AMD phenom II X6 hexacore vs Intel i7 quad HT

simple question, which chip is better in terms of PPD and power consumption? the price of the new 6core AMD chips looks really competitive and i'm considering buying one. i'm not looking for gaming, subzero, or rendering performance, it's solely to crunch with.

it's my understanding that an intel's HT makes each core ~50% more efficient, so a quad with HT is about equal to a a true 6 core chip with no HT. i'm assuming that the i7 will generally OC better and consumes lower power at the higher clocks, but i could be wrong. which chip has the upper hand in computational power and power consumption between 3.6ghz - 4ghz? and please, i do not want to turn this into a battle between fanboys that swear by either chip.

that's a tough question. intel is very good at crunching and both amd and intel chips currently OC well. overclocked i7s consume a lot of power though. using the wcg ppd calc amd's new 6 core phenom at 4GHz should get 26,400ppd. the i7 @4GHz is 30,360. you can probably get power consumption numbers off of some review site when NDA is over.

NDA? some people ordered these retail hexacore like a week ago and already received them.

i'll do some digging around, but if anyone has first hand experience on hexacore crunching, please post :)

lol no more debate on this. the hexacore wins, it's only $125 after rebate and cashback.

$200 shipped from tigerdirect (link)
$50 mail in rebate (link)
$25 bing cash back (link)

For crunching, the hexacore wins easily in price and power consumption ratio. It will basically have the same or slightly better performance. HT = app. 30% extra performance. 4 * 1.3 = 5.2 "real" cores where Thuban has 6.

Quote:

---

Originally Posted by v0dka

For crunching, the hexacore wins easily in price and power consumption ratio. It will basically have the same or slightly better performance. HT = app. 30% extra performance. 4 * 1.3 = 5.2 "real" cores where Thuban has 6.

---

i just always heard it was 50% more performance for HT. how did you figure out the 30% figure?

How much performance increase you get is highly dependent on the load you apply. The worse a given thread is in terms of efficient cache usage and IO waiting time the better the improvement you'll get by having another whole thread "ready to roll". On the other hand, a highly efficient thread will lose very little time to being locked out waiting for cache refreshes, IO etc. In that case you'll gain far less from hyperthreading.
Some folks will remember a while back one of the projects, might have been the original HCC I'm not sure, was an absolute mongrel for cache misses (page faults). Having another thread to switch to, already in cache, was a great boost to those units. My Athlon XP2400+, on the other hand, suffered greatly due to lack of cache (256k compared to the P4 2.4 beside it which had 512k) which made the situation far worse.





Yes, I normally run older gear coz that's what I've got.

Quote:

---

Originally Posted by D_A

How much performance increase you get is highly dependent on the load you apply. The worse a given thread is in terms of efficient cache usage and IO waiting time the better the improvement you'll get by having another whole thread "ready to roll". On the other hand, a highly efficient thread will lose very little time to being locked out waiting for cache refreshes, IO etc. In that case you'll gain far less from hyperthreading.
Some folks will remember a while back one of the projects, might have been the original HCC I'm not sure, was an absolute mongrel for cache misses (page faults). Having another thread to switch to, already in cache, was a great boost to those units. My Athlon XP2400+, on the other hand, suffered greatly due to lack of cache (256k compared to the P4 2.4 beside it which had 512k) which made the situation far worse.

---

lol excuse my ignorance, but is there a dumbed down translation of this? i have no idea how efficient HT is for WCG, and i see a lot of mixed performance on HT. so in summary, would a hexacore beat a quad with HT at the same clocks in terms of PPD?

You may not see much in the way of power savings with the Phenom. Anandtech just posted its article. These are the power numbers they got. This is total system power, not just CPU power consumption.

thanks for the chart sierra. although, that doesn't agree with the article that guru3d just published a few hours ago. they commend the phenom II hexacores for being power efficient (at stock clocks) and run very cool.

The better chip in terms of power savings is probably the 1035T which no one seems to have in stock yet. It's rated at 95W versus 125W for the 1055T and 1090T.

Yes, those results are much different than Anandtech's. I just happened to notice Anand Shimpi posted his article, so I took a quick look. He generally has a favorable impression of the X6.

Quote:

---

Originally Posted by WhiteFireDragon

lol excuse my ignorance, but is there a dumbed down translation of this?

