Athlon X2 5200+ 45nm AM3 may unlock to Phenom FX5200

Uriel · 5 Jan 2011 at 10:17

Looks like there's another deneb in disguise on the loose. This time, we have an Athlon X2 5200+ 45nm, AM3 CPU. There are 2 versions about - some are regor (which are native dual core and any unlock would only give an extra 512kb per core L2 cache).

I've now seen 3 separate reports of them unlocking to Phenom FX5200s, one from UK OEM stock. The FX5200 is essentially a full deneb Phenom II X4 with an 11.5 multiplier, so 2.3 GHz and 6MB l3 cache.

I've not heard reports about stability or overclocking yet but, buoyed by my earlier success unlocking an Athlon X2 5000+ 45nm AM2+, I fancy my chances so ordered one for under £32. Will let you know how I get on.

Uriel · 5 Jan 2011 at 12:02

It's the second one. The AM2+ 45nm Athlon X2 5000+ was 2.2GHz. Didn't stop me getting it up to 3.3GHz though.

I still do wonder if these are made deliberately easy to identify with the Athlon X2 (not Athlon II) naming. Most unlocked Phenoms have fairly boring names. Getting recognised by motherboards and CPUz as Phenom FX5000+ and FX5200+ (rather than Bxx) smacks of easter egg to me.

Uriel · 5 Jan 2011 at 12:27

I checked a few benchmarks and both 45nm cpus perform similarly to their 65 nm counterparts despite the lower frequencies.

It is an odd naming scheme that could lead to consumer confusion. Saw a 45 nm 5000+ on the bay the other day as it wouldn't work in the seller's motherboard. There will be a lot of old AM2 boards that won't take these because of BIOS limitations.

Anyway - handy for me to have a spare AM3 CPU. My motherboard supports either DDR3 or DDR2 and has a separate BIOS for each one. I needed a cheap AM3 CPU to flash the DDR3 BIOS before it will take X6 CPUs and with memory prices the way they are it's getting tempting to switch my old DDR2 for DDR3. I will do if this unlocks and clocks well.

Easter eggs aside, there are two good reasons for AMD to do this. One is clearing stock thad didn't meet QA testing requirements. The other is just using existing stock to meet demand at different market segments. From what I've seen and read I think it's a mix of the two.

My X2 5000+ 45nm is a definite QA reject. Unlocked it needs some hefty voltages on the CPU-NB to get it stable and still generates 'corrected' hardware errors in Windows event viewer which are probably related to the unlocked L3 cache. Perfectly usable though - once I'd got it stable its never crashed in a stress test, benchmark or game. I do run it as dual core most of the time to save power though.

Uriel · 5 Jan 2011 at 13:46

Well there's the additional hurdle with this one that you might just get a Regor - disappointing if you wanted a core and l3 cache unlock.

Someone on HardForum did a straw poll. Of the 5000+ 12 unlocked and 3 failed. No results there on 5200+ yet. http://hardforum.com/showthread.php?t=1499697

From what I've read, early ones had a better success rate. Mine was a late one and although it unlocked, it clearly wouldn't have made it as a 940 or 920. I think many enthusiasts would have given up trying to get it stable too and it just wouldn't have been possible on a lot of motherboards.

Uriel · 6 Jan 2011 at 10:13

Someone on AMDZone has one of these unlocked currently stress testing at 3.6 GHz on stock volts. Mine arrives later today.

Uriel · 6 Jan 2011 at 19:04

It's arrived. Stepping starts CACEC AC. The two ACs mean it's a deneb. No time to install until the weekend though. Let's hope it unlocks

Uriel · 6 Jan 2011 at 22:47

That's one reason it appealed to me. My board takes DDR2 or DDR3 with a separate BIOS for each type. Needed a cheap AM3 CPU to flash the DDR3 BIOS so this won't be wasted even if it doesn't unlock. Only have DDR2 at the moment though.

Anyway here's mine:

Exactly the same batch as gamesaregood - and yes it's from the same retailer (I spotted your recent posts on another forum on Google).

Hopefully this one will be less of a pig to get stable than my 5000+. I had much more trouble than gamesaregood's first 5000+ and only got stable with some hefty voltage on the CPU-NB as I definitely have dodgy l3 cache on mine. Going to nip into my BIOS to write down all current settings before I drop in another CPU - one reason I won't be dropping this in until the weekend.

If this 5200+ matches my 5000+ without the hefty voltage, I'll be happy.

All the other unlocked 5200+s I've seen so far reported as socket AM2+ in CPUz. Just a bug.

Uriel · 6 Jan 2011 at 23:27

Well, I currently have to overvolt my RAM to get above 282MHz FSB. That means my target will be 3.2 - 3.3 GHz.

