Sign in with
Sign up | Sign in
Your question

Anandtech + Phenom 9850BE + SB750 = 3.2ghz

Last response: in CPUs
Share
July 21, 2008 12:37:23 PM

UPDATED: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=336...

http://www.anandtech.com/weblog/showpost.aspx?i=471

Quote:
After a few tweaks and reboots, we reduced our core voltage from 1.4250V to 1.3875V, increased our NB speed from 2.2GHz to 2.4GHz, and changed our clock multiplier from 15.5 to 16 giving us a 3.2GHz core clock speed. This is an instant 100MHz increase in core speed while decreasing core voltages by 3%, increasing NB clock speed by 9%, and improving memory latencies up to 6% (all important L3 cache). These results were completed with ACC on set and forget auto settings.

We have reached 3.3GHz on this CPU with further tuning of system settings including individual core tuning with ACC. While are initial 100MHz speed increase may not sound impressive, we must consider the fact that this CPU was right on the ragged edge of stability at 3.1GHz. We are now able to run higher NB clock speeds and higher core clock speeds at lower voltages while maintaining a fair amount of headroom for further performance improvements. Even more impressive (results tomorrow) is that fact that this CPU will now clock to 3GHz on stock voltages instead of 2.7GHz before tuning with ACC.


Looking good so far. Hope they make it with Captain Dirk.
July 21, 2008 12:56:00 PM

Good news; I wonder how much of a limiting factor the NB clock is. I seem to remember reading somewhere that it was a significant bottleneck for the Phenoms. Maybe 45nm will help in this regard.

Hah, found it: http://techreport.com/articles.x/14093/2
Quote:
As you may know from this page of our Phenom review, among other sources, the Phenom architecture's memory access latency and overall performance is affected quite a bit by the speed of the chip's L3 cache. That cache runs at the speed of the chip's north bridge, not the CPU core clock, so there's some real performance potential hiding in the Phenom's north bridge clock.
July 21, 2008 12:57:07 PM

dattimr said:


Looking good so far. Hope they make it with Captain Dirk.


Captain Dirk "Open the Duron bays, and lets wreck havok with the Thunderbirds"

Good old, geek jokes. For our own good i hope it is successful.
Related resources
July 21, 2008 1:13:08 PM

radnor said:
Captain Dirk "Open the Duron bays, and lets wreck havok with the Thunderbirds"

Good old, geek jokes. For our own good i hope it is successful.


hahahaha!!! i remember the days, brother!
maybe its just me, but i can't help but feel a nudge more hopeful with dirk at the wheel. and if sb750 turns out to be even half of what they said it would...looks like a good time for him to've taken the reins. this reflects well for the green camp!!!
now lets see those 780g boards come down in price around 10-15 bucks!!!
July 21, 2008 1:45:52 PM

I would like to see some video encoding benchmarks. That is the main purpose for my next build.

July 21, 2008 1:58:44 PM

I hope they do a review of the system with a core2quad at 3.2ghz as well. I am fully aware which system would be faster, but I would like to see the results side by side for reference at the higher clock speeds. Nice to see a shed of light (small though it be) from the green guys (even though its more or less Anand producing the results lol) atleast we can see that there is something positive churning about.

Best,

3Ball
a c 126 à CPUs
July 21, 2008 2:15:49 PM

radnor said:
Captain Dirk "Open the Duron bays, and lets wreck havok with the Thunderbirds"

Good old, geek jokes. For our own good i hope it is successful.


WOOT!!!!!!!! Talk about old school.

cah027 said:
I would like to see some video encoding benchmarks. That is the main purpose for my next build.


3Ball said:
I hope they do a review of the system with a core2quad at 3.2ghz as well. I am fully aware which system would be faster, but I would like to see the results side by side for reference at the higher clock speeds. Nice to see a shed of light (small though it be) from the green guys (even though its more or less Anand producing the results lol) atleast we can see that there is something positive churning about.

Best,

3Ball


^Agreed. I unfortunately hope people understand that if there is a review then you have to take OCing into account especially if you are considering this or a C2Q. My only question is whats the power drain and does the TDP raise along with the heat.

