ATI RS780 benchmarks

[Npl]

Guys the sideport memory bank is not for performance. It's more of a power saving measure so that you don't need to keep the CPU active when you want to write to VRAM. The sideport buses are usually 16-32bit, just enough to provide the most basic functionality.

Its not like the IGP will compete with 256-bit discrete cards, but if you look at the graphs, the IGP is compared with a 8400GS, which only has a 64bit Memory path itself. Suddently its not looking so insignificant.

Also, consider that the CPU wont be happy if the GPU steals its bandwidth and adds potential latency, so by adding a LFB you both increase total Bandwidth for the IGP and ease up system bandwidth contention.

 

[Dave Baumann]

Couple of reviews up:
http://www.tomshardware.com/2008/03/04/amd_780g_chipset/index.html
http://www.techreport.com/articles.x/14261
http://www.hothardware.com/Articles/AMD_780G_Chipset_and_Athlon_X2_4850e_Preview_/

 

[Bouncing Zabaglione Bros.]

ATI still hasn't fixed the USB/SATA in the new SB700? Another missed opportunity:

The 780G isn't perfect, though, and our enthusiasm for the chipset is tempered by an SB700 south bridge that feels more like a die-shrunk SB600 than a brand new chip. The fact that AMD has allowed AHCI problems that afflicted the SB600 to persist in the SB700 comes across as more than just a little sloppy, particularly given the hoops one has to jump through just to install Vista in AHCI mode. Mainstream desktops and home theater PCs may see little benefit from the Native Command Queuing support that AHCI mode provides, but it's a feature that should work properly. And it just doesn't.

 

[DSC]

Overhyped but underdelivering on the southbridge again as usual.

 

[Dave Baumann]

ATI still hasn't fixed the USB/SATA in the new SB700?

Not sure if you looked at the benchmarks, but the USB is little slower than Intel reading and a reasonable amount faster writing, with lower CPU utilization. ACPI Vista install issues are not a hardware fix - AFAIK this gets supported in Vista SP1 installs (though I'm not sure).

 

[DSC]

http://www.siliconmadness.com/2008/03/amd-780g-chipset-doesnt-support-ht-30.html

What AMD isn't telling it's potential customers is that there is a major issue with the current A12 silicon: it doesn't support HT 3.0. HT3.0 is marketed as a big feature of Phenom CPUs and the AM2+ socket because it provides support for a maximum speed of 2.6GHz, providing more bandwidth than the HT2.0 link which is limited to 1.0GHz, or 2000MT/s.

The reason why most testers aren't noticing this workaround is because AGESA, a part of the BIOS, is automatically reducing the link to 1GHz HT2.0 when it detects an A12 revision of the chipset. Nvidia's upcoming, competing, solutions will offer full HT3.0 compliance, with production starting this month.

 

[Zaphod]

ATI still hasn't fixed the USB/SATA in the new SB700? Another missed opportunity:

Once again Tech Report is showing much faster results (97 MB/s) for Intel chipsets in the write test than the disk they're testing with is specced for (84 MB/s buffer to disk) and also significantly faster than the measured read speed.

Am I the only one who thinks this has to be a bug in HD Tach, and not some magic "Native Command Queuing" voodoo?

 

[Tchock]

Until then the GeForce 8200 chipset from Nvidia will have a good chance of competing with the 780G for the title of top performing integrated graphics.

My oh my. Someone remind them that this baby alone already gets quite close to a 8400GS, never mind a castrated version.

Perhaps this might be one of the factors why sideport memory's performance delta is that big?

 

[Dave Baumann]

That report is false. With a Phenom 9500 or 9600 you'll get HT3 link speeds out of RS780.

HotHardware additionally tested the link gains:

http://www.hothardware.com/articles/AMD_780G_Chipset_and_Athlon_X2_4850e_Preview_/?page=11

 

[Arun]

My oh my. Someone remind them that this baby alone already gets quite close to a 8400GS, never mind a castrated version.

But in theory, with HT3, you've got access to *more* memory bandwidth than a 8400GS and nearly as much as a 8500GT, so performance should be something between the two assuming the HT and CPU's memory controller efficiencies are good enough [EDIT: and assuming the same unit config of course, which doesn't seem to be the case for the 8200]. But either way, 780G is *very* impressive and I'm certainly not expecting MCP78 to overtake it (in any of its configs). MCP79 (aka MCP7A afaict) against G45/G47M will be an interesting battle though, I presume.