---

lol sorry. Yes there is, but it will take an hour to set up and explain the analogies to make them cover all the bases. If you want to know more do some searching on "memory hierarchy", "page faults", "turnaside lookup buffer" and possibly "memory paging".

Short short short version:
Good programming + big cache + big RAM + HTT = :up::up::up:
Good programming + big cache + big RAM + no HTT + moar coars= :up::up::up::up:
Bad programming + big cache + big RAM + HTT = :up: :up:
Bad programming + big cache + big RAM + no HTT = :up:
Bad programming + small cache + big RAM + HTT = :up:
Bad programming + small cache + small RAM + HTT = :down:
Bad programming + small cache + small RAM + no HTT = :shakes:

My bet is that performance-wise, they'll basically be even at identical clock rates.
HT gives about a 40% boost on WCG, along with the i7's slightly higher IPC, it should level out.
It remains to be seen how far the X6 can be OCed, and how power-efficient it is - price/performance is definitely in favor of the X6, considering the motherboards are about half the price of an x58.

However, as soon as affordable Intel Hexas are available, the game changes again. They do 50% more than a Bloomfield, clock very high, and draw even less power.
Right now, 32nm Westmere is unbeatable when it comes to performance per Watt (Dual X5650: 72000 WCG PPD @ 300W from the wall)

Quote:

---

Originally Posted by jcool

However, as soon as affordable Intel Hexas are available, the game changes again. They do 50% more than a Bloomfield, clock very high, and draw even less power.
Right now, 32nm Westmere is unbeatable when it comes to performance per Watt (Dual X5650: 72000 WCG PPD @ 300W from the wall)

---

The question is when will that happen. AFAIK Intel hasnt even anounced non EE hexacores yet.

http://www.tigerdirect.com/applicati...mgqKNUi_y.GNkQ

AMD 1055T for 125 no tax and shipping in CA after 50 dollars rebate + 25 dollars bing

What a steal!!!

Quote:

---

Originally Posted by v0dka

The question is when will that happen. AFAIK Intel hasnt even anounced non EE hexacores yet.

---

Well there's DP hexacore xeons but well they ain't any cheaper than 980X

Quote:

---

Originally Posted by jcool

However, as soon as affordable Intel Hexas are available, the game changes again. They do 50% more than a Bloomfield, clock very high, and draw even less power.
Right now, 32nm Westmere is unbeatable when it comes to performance per Watt (Dual X5650: 72000 WCG PPD @ 300W from the wall)

---

I have the feeling that this article might be true. Can't say how I know, and have no clue how affordable it'll be, but it'll be less then $1k;)
http://www.fudzilla.com/content/view/17499/1/

as for the main question. Looks like for perf/watt/$ its a pretty close race between the Lynnfield i7's and the new phenom x6's. I look forward to seeing some results here. I am admittedly biased towards the i7's, but I think the performance of each will be about the same, but if I have to guess I would think that the platform power consumption of the P55/i7 8xx combo will be a bit better at comparable overclocks.

http://www.tigerdirect.com/applicati...916&CatId=5291

That barebones isn't a bad deal honestly, but I'm going to see if I can't come up with something more reasonably priced on Newegg (I'm partial towards Newegg, don't much care for TigerDirect). At the price of the Thuban chips, you can hardly resist. I hope they at least hold up to the i7s, would be interesting to see what they'll do.

the big question to me is which has the overall cheaper platform?

what's the lowest you can buy a hexacore setup for, assuming you buy the chip for $125?
-what's the power draw

what's the lowest you can buy a i7 platform for, assuming you can buy a 920 for $199?
-what's the power draw

basically, i7 mobos are $$$$$$, is it the same case for the AMD?

Quote:

---

Originally Posted by Tristor

http://www.tigerdirect.com/applicati...916&CatId=5291

That barebones isn't a bad deal honestly, but I'm going to see if I can't come up with something more reasonably priced on Newegg (I'm partial towards Newegg, don't much care for TigerDirect). At the price of the Thuban chips, you can hardly resist. I hope they at least hold up to the i7s, would be interesting to see what they'll do.