With overvolting RAM I can get to 300MHz FSB, meaning I think I'll be platform limited at about 3.4 - 3.5 GHz. Over that it's touch or go whether I can POST consistently on my current CPU.

Maybe a different CPU will get me higher but I think the 11.5 multiplier and platform are going to limit overclocking of these CPUs.

I did think about getting a Phenom X2 to try at one point but Phenom FX sounds much cooler than Phenom II B**.

Uriel · 7 Jan 2011 at 00:21

Good start there and very encouraging.

Memory prices prices played a part in my decision to try this. Although my motherboard takes either DDR2 or DDR3, it's limited to 4GB DDR2. If I wanted to go higher I would have to upgrade CPU at some point so I can use DDR3.

Uriel · 7 Jan 2011 at 22:03

It lives!

Unlocked at stock voltages and no warnings in Event viewer. Good start.

I need to update CPUz though.

Edit: That's better:

Further Edit: Just blending at 3.24GHz (2.56GHz CPU-NB). Anyone know a good way to check CPU-NB voltage? My BIOS settings are of the + or - mv increment variety and I'm not completely sure what the default should be.

Edit for the blend before bed - not bad

:

Uriel · 8 Jan 2011 at 21:00

Well there's an easy way to tell if it's a deneb without installing the CPU. The second two letters in the stepping code on the CPU's heat spreader will be AC. This is repeated after the first space. Mine is CACEC AC.

My old 5000+ that unlocked is CACVC AC.

Uriel · 8 Jan 2011 at 21:44

Dropping the voltage at. 3.24GHz rather than going for a higher clock at the moment. If this was a 955 I would be undervolting it now

Uriel · 8 Jan 2011 at 22:05

I could never get my 5000+ stable at all with 1T memory. Haven't bothered trying with the 5200+ yet.

"Look at those volts"? I think I'll be posting on that later. Can't believe how low I'm getting it.

Uriel · 8 Jan 2011 at 22:28

Have you looked at Windows Event viewer to see if it's generated any errors or warnings in connection with the BSoDs? That can be handy in finding out exactly what the problem is.

Edit: I definitely can't go below 1.2v vCore at 3.24 GHz.

Further Edit: And who's this Michael? Too far away from my username to be a typo so I'm assuming someone else.

Uriel · 9 Jan 2011 at 08:26

8 hour blend work in progress shot:

Have to say I'm surprised by the voltages and temps.

Edit: Speedfan's Temp1 is my overall CPU Temp. Individual core temps don't work when unlocked.

Uriel · 9 Jan 2011 at 21:16

My 5000+ was a pain to unlock and get stable. I lucked out practically going on a suicide run with voltages and found it worked. Kicked out far too much heat and I ended up removing my 5.25" bezels and fitting an extra vent and 120mm intake fan just to blow more air at it. I got it stable enough to Prime indefinitely but if I ever looked it Event Viewer it reminded me that the cache was knackered with one Machine Check warning (corrected hardware error) every minute it was run as a quad. On the other hand it never crashed once I'd got it stable. Must have been pulling in some serious juice though. I reckon I couldn't have done it if my motherboard hadn't been rated for 140w CPUs.

This 5200+ seems to be a breeze by comparison. Currently 21 hours into a Prime95 blend and the temps haven't got anywhere near 50°C with the case fans on low.

Uriel · 10 Jan 2011 at 00:19

24 hour blend. That'll do, at least until I get onto messing with the CPU-NB overclock.

Uriel · 12 Jan 2011 at 07:37

Slight setback last night. Had an IBT failure just before bed. Not a complete lock up. It sort of hung for a couple of seconds, carried on, then gave me failure message. CPU temperature had been peaking at 47 so I don't think that was the problem. Settings are the same as for the 24 hour prime95 blend posted earlier.

Will come back to it tonight and see if more vCore or CPU-NB voltage helps.

Should have got on to running IBT earlier I guess.

Uriel · 12 Jan 2011 at 18:13

Another 0.025 vCore and IBT 20 runs is fine.

Uriel · 12 Jan 2011 at 19:03

Not a lot. It's a C3 deneb rather than C2 so same differences as the Phenom revisions.

This seems to generate much less heat and require less voltage than my 5000+ (mine looked like it was locked for a reason). The multiplier is 11.5 rather than 11. I think I got lucky and just got a much healthier chip this time round.

Nicely done getting 3.6 GHz out of your 5000+ by the way. What have you set the CPU-NB to? The bus on my rig is limited to 300, being hit or miss about whether it'll post consistently above that so never saw how high the chip could go. Getting over 282 means I need to over-volt my RAM, which is why I'm sitting at this speed.