AMD Shanghai/Deneb Review Thread

[donitsi]

Have you all seen this CF with PhII??

http://www.ocworkbench.com/2008/amd/...benchmarks.htm

They forgot to add the resolution they tested on?
(Crysis has 1600x1200 on the SS)

[Hornet331]

http://www.xtremesystems.org/forums/...9&postcount=42

riped out to keep this thread clean.

[iocedmyself]

quick question, anyone know how to get around the 1.45 voltage limit for cpu using Amd overdrive? know there is a way but too wiped out to dig around for it again, help would be much apprieciated

[charged3800z24]

quick question, anyone know how to get around the 1.45 voltage limit for cpu using Amd overdrive? know there is a way but too wiped out to dig around for it again, help would be much apprieciated

go into the AOD directory (ie: C:\Program Files\AMD\OverDrive)

open the preferences file (right click, open with, notepad) and change the max vcore below from "0" to "1". if it crashes this file is reset sometimes back to default. I just made a shortcut to the folder on my desktop for lazyness LOL..

- <Preference>
<AutoStartOnBoot>1</AutoStartOnBoot>
<NeedConfirmOnBoot>0</NeedConfirmOnBoot>
<AutoApplySettings>0</AutoApplySettings>
<FanBrokenDetectionEnabled>0</FanBrokenDetectionEnabled>
<LogFileEnabled>0</LogFileEnabled>
<LanguageType>0</LanguageType>
<BoardVendor />
<SkinScheme />
<MaximumVCoreEnabled>1</MaximumVCoreEnabled>
</Preference>

[imsochobo]

Use higher default voltage in bios also helps things out. running 1.4 V in bios (+150 MV) 1.392-1.408 V to be exact.

MY end voltage is with maximum Vcore in AOD is :
1.600 -1.616 V According to Cpu-z and DFI smart ITE monitor.

IF youre using Gigabyte MA790FX DQ6 you will notice the voltage drops below bios voltage setting, and on some motherboards it resets it completly and decreases it by 0.25% (unconfirmed 2.1.4 and up)
DFI Lanparty UT 790FX M2R dropped belov 1.25 V on phenom 9850 which is stock.

Gigabyte just drops a notch.

Hope this and the previous post helped you're question.

*Charged3800z24
Didnt know of it, thankya =)


**Smoothness talk***

Im running P4's A64(sempron) S754 A64 X2 AM2. Phenom 9850 BE. core 2 quad(little), core 2 duo(much exp) what envoriment, and what system, OS made me just wooow.
P.S| Cant understand it either, but its like that! here's my experience.

In order slowest to fastest smoothness/speed with OS | CPU | Mem Amount: P43GHZHT|Vista|2GB, P43GHZHT|XP|2GB, C2D3ghz|Vista|4GB, A64X23.4ghz|Vista|4GB, 9850BE/Q6600|Vista/win7|4gb, Sempron2800+|XP|1GB.
So there you have it. my list, i dont get it either, my feelings, experience, the A64 sempron 2800+ 1.6 ghz with 1gb of ram was the fastest in my experience, this is surfing the web, managing servers(vmware esx) configuring switches and it included multitasking.
The a64 X2 3.4 ghz didnt feel faster.

The P4 XP and Sempron 2800+ both utilized an "Maxtor 20GB Model :2B020H1 C2D utilized an 250 GB seagate. the rest of the system samsung 1TB/maxtor 200 gb tested both over time.

This is computers i actually work on, do productive work, and my experience can be described over many pages, but gives me no answer what so ever, i use them all several hours while the 9850BE is my primary computer and its running 3400 mhz.
The 2800+ is also the fastest system to boot, open cpu-z and firefox, and running the oldest install of windows which makes the experience more strange for me.

I'd totally see the K8 K10 architecture feeling smoother than Core 2, but compared to core 7 in 4 threads vs 4 threads, i'd say its just the same by my wild guess, but on this subject there cannot be any benchmarks, and when using an 2800+ i kind of getting the feeling that all the new computers ive bought lately is just a pure waste

[iocedmyself]

thank you fella's, had been looking in the config file and just couldn't remember what to do.

