Phenom II and oc'ing HT link/L3 cache to max of 2.6 GHz

Anarchist420

Diamond Member
Feb 13, 2010
8,645
0
76
www.facebook.com
I'm not going to do it, but how much of a performance gain clock for clock (in percentage) would it give if you increased the HT link (which is the speed of the L3 cache also) to the max of 2.6 GHz? That's a 30% increase in speed of the HT Link and L3 cache, although the cores are unaffected.

Has anyone ever Oc'd the HT link on a PII? If so, by which percentage did performance increase, and how much did you oc the HT Link/L3 cache?

I was thinking if they ever made a 32 nm PII x4 at 3.6 GHz core and 2.8GHz HT Link/L3 cache then it would be fast as ****. It would require a new socket, since AM3 only goes up to 2.6 GHz, but if they did that they could make that a really cheap alternative to bulldozer, if it ever comes out.
 

BD231

Lifer
Feb 26, 2001
10,568
138
106
HT Link = HT Link
FSB/NorthBridge Clock = Cache (and then some)

I think 240 to 260 is the best you can do @ 1.3v or so. Leave your HT link as close to 2000mzh as possible. There are no gains to be had overclocking the HT link.

FYI, PII kinna sucks when compared to the competition, if there was anything AMD could have done to make it a more competitive part they've already done it.
 
Last edited:

Anarchist420

Diamond Member
Feb 13, 2010
8,645
0
76
www.facebook.com
HT Link = HT Link
FSB/NorthBridge Clock = Cache (and then some)

I think 240 to 260 is the best you can do @ 1.3v or so. Leave your HT link as close to 2000mzh as possible. There are no gains to be had overclocking the HT link.

FYI, PII kinna sucks when compared to the competition, if there was anything AMD could have done to make it a more competitive part they've already done it.
Okay thanks:) I thought the HT Link was the Northbridge.

How much is there to gain going from from 200 (core multiplier 15.5x) to 260 (core multiplier 12x)?
 

veri745

Golden Member
Oct 11, 2007
1,163
4
81
I don't recall benchmark numbers offhand, but I remember that in certain workloads, the gains were quite large. On the order of 10% I think.
 

formulav8

Diamond Member
Sep 18, 2000
7,004
523
126
I've done my own tests and even the Non-L3 cache northbridges get gains with increased northbridge speeds. I was able to decrease memory latency and also increase bandwidth. I even have a tiny article I made that shows the benchmark tests. Although I compared northbridge speeds below stock (I started around 1.4ghz), and comparing to 2800 mhz (I specifically wanted to test northbridge results in general and not in actual stock to oced instance like showing stock 2ghz compared to 2.8ghz) it shows that performance responds to northbridge speed increases. Around 1ghz increases in nb speed got me over 10% increases in a some area's. One was a game or 2 and another was cinebench opengl rendering. Otherwise the gains were 4% to 7%. There was some increase in power consumption as well.

When it comes to having L3 cache on the northbridge then the gains per mhz will be even higher than I got on the cacheless northbridge results. But you won't be able to increase the nb clockspeed as high as a cacheless northbridge. But there are gains to be had either way.

HT on the other hand I never tested.

Anyways, I figured I would let you knowe my actual results. :)


Jason
 

SickBeast

Lifer
Jul 21, 2000
14,377
19
81
I haven't run benchmarks, but my system seems to be snappier with my CPU-NB at 2400mhz instead of the default 1800mhz. Games also seem to run a bit faster, say 34fps instead of 30fps.

I've also got my HT Link at 2400mhz up from the default 1800mhz. I've heard that this helps improve minimum framerates in games.
 

Dice144

Senior member
Oct 22, 2010
654
1
81
HT link should be 2000 default not 1800. From personal testing OCing the NB abit gives you alittle bit of a boost. I have not notice any when OCing the HT link.
 

Fox5

Diamond Member
Jan 31, 2005
5,957
7
81
Anandtech had a review a while back (I think when the amd 6 cores came out) that showed that overclocking the Northbridge gave substantial (20%-30%) performance improvements in games. That is, a 50% overclock of the northbridge gave about half that to games. Nothing else really seemed to benefit though.
 

Dice144

Senior member
Oct 22, 2010
654
1
81
I finally stopped caring about power draw so I OC my X6 1090 to 4Ghz and 2.6 NB. I got it stable at 2.8 but had to push the volts up higher then I was willing to keep it at. 8 hours Prime stable and 20 or so passes of Intel Burn Test.
 

BD231

Lifer
Feb 26, 2001
10,568
138
106
Okay thanks:) I thought the HT Link was the Northbridge.

How much is there to gain going from from 200 (core multiplier 15.5x) to 260 (core multiplier 12x)?

I have a Phenom II x3 myself and notice substantial gains overclocking the northbridge gaming like others have said, well worth it IMO. My chip was no good past 240, but I hear the C3 (PII 955's and up) chips have a more capable northbridge so you might fair better.
 

formulav8

Diamond Member
Sep 18, 2000
7,004
523
126
HT link should be 2000 default not 1800. From personal testing OCing the NB abit gives you alittle bit of a boost. I have not notice any when OCing the HT link.

It depends on his cpu. The first PI's had 1800 (3600) ht speeds (9850/9950 has 2ghz (4ghz) speeds, the AM2+ PII's also had just the 1800mhz(3600) ht speeds. AM3 has 2ghz(4ghz) ht speeds. There is a x2 5000+ AM2+ that has only 1600mhz(3.2ghz) ht speed.

I have a B53 PII (2.8ghz core) as well and it comes with a 2200mhz (4.4ghz) Northbridge instead of the typical 1800/2000mhz. I tried in 3 different mobo's and it still defaulted to the 2.2ghz nb/ht speeds. Its a non-black edition that has a max nb multiplyer @ 12x (I think). I'm not sure if this is a fluke or not.

Something else interesting is that I have a X3 740 black edition cpu and I left it oced by the multiplyers set to 3.2ghz when I took it out of the motherboard. I then put in that B53 and the multiplyers stayed @ 3.2ghz! I meant to test other cpu's to see if the multiplyer would stick or not but haven't had time. Anyways...


Jason