790GX Vs new 790FX "hawkfish" MB's

[acreda]

Hi Everyone, as a long suffering AMD fan i have been waiting for a decent AM2+ Motherboard with the SB750 and have been putting off my upgrade until they appear

I have now had my interest peaked by the new 790GX chipset. to me the are identical feature wise with the 790FX except from the Dual 8x pci-e slots (which is a mute point as Crossfire maybe coming but not here yet)

so considering that most MB with the 790GX set may sell cheaper than the new hawkfish board because of IGP that is still yet to see the light of day, and also that the only games I regularly play are UT2004, their mods and ut3 (fingers crossed) am i better just getting a 790GX board + 4850 card and saving some money with a cheaper board? will the restricted lanes ever hinder the performance if we are ever able to go crossfire?

[chithanh]

to me the are identical feature wise with the 790FX except from the Dual 8x pci-e slots (which is a mute point as Crossfire maybe coming but not here yet)I think it was already pointed out elsewhere: the 790GX supports PCIe 2.0, a PCIe 2.0 x8 slot has the same bandwidth as a PCIe 1.0 x16 which is already plenty.

[Redeeman]

you should think twice about SB750.. apparently the sata and usb performance is not very good..

[Melcar]

you should think twice about SB750.. apparently the sata and usb performance is not very good..

I thought it was "improved" over the SB600. SATA and USB performance isn't *that* bad with SB600; it could be better, but it's nothing that will make you cut yourself. If overall performance is better than with SB600 that would be a plus.

[Redeeman]

deanjo says write performance on sata is 45% that of competing chipsets....

as for usb, he also says its alot worse than other chipsets...

[Melcar]

I can't get the bonnie++ tets to run properly, so here is a run with iozone. SB600 with a single SATAII HDD:


http://www.phoronix.net/phoronix-test-suite/remote-graph/graph-remote.php?g=BAR_GRAPH&t=IOzone&s=Write%20Performance&n=3.291&u=MB/s&i=SATAII;&v=65.82;&p=HIB&x=1.0.5

http://global.phoronix-test-suite.com/index.php?k=profile&u=melcar-14281-27156-31933

Someone has a competing Intel/nvidia solution to compare to?

[deanjo]

http://www.phoronix.net/phoronix-test-suite/remote-graph/graph-remote.php?g=BAR_GRAPH&t=IOzone&s=Write%20Performance&n=3.291&u=MB/s&i=1;&v=106.41;&p=HIB&x=1.0.4

Nforce 780a on a old Maxtor 6B250S0 drive.

[deanjo]

When I had a couple of Maxtor 6B250S0's running dmraid in raid 0

http://www.phoronix.net/phoronix-test-suite/remote-graph/graph-remote.php?g=BAR_GRAPH&t=IOzone&s=Write%20Performance&n=3.291&u=MB/s&i=1st_Run;&v=184.82;&p=HIB&x=1.0.1

[Melcar]

Is there a way to test USB performance?

[acreda]

This may explain it, there is a set of benchmarks from the guys at Tech Report http://www.techreport.com/articles.x/15256 stating that the South bridge is runing in IDE and therefore the SB750 is not doing NCQ which would indicate the poor performance for writing speeds (burst speeds are fine) I understand that this ACHI problem existed in the SB600, why has it been carried over if its true?!?!?

[Melcar]

This may explain it, there is a set of benchmarks from the guys at Tech Report http://www.techreport.com/articles.x/15256 stating that the South bridge is runing in IDE and therefore the SB750 is not doing NCQ which would indicate the poor performance for writing speeds (burst speeds are fine) I understand that this ACHI problem existed in the SB600, why has it been carried over if its true?!?!?


ACHI mode does not really work with the SB600 (some BIOS on certain boards even remove the option). According to that review, the new SB still has issues with AHCI, so I think that's why they did not use it. AMD hyped this SB750 so much that I'm rather disappointed; the "extra overclocking" seems more like a gimmick as well.

[bridgman]

AMD hyped this SB750 so much that I'm rather disappointed;

Just curious, where are you finding all this hype ? I hadn't even seen the SB750 mentioned in the press other than as a companion to the 790GX.

[Melcar]

Just curious, where are you finding all this hype ? I hadn't even seen the SB750 mentioned in the press other than as a companion to the 790GX.

Maybe it was just the particular circles I frequent that were guilty of this over hyping.

[acreda]

Maybe it was just the particular circles I frequent that were guilty of this over hyping.

Well we will have to wait for the Linux benchmarks from Phoronix, Generally i find their results accurate, i hope it's going to be tested against the N 750 chipset and the old 708G with the old south bridge. do you think anyone will advise when article will be out, i assue there will be one? if it is reasonably close then i will want to stay with AMD for the Crossfire and ACC. after all if something is taking a while to copy over I'll just go a get myself a drink!

[deanjo]

Just curious, where are you finding all this hype ? I hadn't even seen the SB750 mentioned in the press other than as a companion to the 790GX.

How about these:

http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543_15434~127446,00.html

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3360

http://www.hardwarezone.com/articles/view.php?cid=6&id=2618

http://www.bit-tech.net/hardware/2008/08/06/amd-790gx-igp-and-sb750/1

http://news.softpedia.com/news/SB750-Allows-Phenoms-to-Be-Overclocked-90672.shtml

http://www.fudzilla.com/index.php?option=com_content&task=view&id=6956&Itemid=1

Plus hundreds of others, people expected alot from this chipset. Of course the hyped OC'ing capability of the chipset is pretty much a useless feature in linux as there is no overdrive utility for linux (such utilities are badly absent in linux).