I am also pretty sure that certain CPUs will best it but a few tests would be nice. What sucks though is for those who bought a AM2 board just for the upgrade path and they wont be able to OC as easily like those who are willing to buy a new mobo if they had one before or are building a new system.
July 21, 2008 2:41:33 PM

they probably didnt release the temps cause they are through the roof. i wouldn't be too surprised if they are liquid cooling it.
a c 126 à CPUs
July 21, 2008 2:46:01 PM

^LMAO

Thats funny right there.....

I don't think so but still the voltage isn't that bad. But they need to Prime95 stability test it for 8-12 hours before I believe its stable. Sorry but I cannot trust a OCing system made by the CPU company.
July 21, 2008 2:46:07 PM

Just asking, since I'm not an expert at all: don't those memory benchmarks look "strange" if compared to C2Q's? The latency looks a bit high - and the read and write values a bit low -, if taking into consideration that it uses an IMC.
July 21, 2008 2:48:10 PM

Quote:
Even more impressive (results tomorrow) is that fact that this CPU will now clock to 3GHz on stock voltages instead of 2.7GHz before tuning with ACC.


In this scenario I don't think that temps will be an issue. However, above 3.2...
a c 108 à CPUs
July 21, 2008 2:51:24 PM

The 3.2GHz - 3.3GHz OC was with a 9850BE - - not the 9950BE
July 21, 2008 2:54:47 PM

Wisecracker said:
The 3.2GHz - 3.3GHz OC was with a 9850BE - - not the 9950BE


Sorry and thanks about that. Fixed. =)
July 21, 2008 3:16:22 PM

dattimr said:
Just asking, since I'm not an expert at all: don't those memory benchmarks look "strange" if compared to C2Q's? The latency looks a bit high - and the read and write values a bit low -, if taking into consideration that it uses an IMC.

They look about right when considering the castrated NB (IMC and L3) clocks. Like I said earlier, this is something that will hopefully be addressed with Deneb.
a b à CPUs
July 21, 2008 4:13:17 PM

Funny thing really that the southbridge has such a significant effect on performance...
I'm really looking forward to the part two tomorrow :) 
July 21, 2008 5:57:16 PM

Kari said:
Funny thing really that the southbridge has such a significant effect on performance...
I'm really looking forward to the part two tomorrow :) 


More of an issue of the fact that the clock generator on the sb600 and sb700 were craptastic, too much fluctuation and spikes on a core that is very sensitive to such things. Would be tolerable at stock speeds but causes lock ups and crashes while OCing. Looks promising, and if it does what it is suppose to, I'll be picking up a new 790fx board, or possibly a 790gx board soon.
July 21, 2008 6:08:30 PM

Hmm, so we know that Phenom can go that high, anyone seen any OC reviews of the nVidia 700 series boards for AMD?
a c 108 à CPUs
July 23, 2008 5:50:43 PM

I just wanted to bump this thread because I'm tired of looking at that 'Phenom Flaky 3rd core' crap. That ... and because Anand & Gary have yet to follow up on their Phenom preview.

B-Unit said:
Hmm, so we know that Phenom can go that high, anyone seen any OC reviews of the nVidia 700 series boards for AMD?


No.

And it keeps getting harder to find reasonable, consistent and objective reviews of motherboards (and everything else). AT has been promising an mATX roundup for 6 months or more and they keep delaying it. The only conclusion is that they won't run it until G45 comes out (maybe next month?).

I got no faith in the 'lot' of them anymore.


PS - I found this ...

NFORCE 780a SLI Motherboard Comparison : A Scooter with a Rocket Engine?
http://www.tomshardware.com/reviews/nvidia-motherboard-...
July 23, 2008 6:07:53 PM

jimmysmitty said:
Sorry but I cannot trust a OCing system made by the CPU company.


Talk about f**ked up reasoning!!!

What interests do you think the CPU company would have in your over-clocking failing?