On asus 790GX setting vcore to 1.55 in bios gives actual vcore of 1.586, AOD maxes volts at 1.55. Still, means less time i have to spend resetting the bios now so much obliged, and lotsa benchs to post (again actually ) so look forward

[ajaidev]

New world record :-

Nº 1 : authienvu8@OCCLub_amtech.vn reached 6231.01 MHz with an AMD Phenom II (45 nm)

http://valid.canardpc.com/show_oc.php?id=474771

Nice

[Throwed]

New world record :-

Nº 1 : authienvu8@OCCLub_amtech.vn reached 6231.01 MHz with an AMD Phenom II (45 nm)

http://valid.canardpc.com/show_oc.php?id=474771

Nice

New World Record? Where have you been? 6.3 is the World Record. Remember Team Finland & AMD? Theres videos out with them hitting 6.3Ghz.

[Macadamia]

Notice the motherboard used again.

Might be our new champ.

[ajaidev]

New World Record? Where have you been? 6.3 is the World Record. Remember Team Finland & AMD? Theres videos out with them hitting 6.3Ghz.

Its a new world record allrite the mb used is a 790FX not a 790GX..

[knopflerbruce]

Its a new world record allrite the mb used is a 790FX not a 790GX..

Does it exist a WR for each chipset?

However, IMO it should exist a decent picture (or a reasonably steady video recording) of the cpuz tab, or a validation, if it's supposed to be called a WR. If that kind of stuff exist for the 6.3ghz-session, then allright.

[Blaber]

I think a new world record for PHII overclocking will be set CES 2009 when AMD will attempt to use Liquid Helium (-269°C) to cool down the Phenom II and clock beyond that 6.3Ghz mark (Source MADSHRIMPS Phenom scaling article).

I expect PHII to hit 7ghz at ces09 (Hopefully) .

[Raiderman]

I think a new world record for PHII overclocking will be set CES 2009 when AMD will attempt to use Liquid Helium (-269°C) to cool down the Phenom II and clock beyond that 6.3Ghz mark (Source MADSHRIMPS Phenom scaling article).

I expect PHII to hit 7ghz at ces09 (Hopefully) .

SaaaWeeet!

[Uncle Jimbo]

Which would at least be a lot better than what supporters of the "smoother" have now, which is no quantifiable evidence at all. At the moment, it's no better than the people trying to sell $1000 power cables for audio or those who still think vinyl records sound better.


And a lot of the P4 sales comes from AMD's inability to build and sell 200 million processors a year.

I have been involved in high end tube audio design for over 25 years. Many component effects were verified by blind testing when no one had any technical justification. Selection of components with specific characteristics can make an obvious difference, but the science to explain those differences is sometimes hard to find. Audio cabling is one area where some work has been done - copper stranded wire develops surface oxides which can have diode effects, which cause nonlinear transmission. Creating speaker cables from wire wrap (which are silver coated and gas tight) virtually eliminates those effects. You will see cables made with multiple gauge wire, transmission line designs, and other exotic stuff, each with advocates and some science. This is mostly art, but hand-selected parts put together with rules that are part witchcraft can still produce results that are clearly superior, even if the makers don't know exactly why.

In multisocket servers, AMD has shown superior performance in virtualization, many database applications and in some web serving, even though Intel wins on benchmarks. Look at the differences in Nehalem versus previous Intel hardware - Intel did not reduce the L2 cache by a factor of 10, or add on-die memory control, just for fun. They did it to be more competitive.

If a majority of large sample of users feel that some combination of AMD components produce a 'smoother' gaming experience, there are probably technical reasons for that. It is probably also dependent on the game engine and a ton of other parameters that are not well controlled, because users don't know what to control. A group of disciplined (and unbiased) amateurs could probably get to a better understanding of why that is true, but it would take a good sample of different systems, and some thought about what to test. This kind of forum could be a place to do that, if the testing group gets away from thinking they know the answer, and just follows the data.

[roofsniper]

would be nice to test it, we made another thread for it too. if we can find some unbiased testers and find a good testing method it would be nice but at this point we can't think of a good testing method and no matter what people come up with intel users will say that it is biased and won't work. so it would be nice to find some unbiased people willing to give it a test.

[Uncle Jimbo]

duplicated for some reason - original is above

[iocedmyself]

I recall seeing 1800HT and usually something like 800-1200 HT when going for 4ghz but just Figured i'd share my 4ghz cpuz anyway since i just got around to verifying

[Caveman787]

Theres been some with up to 2667 nb / hypertransport with 3.8 to 4.2 core clock.

I think just not everyone's willing to test the limits of the chip since they don't wanna risk any issues.

[iocedmyself]

well on NB i've hit 2742 i think, HT is just limited by NB speed, but if using cold on the chips or going for uber high clocks you have to set HT to 1ghz, hench the records being at 800-1200 HT speed.