---

i just bought this case: http://www.newegg.com/Product/Produc...-11154098-LC1C

and applied code: EMCYRNV34 for $4 off. comes out to $35.99 free shipping for what appears to be a pretty darn good case. i particularly like the cut out behind the cpu on the mobo tray. i think i'll be using that feature often with as much as i'm switching cpus around trying to find the best PPD :)

i'll chime in when i get it.

Quote:

---

Originally Posted by Rise

the big question to me is which has the overall cheaper platform?

what's the lowest you can buy a hexacore setup for, assuming you buy the chip for $125?
-what's the power draw

what's the lowest you can buy a i7 platform for, assuming you can buy a 920 for $199?
-what's the power draw

basically, i7 mobos are $$$$$$, is it the same case for the AMD?

---

Very nice mobo 890GX/AM3 for $140 with free shipping on Newegg:
http://www.newegg.com/Product/Produc...82E16813128435

I just finished configuring what my ideal would be for a Thuban WCG node. Compared to an i7-860 it's about $100 more, but that's partly because I got the 1090T vs the 1055T, which was exactly $100 difference. I figured with the unlocked multi I might be able to push the clocks further (are there people doing 4.0 with these without unlocked multi? I can hit 4.0 HT on with the 860).

Quote:

---

Originally Posted by Rise

basically, i7 mobos are $$$$$$, is it the same case for the AMD?

---

You should be able to find ~$60 boards that support Thuban (check for updated BIOS), running DDR2 if you still have any around. Overclocking range is the limiter on the low end boards.

Quote:

---

Originally Posted by Rise

the big question to me is which has the overall cheaper platform?

what's the lowest you can buy a hexacore setup for, assuming you buy the chip for $125?
-what's the power draw

what's the lowest you can buy a i7 platform for, assuming you can buy a 920 for $199?
-what's the power draw

basically, i7 mobos are $$$$$$, is it the same case for the AMD?

---

Best buy is Asrock 890gx extreme3 (129$) and 1055T :up:

http://www.newegg.com/Product/Produc...asrock%20890gx

And the 890GX can unlock quads to X6 too.

Quote:

---

Originally Posted by v0dka

And the 890GX can unlock quads to X6 too.

---

If the quad Thubans were forsale. :D

well i decided to jump on it and bought one of these. great, now i have to learn how to OC these things. i don't know squat about AMD, haven't used one ever since conroe came out. will be crunching with one of these soon.

do these AMD boards support SLI, or at least can i use two nvidia cards to crunch for GPUgrid as well? if so, can anyone recommend the lowest costing board that is mATX, support multi GPU's, and able to mildly OC these new hexacores?

on the intel side, the Gigabyte uATX UD2 is a great clocker for $100
http://www.newegg.com/Product/Produc...82E16813128405

Quote:

---

Originally Posted by WhiteFireDragon

well i decided to jump on it and bought one of these. great, now i have to learn how to OC these things. i don't know squat about AMD, haven't used one ever since conroe came out. will be crunching with one of these soon.

do these AMD boards support SLI, or at least can i use two nvidia cards to crunch for GPUgrid as well? if so, can anyone recommend the lowest costing board that is mATX, support multi GPU's, and able to mildly OC these new hexacores?

---

No, I don't think AMD board can be able to support SLI without hacking it. But AFAIK GPUGrid will be able to recognize. If you really want to enable SLI on those things, you can check this out; I don't know if this works or not, though:shrug:

By the way, tell us about the 1055T when you got it on the grid. It would be a tie breaker for me if it is able to clock right (4-4.2GHz).:)

well perhaps i don't need to connect in SLI by using the bridge, but i'd like to use at least 2 video cards on the board to crunch two WU for GPUgrid. is multi nvidia GPU possible without being in SLI configuration?

i'll surely post PPD results no this hexacore when i get it. my only worry is the power consumption when OC'ed.

Possibly. You can get a system with a single vid card but six Telsa cards in it if you don't mind taking on the national debt. No SLi but lots of GPUs.

Quote:

---

Originally Posted by WhiteFireDragon

well perhaps i don't need to connect in SLI by using the bridge, but i'd like to use at least 2 video cards on the board to crunch two WU for GPUgrid. is multi nvidia GPU possible without being in SLI configuration?

i'll surely post PPD results no this hexacore when i get it. my only worry is the power consumption when OC'ed.