They are selling the overclocking parts as Black Editions anyway - and can choose to charge more for them if they wish.
July 23, 2008 10:38:39 PM

the article has been updated. not yet a full review, but things are certainly looking good, if not a bit confusing...

this is going to be fun. my blackie 5k hit 3.2 easy...just upping multi. but thats just too easy, and a little ho-hum. anyone who can apply thermal paste can oc that chip.
but with this level of tweaking...sheesh! i mean, even the cats at anandtech are still testing and sorting out the myriad ways one can squeeze the juice outta her!!! thats like making the iron chefs all battle eachother, with an ingredient that none of them has even heard of before...

good stuff, this. can't wait to see the full reports...
July 24, 2008 10:21:54 AM

I'm particularly interested in seeing performance reviews focusing on the impact of a faster IMC.

If, as many suspect, the IMC is starving the processor, and AMD fix this in K10.5, we could see the 10-15% IPC (since that is relative to core clock) improvements AMD talk about.
July 24, 2008 12:56:19 PM

Here's the new link: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=336...

Anyway, I found quite disappointing that they didn't show any benchmarks with the improved NB/HT frequencies. Perhaps that could show the gains of a "true quad-core design"?

I'm Intelish by now, but I must confess that I have some hopes for AM3.
July 24, 2008 1:07:45 PM

dattimr said:
Here's the new link: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=336...

Anyway, I found quite disappointing that they didn't show any benchmarks with the improved NB/HT frequencies. Perhaps that could show the gains of a "true quad-core design"?

I'm Intelish by now, but I must confess that I have some hopes for AM3.


Already read it, and i guess, they were sure they were Overclocking but wasn't sure what they were doing. We shall see in the next Article i guess. The best results they gto was in "auto" mode. So, they will still tinker with it a it more, than write about true conclusions.

I'm looking forward also.
a c 126 à CPUs
July 24, 2008 1:10:32 PM

Amiga500 said:
Talk about f**ked up reasoning!!!

What interests do you think the CPU company would have in your over-clocking failing?

They are selling the overclocking parts as Black Editions anyway - and can choose to charge more for them if they wish.


I wouldn't trust a in Windows OCing program from Intel stating its stable either. I would prefer Prime95 since it fully stresses the system.

Sorry but thats just my own feelings. If you feel its stable based on that then thats you..
July 24, 2008 1:18:20 PM

radnor said:
Already read it, and i guess, they were sure they were Overclocking but wasn't sure what they were doing. We shall see in the next Article i guess. The best results they gto was in "auto" mode. So, they will still tinker with it a it more, than write about true conclusions.

I'm looking forward also.


Man, I hope you are right, however, since they (and every major reviews site) don't even have a damn article comparing all the Wolfdales/Yorkfields, I doubt they'll benchmark Phenom again (at least until Deneb comes out).

I'm looking forward too. : S

(PS: Actually, Tech Report had an article about the Q9300 which almost included all those CPUs)
July 24, 2008 1:19:47 PM

jimmysmitty said:
I wouldn't trust a in Windows OCing program from Intel stating its stable either. I would prefer Prime95 since it fully stresses the system.

Sorry but thats just my own feelings. If you feel its stable based on that then thats you..


Agreed. The BIOS is still the godfather of overclocking. =P
July 24, 2008 3:02:40 PM

dattimr said:
Man, I hope you are right, however, since they (and every major reviews site) don't even have a damn article comparing all the Wolfdales/Yorkfields, I doubt they'll benchmark Phenom again (at least until Deneb comes out).

I'm looking forward too. : S

(PS: Actually, Tech Report had an article about the Q9300 which almost included all those CPUs)


They aren't benchmarking phenoms. They are benchmarking the sb750. So i guess they will rip the crap of those phenoms.
Ill buy a mobo in Decembre, so this reviews feel quite right.
July 24, 2008 4:55:15 PM

radnor said:
They aren't benchmarking phenoms. They are benchmarking the sb750. So i guess they will rip the crap of those phenoms.
Ill buy a mobo in Decembre, so this reviews feel quite right.


Yeah, you're right: the main point of the article is the SB750. But, if it allows a 100-200 mhz increase in the core frequency AND a 400-600 mhz in the NB/HT frequencies, what's the point of knowing that if you don't really know what that means in the "real world"? I mean, perhaps we'll see a 5-8% increase in overall performance, which can be irrelevant for some people who were expecting to upgrade, OR it can show a 10-15% - or even more - increase, which would make it quite a fair investment. That's my point: if you don't know what are the results of such an increase, how can you decide if it's worth?