[biohead]

I think a new world record for PHII overclocking will be set CES 2009 when AMD will attempt to use Liquid Helium (-269°C) to cool down the Phenom II and clock beyond that 6.3Ghz mark (Source MADSHRIMPS Phenom scaling article).

I expect PHII to hit 7ghz at ces09 (Hopefully) .

except for the fact that liquid helium sucks at removing heat and such a cooling system will be way too sophisticated.

but above all, beyond the -200 °C mark most electronics will fail to function the way they should. you need to stay above a certain temperature threshold.

[Uncle Jimbo]

except for the fact that liquid helium sucks at removing heat and such a cooling system will be way too sophisticated.

but above all, beyond the -200 °C mark most electronics will fail to function the way they should. you need to stay above a certain temperature threshold.

While liquid helium cooling is not for amateurs, it is possible to make large capacity coolers. The usual technique is to use a sealed system where the helium is always in a liquid state. LN2 is sometimes used in a dual dewar setup to reduce heat loss from the inner dewar. LN2 may also be used in the cooling stage.

This works like any refrigeration system - the closed loop system would need to provide adequate flow to cool 150W or so.

Both electron and hole mobility in silicon actually increase with decreasing temperature down to very low levels - both go up by a factor of 10 between 300K and 70K. This appears to be linear - there is no 'wall' at -200C, and it is just a question of where things get sufficiently far away from design parameters to quit working. I would assume that a part that works well at -190C would work at lower temperatures, and only experimentation would show where the degradation becomes significant. Silicon is not a superconductor at any temperature, with any doping levels used in semiconductor design.

And there is a fair amount of work on large scale liquid helium cooling at the University of Texas at Austin, so it seems possible that this could be a workable stunt.

references:
Centre national de la recherche scientifique
http://cat.inist.fr/?aModele=afficheN&cpsidt=1809679
Forced flow cooling at liquid helium temperature
Résumé / Abstract
Single phase helium at supercritical pressure is often used as refrigerant for super conducting components. At forced flow conditions it enables high heat transfer rates and it prevents the difficult conditions of a two phase flow. The advantageous flow conditions are paid with a considerably reduced capacity of the refrigerant since only the sensitive heat and not the evaporation heat is used.

Electrical Properties of Silicon - mobility versus temperature for different doping levels
http://www.ioffe.rssi.ru/SVA/NSM/Sem.../electric.html

CERN 1.4KG/S helium cryopump
http://www.barber-nichols.com/pdf/at..._pump_test.pdf

[iocedmyself]

http://valid.canardpc.com/records.php

Just have to smile at those 5.5ghz core I7 records there when the 6.3 phenom IIs are there in view.

[roofsniper]

http://valid.canardpc.com/records.php

Just have to smile at those 5.5ghz core I7 records there when the 6.3 phenom IIs are there in view.

yea but they aren't day to day clocks. it doesn't seem like i7 gets that great of a bonus from extreme cooling. once deneb comes out and we see what the average oc is that will decide it. not just the maximum overclock. only 4 more days.

[iocedmyself]

well of course they aren't day to day clocks which is kind of my point, the glass cieling for i7 is 800mhz below phenom II (thus far anyway) I7 has trouble hitting anything past 3.6 on air as has been shown by several reviews, at least one posted in this thread. While Deneb routinely seems to hit 3.9-4.1 on air, someone here had hit 5.6 or 5.7 with cold which isn't far off from record.

Though i think it has less to do with i7's not benifiting much from cold and more to do with they run so hot to begin with that there isn't much headroom

So keep meaning to repost that, memory running at 570 (1140) 5-5-5-5-18 ganged, also running a single thread at 3.8 scaled almost liniarly taking 39.922 compared to 4 threads at 10.25

[Hornet331]

well of course they aren't day to day clocks which is kind of my point, the glass cieling for i7 is 800mhz below phenom II (thus far anyway) I7 has trouble hitting anything past 3.6 on air as has been shown by several reviews, at least one posted in this thread. While Deneb routinely seems to hit 3.9-4.1 on air, someone here had hit 5.6 or 5.7 with cold which isn't far off from record.

4ghz is possible with air on Ci7 with HT off, just look in the Ci7 ocing thread. The only problem are temps with HT on and they depend strongly on voltges.
It make a huge difference if HT is enabled or disabled, disabling HT reduces temps for ~10-15°C

3.8ghz is no problem on air for most Ci7 with HT on and a good aircooler such as a TRUE or a Noctua U12P. Everything above is still possible but temps get in the critical range 90°C+ so WC is recommended or disable HT.