---

No, you will not need a SLI bridge for GPUGrid to recognize the video cards. I've seen in the other reviews that the Thuban does not drain as much power as i7s. I'm not really sure, but you can always check reviews from the other sites.:)

MSI NF750 G65 has SLI (2x8).
Around 100$,single chip design,8200 IGP.
MS-7578 model number.

Any verdicts on the X6?

I would like to see how these do on wcg myself. too bad we missed out on the killer td deal

I have my 1055T coming in tomorrow to go into my WHS box which will start crunching for WCG. I figure I won't be streaming often, so I might as well put the cycles to good use. I doubt I will overclock it, but who knows. For $125 I couldn't pass on the upgrade from my Triple core.

The problem with AMD and SLI is simple.. all SLI capable boards for AM2/AM3 use NV chipsets, and they basically all suck. They run hot, don't overclock as well as AMD or Intel ones and are a bug-ridden PITA to deal with. I once had a 780a (M3N78-HT) and let me tell you this: Never again :p:

Only decent way to go SLI these days is with an x58.

I am curious how well these X6's can do in performance/watt. AMD is always far too conservative with their voltages, and these new chips seem to be even more power efficient than the Deneb cores which were already capable of getting huge gains (or loss?) in power savings by undervolting them. My 955BE (C2 stepping) has a stock vCore of 1.35v and I can run it at 1.1875v at stock speeds which is a pretty huge drop in power consumption under full load. If these 6-cores could run into the 3Ghz territory with similar undervolting they could generate a pretty huge amount of points for little wattage.

I am in for now. I am gonna oc it later. I am new to AMD overclocking, so we will see...

http://www.abload.de/img/x67d1v.png

ah first person to crunch with these hexacores. that's great. soya, post EVERYTHING you can about it crunching, mainly power consumptions, overclocks, points, etc.. i'll be getting mine this monday and i'll also join the results

Yeah, from my oberservations right now. I think you can compare it pretty much with my old Q6600 Clock-to-Clock wise. It just got 2 more cores.

WFD where do you stand right now with your X6?

lol nowhere, because the package hasn't came yet. i'll get it this monday and plug it in as soon as i can. i'll let it start crunching first while i learn how to OC amd chips

instead of learning to OC it when i get it, it'll probably be wise to learn the basic of it first and when i get this chip on monday, i can apply it to OCing. i skimmed some of the n00b guides, and a few questions came to mind. running the HT frequency and NB frequency in sync is beneficial, but how important is it for real world applications? will it actually increase the computational power? or will it be almost useless sort of like tweaking the memory and triple channel on the i7? and i'm assuming AMD platform is not limited to 1.65v memory only right? i can use older ddr3 2.0v memory?

Quote:

---

Originally Posted by jcool

The problem with AMD and SLI is simple.. all SLI capable boards for AM2/AM3 use NV chipsets, and they basically all suck. They run hot, don't overclock as well as AMD or Intel ones and are a bug-ridden PITA to deal with. I once had a 780a (M3N78-HT) and let me tell you this: Never again :p:

Only decent way to go SLI these days is with an x58.

---

I thought you can mod/hack it into using SLI:yepp: I remember giving WFD a link earlier to a guide for hacking any chipset into recognizing SLI. I've never tried it, though.

Quote:

---

Originally Posted by WhiteFireDragon

lol nowhere, because the package hasn't came yet. i'll get it this monday and plug it in as soon as i can. i'll let it start crunching first while i learn how to OC amd chips

---

Is it the 1055T?

Quote:

---

Originally Posted by WhiteFireDragon

instead of learning to OC it when i get it, it'll probably be wise to learn the basic of it first and when i get this chip on monday, i can apply it to OCing. i skimmed some of the n00b guides, and a few questions came to mind. running the HT frequency and NB frequency in sync is beneficial, but how important is it for real world applications? will it actually increase the computational power? or will it be almost useless sort of like tweaking the memory and triple channel on the i7? and i'm assuming AMD platform is not limited to 1.65v memory only right? i can use older ddr3 2.0v memory?

---

HT & NB in sync? No, not that important. HT at 2-2.2GHz is more than enough, as higher HT has almost no gain.
NB actually affects performance, so as high as possible normally 2.5-2.9GHz.