You can't - unless you are going for bragging rights and the latest tech.
a c 108 à CPUs
July 24, 2008 6:04:10 PM

dattimr said:
Agreed. The BIOS is still the godfather of overclocking. =P


The Godfather may soon be sleeping with the fishes :lol: 

By my nature I would tend to agree with you but AMD Overdrive is slick and more importantly, getting better.

Anand got 3GHz+ at stock volts with the Phenom with the IMC at 2600MHz. This was unheard of a few months ago. Additional tweaking with AOD got 3.4GHz (granted - this was at 1.47v).

There are over 50 RAM adjustments in AOD. The article notes over 600 different settings that they themselves tried. There are alot more to go.

From what I can tell, the more advanced the BIOS the more capable AOD becomes. I hope they continue to expand the capabilities of AOD.

July 24, 2008 6:07:49 PM

Wisecracker said:
The Godfather may soon be sleeping with the fishes :lol: 

By my nature I would tend to agree with you but AMD Overdrive is slick and more importantly, getting better.

Anand got 3GHz+ at stock volts with the Phenom with the IMC at 2600MHz. This was unheard of a few months ago. Additional tweaking with AOD got 3.4GHz (granted - this was at 1.47v).

There are over 50 RAM adjustments in AOD. The article notes over 600 different settings that they themselves tried. There are alot more to go.

From what I can tell, the more advanced the BIOS the more capable AOD becomes. I hope they continue to expand the capabilities of AOD.


Agreed! =P However, there's still the EFI thing... Don't know for sure how much of an improvement it would offer in this area.
July 24, 2008 7:41:19 PM

dattimr said:
UPDATED: http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=336...

http://www.anandtech.com/weblog/showpost.aspx?i=471

Quote:
After a few tweaks and reboots, we reduced our core voltage from 1.4250V to 1.3875V, increased our NB speed from 2.2GHz to 2.4GHz, and changed our clock multiplier from 15.5 to 16 giving us a 3.2GHz core clock speed. This is an instant 100MHz increase in core speed while decreasing core voltages by 3%, increasing NB clock speed by 9%, and improving memory latencies up to 6% (all important L3 cache). These results were completed with ACC on set and forget auto settings.

We have reached 3.3GHz on this CPU with further tuning of system settings including individual core tuning with ACC. While are initial 100MHz speed increase may not sound impressive, we must consider the fact that this CPU was right on the ragged edge of stability at 3.1GHz. We are now able to run higher NB clock speeds and higher core clock speeds at lower voltages while maintaining a fair amount of headroom for further performance improvements. Even more impressive (results tomorrow) is that fact that this CPU will now clock to 3GHz on stock voltages instead of 2.7GHz before tuning with ACC.


Looking good so far. Hope they make it with Captain Dirk.


Quote:
AMD says that the settings tweak doesn't impact performance and doesn't change thermals or voltages, it simply can allow a Phenom processor to clock higher when overclocking. The BIOS exposes the parameter being changed, which AMD refers to as the Advanced Clock Calibration (ACC) value. Typically this value has a range of -2 to 0, on motherboards with the SB750 that support ACC the value can be set from -12 to +12. Higher numbers should allow for higher clock speeds, while lower values should allow for lower voltages/lower power operation.


Thats needs to be explained further are they looseing their signal quality to allow for higher overclocks or they loosening memory timings? There is something amiss here.

Word, Playa.
a c 108 à CPUs
July 24, 2008 8:47:31 PM

spud said:


Thats needs to be explained further are they looseing their signal quality to allow for higher overclocks or they loosening memory timings? There is something amiss here.

Word, Playa.


As I understand it - - the opposite.

SB750 tightens and improves PLL timings and reduces voltage requirements - hence, 3GHz and 2600MHz IMC at stock volts for the Phenom 9850BE.
July 24, 2008 11:51:37 PM

Wisecracker said:
As I understand it - - the opposite.