[Melcar]

How about these:

http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543_15434~127446,00.html (http://www.amd.com/us-en/Corporate/VirtualPressRoom/0,,51_104_543_15434%7E127446,00.html)

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3360

http://www.hardwarezone.com/articles/view.php?cid=6&id=2618

http://www.bit-tech.net/hardware/2008/08/06/amd-790gx-igp-and-sb750/1

http://news.softpedia.com/news/SB750-Allows-Phenoms-to-Be-Overclocked-90672.shtml

http://www.fudzilla.com/index.php?option=com_content&task=view&id=6956&Itemid=1

Plus hundreds of others, people expected alot from this chipset. Of course the hyped OC'ing capability of the chipset is pretty much a useless feature in linux as there is no overdrive utility for linux (such utilities are badly absent in linux).

They're hardly useful, except for GPU based overclocking, so I for one don't miss them. Software based overclocking for the CPU/chipset always ends up with stability problems. GPU overclocking with Overdrive is not that great either (very small margins for frequency adjustment and the stress tool could be better), but for now it should suffice. Still, if you're going for the pure principle of the thing, Overdrive should be fully supported in Linux as well (even if it does not end up being too useful). As for the SB750 itself, given that disk and USB controller performance is the biggest reason to choose one SB over another, I say that it is a big letdown.

[deanjo]

They're hardly useful, except for GPU based overclocking, so I for one don't miss them. Software based overclocking for the CPU/chipset always ends up with stability problems. GPU overclocking with Overdrive is not that great either (very small margins for frequency adjustment and the stress tool could be better), but for now it should suffice. Still, if you're going for the pure principle of the thing, Overdrive should be fully supported in Linux as well (even if it does not end up being too useful). As for the SB750 itself, given that disk and USB controller performance is the biggest reason to choose one SB over another, I say that it is a big letdown.

Granted software OC'ing is the bastard child of overclocker enthusiasts, but is improving with each generation. Where it is nice is that gives usually a good starting base with it's results so that other methods can be applied such as the soon-to-be-dead bios methods. With systems gradually making the transition to software based alteratives (EFI for example) it is the future of OC'ing. It would be in the best interest of all parties to start refining the technology and spreading it across all platforms so when it does become the defacto standard we are not starting from scratch in linux and other OS's.

[Kacper86]

Of course the hyped OC'ing capability of the chipset is pretty much a useless feature in linux as there is no overdrive utility for linux (such utilities are badly absent in linux).
I totaly disagree. You do not have to use AMD overdrive utility to oc you cpu/memory/gpu! It's possible with bios tweaking - as always. 790GX is about ACC (advanced clock calibration).

[deanjo]

I totaly disagree. You do not have to use AMD overdrive utility to oc you cpu/memory/gpu! It's possible with bios tweaking - as always. 790GX is about ACC (advanced clock calibration).

I did not say you cannot OC with the BIOS. ACC is all about overclocking, that's it's sole purpose. It has also been proven time and time again that with ACC people are getting substantially higher OC's with it. What a utility allows via software is the ability to OC when needed and cut back when it's not without having to jump into the bios everytime to revert settings when OCing is or not needed. Nothing like roasting up a system websurfing and then having to reboot and the OC back up when you want to do something like video encoding, rendering or something else that can benifit from the extra boost in speed. One of the key features of linux is the ability to go eons without rebooting. A software app allows the dynamic clocking based on the current need without having to reboot.

[energyman]

well, there is one thing - all the tests are windows test. So they aren't very useful. I am at the moment looking for a new board. And I found that Vista SP1 greatly reduces the usb performance of amd southbridges. So.. if the tests were done with Vista SP1 you can ignore them.
http://www.pctreiber.net/thread.php?threadid=9448

[MetalheadGautham]

well, there is one thing - all the tests are windows test. So they aren't very useful. I am at the moment looking for a new board. And I found that Vista SP1 greatly reduces the usb performance of amd southbridges. So.. if the tests were done with Vista SP1 you can ignore them.
http://www.pctreiber.net/thread.php?threadid=9448

Really ? Then I guess we need some "standardised" tests...

How about Linux 2.6.26 i686 Kernel and EXT3 file system ?

PS: I suppose the speed would be higher with XFS ?

[ajw1980]

I did not say you cannot OC with the BIOS. ACC is all about overclocking, that's it's sole purpose. It has also been proven time and time again that with ACC people are getting substantially higher OC's with it. What a utility allows via software is the ability to OC when needed and cut back when it's not without having to jump into the bios everytime to revert settings when OCing is or not needed. Nothing like roasting up a system websurfing and then having to reboot and the OC back up when you want to do something like video encoding, rendering or something else that can benifit from the extra boost in speed. One of the key features of linux is the ability to go eons without rebooting. A software app allows the dynamic clocking based on the current need without having to reboot.

I agree that a software tool for overclocking would be nice. Apparently phenom use msr's to set a lot of these parameters, which I would guess AOD might use, too. I played around with setting the msr's, but didn't see any noticeable effects for over/underclocking. I was able to disable the TLB fix on a phenom 9600 by modifying the proper msr (my bios doesn't have an option for it). I posted that in the forums here.

There are details on setting speeds/voltages/etc here:http://www.xtremesystems.org/forums/showthread.php?p=2806501

Maybe someone with some more time can figure out if it is possible to change this on the fly in linux.

[energyman]

the speed would be highest with reiser4 ;)
and don't forget the barrier option - ext3 hates your data so it is turned of, while reiserfs and xfs love your data and have it turned on.