is not an issue in this test? Whether BIOS revisions would make any difference is anyone's guess.

Yeah First post.

I found another review of the phenom 9600 that has the exact same CPU result as your am2 system.

system:
Processor: AMD Phenom 9600 (2.3GHz)
Motherboard and Chipset: ASUS M3A32 MVP Deluxe (790FX chipset) (AM2+)
Memory: 2 x 1GB DDR2 800
Graphics: Dual Radeon 3850 (256MB) cards in CrossFire, driver rev. 8.43

0,1425,i=195344,00.gif


So if i look at this i dont see any slowdown on am2 with the 3dmark-cpu test)
 
Very interesting indeed. Thanks for spotting that review and those results.

Most of the stuff I've seen about HT3 makes it sound like its biggest payoff for many people would be communication between the CPU & graphics card. A decent graphics card shouldn't place that much of a load on the CPU but integrated graphics could. Not that anybody on these boards would probably use integrated graphics with a phenom.
 
Back
Top