SB750 tightens and improves PLL timings and reduces voltage requirements - hence, 3GHz and 2600MHz IMC at stock volts for the Phenom 9850BE.

I don't know... there is something strange going on here. AnandTech should be able to give us the real details on this soon enough; hopefully whatever AMD is "relaxing" to allow these processors to clock higher with lower voltages doesn't have any unforeseen consequences..
a c 126 à CPUs
July 25, 2008 3:20:35 AM

^Do you think they may cause the Resonance Cascade??????? :o 
July 25, 2008 3:38:33 AM

^No, but I wouldn't rule out a portal storm ;) 
July 25, 2008 5:34:21 AM

yeah finally amd+ati came back to the race
July 25, 2008 9:23:27 AM

dattimr said:
Yeah, you're right: the main point of the article is the SB750. But, if it allows a 100-200 mhz increase in the core frequency AND a 400-600 mhz in the NB/HT frequencies, what's the point of knowing that if you don't really know what that means in the "real world"? I mean, perhaps we'll see a 5-8% increase in overall performance, which can be irrelevant for some people who were expecting to upgrade, OR it can show a 10-15% - or even more - increase, which would make it quite a fair investment. That's my point: if you don't know what are the results of such an increase, how can you decide if it's worth?

You can't - unless you are going for bragging rights and the latest tech.


Nope, honestly im just curious. It has the potential, now lets see if it delivers :)  The 9850 BE is a good CPU, just too expensive atm. so ill wait till Dec/Jan and then decide depending on price. Anyway, it will be an Upgrade from my 4800+ X2.
Ofc i already got the cooler/case and PSU for it.
July 30, 2008 1:33:47 AM

radnor said:
Nope, honestly im just curious. It has the potential, now lets see if it delivers :)  The 9850 BE is a good CPU, just too expensive atm. so ill wait till Dec/Jan and then decide depending on price. Anyway, it will be an Upgrade from my 4800+ X2.
Ofc i already got the cooler/case and PSU for it.


unfortunately, the 9850BE is being phased out and replaced with a non-BE 95w part sometime in q3. hope BE's are still around come dec/jan...
we might see a reduction around the last gasp of them, a'la the 5000BE just before they ran stock out to make room for the new 5400BE. if it hits $150-175ish, i'd buy it. and quickly, cause that means the dead last of them, for sure.
a c 126 à CPUs
July 30, 2008 1:39:59 AM

^They are phsing it out that fast? Ouch. They need to at least just replace it. Or do they plan on only one BE thats a high end and priced to fit that area?
July 30, 2008 5:30:26 AM

jimmysmitty said:
^They are phsing it out that fast? Ouch. They need to at least just replace it. Or do they plan on only one BE thats a high end and priced to fit that area?


Well, they've already got the 9950be out to replace it, granted at a much higher TDP. Pretty much at the same price the 9850be was for a long time. The sad part is, going by the older roadmaps, the 9950be was actually suppose to have been the socket AM2+ version of the Phenom FX 81/82 (can't remember original number). I wouldn't worry too much about the phasing out though. They'll still likely have an inventory of them around for some time. After all, the 9600be was suppose to have been phased out when the b3 based 9850 came out, but you can still get ahold of one on most etailers. Not to mention once Shanghai/Deneb release it will introduce lower wattage version of all the phenom procs.

I'm actually kinda interested in what kind of OC the sb750 can produce with the 9600be. Even at the same frequencies it has a lower TDP than the 9850/9950 because of the lower voltage requirements on the IMC and cores. Either that or I'm interesting in trying to see if I can get my 9850 to run below stock voltages on the newer boards. Theoretically it should be possible, as there shouldn't be much difference between the 9850 and 9600 other than the TLB fix.

And on another note, another guess I have would be that it may appear that what is actually happening with those acc settings revolves around the cache. I noticed in some of the Everest memeory benchmarks, cache latency on the L2 and L3 went up very slightly with the ACC enabled, a bit higher than what they should of been at those clocks. Also RAM latency is slightly higher than it should be at the higher IMC clocks as well. Mine on DDR2 800 5-5-5-12 timings was running a bit above 47ns with the IMC at 2.4ghz. Now if I did the same tests with the DDR2 1066 RAM I have now it would be lower, as at stock latency is current a little over 51ns.
a c 126 à CPUs
July 30, 2008 1:47:35 PM

^Theoretically it should be able to run at a lower voltage stock much like the Q6600 G0s can (like mine at stock was only on 1.118v :o  ).