Cheers

This 1090T ran 100% on all cores overnight (still running) for the XtremeSystems team and is only at 48C. Stock speeds, volts and cooler. 3210@1.236v.

It would be interesting to see the performance on windows and linux. My bet is that AMD will outperform the 920 slightly in linux while the opposite will be the case in windows.

Quote:

---

Originally Posted by PaganII

This 1090T ran 100% on all cores overnight (still running) for the XtremeSystems team and is only at 48C. Stock speeds, volts and cooler. 3210@1.236v.

---

How many points have you knocked out already?

Waiting validation

Do you recall the numbers you got from the BOINC benchmark? The benchmark runs automatically when you first install BOINC.

From my past experiences with WCG, Intel CPU has always had 5 ~ 10% performance advantage over AMD cpu. In some cases, AMD CPU suffered more performance loss.

If you are after pure points generation, it might be better to stick with Intel.

I've had my 1090T running since yesterday too, most results still pending validation though. Heres the CPU benchmark results. When I get a better idea of how its going to produce I'll let you all know

OS: Ubuntu 9.10 64-bit

http://img51.imageshack.us/img51/459...inccpubenc.jpg

What clock speeds is that at?

If you run HFCC or FAAH exclusively for a few days you'll get your exact output as they have quorum 1. Aka, no PV.

I have made no changes to BIOS so it should be running stock, but they are all running at 3.6. I suspect there is something wacky going on with ubuntu and the turbo mode on these chips. CPU-Z does not work on Ubuntu Linux so I ran CPUINFO and took screenies. Hope thats good enough for you all. I took screenies of all 6 cores so I could make sure it wasn't throttling 2 of them(like its supposed to...)

http://img338.imageshack.us/img338/1...tcpu0speed.jpg
http://img94.imageshack.us/img94/720...tcpu1speed.jpg
http://img685.imageshack.us/img685/1...tcpu2speed.jpg
http://img685.imageshack.us/img685/9...tcpu3speed.jpg
http://img704.imageshack.us/img704/9...tcpu4speed.jpg
http://img222.imageshack.us/img222/4...tcpu5speed.jpg

I have done no over clocking with it yet, not intentionally anyway. I wanted to get it on for the supercomputer challenge and I will play with it after.

I'm going to take a wild guess and say 25,000 ppd. If it ends up being more than that, my interest in these new Phenoms increases a lot.:D

Quote:

---

Originally Posted by Someguy1982

I have made no changes to BIOS so it should be running stock, but they are all running at 3.6. I suspect there is something wacky going on with ubuntu and the turbo mode on these chips. CPU-Z does not work on Ubuntu Linux so I ran CPUINFO and took screenies. Hope thats good enough for you all. I took screenies of all 6 cores so I could make sure it wasn't throttling 2 of them(like its supposed to...)

<snip>

I have done no over clocking with it yet, not intentionally anyway. I wanted to get it on for the supercomputer challenge and I will play with it after.

---

I find disabling "Cool 'n Quiet" in the BIOS is worth the minute or so to do. :up:

Quote:

---

Originally Posted by sierra_bound

I'm going to take a wild guess and say 25,000 ppd. If it ends up being more than that, my interest in these new Phenoms increases a lot.:D

---

I would be happy to see this, I was hoping for 23k or so. I would definatly take 25k tho :yepp:

Quote:

---

Originally Posted by D_A

I find disabling "Cool 'n Quiet" in the BIOS is worth the minute or so to do. :up:

---

Its already been disabled, I'm wondering if that is what is keeping it from underclocking a few cores. It's supposed to be running at 3.2... The fact that is running at the "turbo mode" speed without underclocking a few cores had me a little worried at first, but it seems like all cores are running at 3.6

Quote:

---

Originally Posted by Someguy1982

I would be happy to see this, I was hoping for 23k or so. I would definatly take 25k tho :yepp:

---

Hard to say with a totally new CPU which is why I said I was taking a wild guess. Yes, 25K would be very good. Anything on top of that would be gravy. If it's above 27K, then I'm driving to the nearest Fry's right away.:yepp:

I like your estimate better then mine, and I hope you are closer to the truth then I! :)

At this point I'm guessing that "turbo mode" is just a more aggressive version of Cool 'n Quiet, or at least uses some of it's functionality. Regardless, that's a 400MHz overclock for free.