But I guess you will have to test it and post your results so we can see what happens. If so that would be nice for people who just want a stock quad core as it will lower the heat output.

For the 9600BE I would be interested to see if it can OC as well as the 9X50BE chips or if the SB750 only works best with the 9X50BE chips....
July 30, 2008 2:05:55 PM

jimmysmitty said:
^Theoretically it should be able to run at a lower voltage stock much like the Q6600 G0s can (like mine at stock was only on 1.118v :o  ).

But I guess you will have to test it and post your results so we can see what happens. If so that would be nice for people who just want a stock quad core as it will lower the heat output.

For the 9600BE I would be interested to see if it can OC as well as the 9X50BE chips or if the SB750 only works best with the 9X50BE chips....


You were able to run your Q6600 at 1.118v, jimmy? O: What about overclocking? Have you tried? : P

Well, as I stated before, I would like to see benchmarking sessions. I wonder if AMD isn't changing 6 for half a dozen, like decreasing the value of some critical setting in order to achieve higher clocks, but with the same performance as before. Mathos' statement can be seen as an evidence of this possibility. One example of that would be overclocking Intel chipsets, since the "Performance Level" is usually increased by default when you use high FSBs, thus negating most benefits of overclocking - supposing you don't decrease it manually.
a c 126 à CPUs
July 30, 2008 2:18:39 PM

^I have mine @ 3GHz and @ 1.225v right now and its stable with 12.5 hours of Prime95. I haven't tried to go higher yet but might to see what it can do. But I may wait till winter because its summer and I live in AZ, the desert.
a c 108 à CPUs
July 30, 2008 2:20:13 PM

I'm not sure about the L3/IMC latency.

IIRC Gary got 3GHZ at stock volts // 200-15-13-??(HT) - - the point being the L3/IMC clocked to 2600MHz without a voltage increase.

In the 'early' days the guys at XS were able to substantially reduce latency as L3/IMC speed increased. It wasn't 1:1 but it was pretty close ...
July 30, 2008 2:20:43 PM

^That's kinda what I'm thinking, but until we get some real analysis I will reserve judgment. The lack of details from AMD is worrisome though...
July 30, 2008 2:41:43 PM

jimmysmitty said:
^I have mine @ 3GHz and @ 1.225v right now and its stable with 12.5 hours of Prime95. I haven't tried to go higher yet but might to see what it can do. But I may wait till winter because its summer and I live in AZ, the desert.


I would like to test how much Intel Quads are dependant on the FSB by raising the NB to a nice 450 FSB - using something like a Q9950 E0, since it may be too much for a Q6600 - while decreasing the multi to still make it run at 2.8. I would like to see if the CPU would need a bump in the vcore - and, if yes, how that would compare to using the default multi - or just the NB would require higher voltages. Any ideas?
July 30, 2008 2:44:54 PM

jimmysmitty said:
^Theoretically it should be able to run at a lower voltage stock much like the Q6600 G0s can (like mine at stock was only on 1.118v :o  ).


Serious???



Right now, mine is at stock with a core voltage of between 1.296 and 1.312V (according to CPU-Z).

(It is G0 stepping)
July 30, 2008 2:58:50 PM

Mines... 1.232 volts, but OC'd with speed step / C1E enabled. :D 

At stock, sure.. I remember seeing my G0 at 1.144v (jumped around, but I did see it go as low as 1.118v):



Guess your not using speed step?
a c 126 à CPUs
July 30, 2008 3:00:43 PM

^Is that self set or auto? And what mobo? I have a Asus P5K-E that has nice quality caps and mosfets.

Plus remember each chip is different and some will run at lower than stock voltage and some will run higher. This helps determin how well it will OC.
July 30, 2008 3:02:52 PM

Wait.. you guys talking at load? :oops: 
!