Yep, I wont complain about that. I think it's supposed to boost some cores and slow the other cores. It just isn't working right.

AMD's turbo mode is similar to Intel's version. This is what Anandtech said about it:

Quote:

---

Turbo Core kicks in when 3 or more cores (on a 6-core part) are idle. When this happens, the frequency of those three cores is reduced to 800MHz, the voltage to the entire chip is increased, and the remaining three cores are turboed up by as much as 500MHz. It doesn’t get any more granular than this. If you have 3 or more cores idle, then the remaining turbo up. In any other situation the CPU runs at its normal clocks.

---

http://www.anandtech.com/show/3641/a...o-core-enabled

Looks like that chip is running at full "turbo" with all six cores then.

Just a thought ... CPUinfo might be reporting the LISTED speed of the cores rather than the actual running speed.

Yeah that's what it looks like. Only change I made in BIOS(besides setting the time) is turning CnQ off.

I would think the listed speed should be 3.2 though. You're a more experianced linux user then I, D A, you have any suggestions on better programs to use to monitor core speed? Its also reporting 3.6 at the system monitor.

Quote:

---

Originally Posted by D_A

Just a thought ... CPUinfo might be reporting the LISTED speed of the cores rather than the actual running speed.

---

Could be, in Windows you can get the CPU speed from the registry but its the max speed, not real time.

fwiw, heres an X6 @ 4138 scoring 3414/10815 in bionc on win7-64.
SS is in second post... http://www.xtremesystems.org/forums/...d.php?t=250930

Quote:

---

Originally Posted by Someguy1982

I would think the listed speed should be 3.2 though. You're a more experianced linux user then I, D A, you have any suggestions on better programs to use to monitor core speed? Its also reporting 3.6 at the system monitor.

---

There's a CPU frequency scaling applet that you can run in your task bar. I've used it before to let me monitor the CPU frequency but with the CnQ type stuff enabled it also lets you UNDERclock your CPU from the desktop. I disabled that part. ;)

Attachment 103741

Attachment 103742

There's also an app called "Perlmon" which just gathers all that kind of info from the various system files.

I run that one too, it says 3.6. Is that real-time though? In BIOS it says 3.2 Ghz but in Ubuntu everything says 3.6. I get a warning message when Ubuntu starts saying "cannot use CPU scaling" and I assume It's from disabling CnQ. I suppose I could Reboot and see if it changes. Do I REALLY want to lose that 400mhz though...

Well ... it's only supposed to be for when you're running a heavy single threaded load and it down clocks the other cores to help keep the heat down. Now I'm not gonna take any responsibility if you cook your shiny new hexa-core chip, but I'd GUESS that if you can keep the heat under control it SHOULD be ok.

Enabling Cool 'n Quiet in BIOS brought the speed down to 3.2. Now to restart and set it to run at 3.6, since I know that works.

Here is the BOINC benchmark at 3.2ghz
http://img265.imageshack.us/img265/2...chmark3200.jpg

What are you using to monitor your temperatures and what cooler are you using?

I've got a TRUE on it, and LM-sensors but I haven't gotten it working yet.

have you run sensors-detect? (sudo sensors-detect)

yep, this is what i get...

#----cut here----
# Chip drivers
# no driver for AMD K10 thermal sensors yet
#----cut here----

Says it all, really. Gotta wait for the update OR go get the latest source and roll your own ;)

Wish I knew how...

It will defineatly be reporting frequency wrong if the reading is under load. The chip would never allow 3.6Ghz on all cores loaded at stock.

What does do though is essentially sit in Turbo mode whenever it's completely idle or less than 50% CPU usage(load distributed across any cores). when load goes above that it'll be at 3.2.
I
-edit , my mistake, it does only boost active cores!

It was definitely under load when I checked. When I enabled CnQ everything went to 3.2 though.

just edited my post, it does only boost the 3 cores.

CnQ is a bit screwy at the moment, I haven't seen anyone be able to disable it fully, I think the HW turbo mode overrides

It might use CnQ to regulate itself to 3.2, running up to 3.6 with the controls disabled. I really don't know, but it wouldn't surprise me.

In the BIOS screen when I was Re-enabling CnQ it mentioned that in processors which support turbo mode CnQ should be set to disabled. I had it disabled though and got an error message on boot-up in Ubuntu saying it would be unable to scale the frequency properly. I actually set CnQ to "auto" instead of enabled or disabled. No freq. scaling error message and it runs at 3.2 under load. When I go to overclock it all this will be moot anyway, but until LM-sensors works I'm going to refrain. It ran 24 hours at 3.6 to the best of my knowledge and it didn't seem to be running very hot, but without the sensor working properly there is no real way to tell yet. Not one available to me, anyway.

mm I need to get AMD Overdrive to work with my mobo so I can check it out. No other freq reporting tool reports what's going on properly.

But I think that all totally idle cores will always drop into a lower P state / frequency even with CnQ off. Perhaps Turning it OFF just prevents the legacy operation of CNQ from occuring..

Try this:

enter "sudo modprobe k10temp" (then your password of course)

then "sensors" and see what it says.

Gave it a shot D_A this is what returned

Code:

---

$ sudo modprobe k10temp
[sudo] password for XXXXXX:
FATAL: Module k10temp not found.

---

What kernel version are you running?

(type "uname -r" in the terminal)

2.6.31-21-generic

Code:

---

$ uname -r
2.6.31-21-generic

---

The K10temp module is listed as being included in the 2.6.33 kernel series .... how much do you want it?

A bit, but as I don't plan on bringing this computer down until supercomputer week is over I can wait a week. Ive found a few websites with instructions on adding k10temp manually, but I'm afraid I have run out of time to do it right now. Later tonight I will be back to mess it up :p:

Honestly I think you're better off just compiling the kernel from source. I can help you with that, it's actually not that hard (when you have a script to work from ;) )
Wait till after the challenge, though, so you have time to tinker with the MANY new settings you'll discover.

Sounds good, I look forward to it.

Just a reminder, when compiling the kernel, don't forget to build in support for your file system, network, and SATA. I've done it before, and it is not fun to debug when you're totally new to it:D
Just take your time in menuconfig. I love compiling my own kernel now. Compiling gnome or KDE on the other hand...

So I am up and running on my WHS server with a 1055T, however I have one quick question, when I go and stream from the server what happens? Does WCG suspend, share or other?
Thanks.

usually it suspends until your cpu is freed back up. i can leave boinc crunching, start a game and never miss a frame.. i hope that's what you're asking

Yep. Just want to make sure it doesn't mess with my streaming. Thanks.

Quote:

---

Originally Posted by sierra_bound

Do you recall the numbers you got from the BOINC benchmark? The benchmark runs automatically when you first install BOINC.

---

Hello, Just got home, checked stats after a little over 26 hrs running, results upload 1 hour ago and shows average points per calander day as 1,266.83 but probably because I joined 3 days before actually running. Points generated shows 7,601 but did not start running till 27hrs ago. Team History and stats updated 5 hour ago so does that make it 22 hours? 345.5 points per hour?

Edit: My Up Time now shows 1:04.43.33

Don't remember benchmark, sorry. Wanted to get some idea of performance before activating other machines.
Found it, Thanks soya! 3275Whetstone/8011 Dhrystone for stock speed, volts, cooler +2600NB

Quote:

---

Originally Posted by jspace

Just a reminder, when compiling the kernel, don't forget to build in support for your file system, network, and SATA. I've done it before, and it is not fun to debug when you're totally new to it:D
Just take your time in menuconfig. I love compiling my own kernel now. Compiling gnome or KDE on the other hand...

---

My intention was/is to nurse folk through the process the first time and just build a new kernel with everything the same as the original generic kernel. That way everything should work straight away, but they get the additional modules that come with the new kernel version.
For my own kernels, I strip them worse than an open cut mine. I even cut multicore support down to only and exactly the number of cores in the precise rig the kernel is for, just to save a little bit more memory.

Well, thats the Bench at 4GHz. I do know that it's quite much voltage, but it isn't stable at all beyond 3,6GHz.

W7 64 Bit
http://www.abload.de/img/x694j5.png

For some reason I'm getting better results with less Mhz

1055T @ 3.92Ghz NB @ 2800 RAM 1120 CAS 7

3246 FP (whetstone)
10344 INT (Dhrystone)

so at 4 I'd expect a couple % higher even.. Sensitive to NB/L3 cache ??