 

[v_rr]

http://www.siliconmadness.com/2008/03/amd-780g-chipset-doesnt-support-ht-30.html

How can someone beleave in such crap writen by someone that probably don´t even have the board :smile:

If you read reviews you will se that 780G is far away the best of the best in its class, and a garanteed sucess.

Only rest waiting for PUMA with Griffin + 780M + Hd 6xxx mobile.

 

[Tchock]

From an absolute standpoint, it definitely seems great (I for one am waiting for a mobo that can shoo off my [censored] R600 before a solid purchase though) and moreso for OEMs if they execute properly and deliver on time (it seems now, with this and the 4850e)


I'll say what AMD *really* needs to advertise: Phenom 9100e/8400 + 780G. Lots of flab to market about, delivering what a lopsided config can't do, overall shouldn't be a bad bet.

With the risk of derailing the whole convo, any more whispers about the Mobility HD3870 ?

 

[satein]

From an absolute standpoint, it definitely seems great (I for one am waiting for a mobo that can shoo off my [censored] R600 before a solid purchase though) and moreso for OEMs if they execute properly and deliver on time (it seems now, with this and the 4850e)


I'll say what AMD *really* needs to advertise: Phenom 9100e/8400 + 780G. Lots of flab to market about, delivering what a lopsided config can't do, overall shouldn't be a bad bet.

With the risk of derailing the whole convo, any more whispers about the Mobility HD3870 ?

Sound likely that they are waiting for the OEM to show up the Moby HD 3870 part, that was the last news I found (sorry I cannot remember where is the internet source). Probably, it might be around Mar-May time frame. But for now I could find only link to show off Moby HD 3400 & 3600 MXM modules here...
http://www.legitreviews.com/article/644/2/

 

[AnarchX]

http://i.i.com.com/cnwk.1d/i/bto/20080213/amd-q208-roadmap-small.jpg

 

[swaaye]

So, I built a sys with a Gigabyte MA78GM mobo. I wanted to have a nice, low power draw system that still had some quickness to it. I had been looking at Mini ITX stuff, but just didn't really like the prices and limitations.

Athlon 64 X2 3800+ @ 1.0v
GA-MA78GM-S2H 780G mobo
4GB DDR2 800
3x 500GB
DVDRW
1920x1200 LCD

So far my biggest prob with it is slow screen redraws in some circumstances. It is most obvious on web pages in both Firefox and IE. Scrolling up and down is dramatically slower than with a discrete vid card. It reminds me of old 2D only cards from the '90s, honestly. It's that slow. And it's strange to see this because the board will happily play 3D games and 720p/1080p vids!

Has anyone else seen this happen? I haven't seen screen redraws like this since, like I said, 10 year old video cards. I put in my Radeon 3850 as a quick test and it was buttery smooth. I sure hope it's some driver bug.... I tried various BIOS settings and verified that the RAM was in dual channel. Overclocking the IGP, CPU, RAM, and HT bus like crazy speeds up everything except web page scrolling lol.

The system otherwise is really great. I'm seeing 66W power use while idle, and that's with a 4 yr old Enermax 20pin PSU that's probably running 70-75% efficiency at those levels. May look into getting a 80+ PSU that does well in the 50-75W range.....

 

[chavvdarrr]

So far my biggest prob with it is slow screen redraws in some circumstances. It is most obvious on web pages in both Firefox and IE. Scrolling up and down is dramatically slower than with a discrete vid card.

can you try with Opera also? preferably the latest 9.50 beta from http://my.opera.com/desktopteam/blog/
, there were significant webpage redrawing speedups, so if you get same problem there, it will be 99.9% prrof to be a driver bug imho.

 

[nas_34]

Hybrid Crossfire PWS sought

Do you have to have a PWS with CrossfireX support to run hybrid crossfire? There is no additional connection on the mboard for the 780 chipset?

 

[nutball]

Has anyone else seen this happen? I haven't seen screen redraws like this since, like I said, 10 year old video cards.

Can't say I have to be honest, I'll try to pay more attention next time!

The system otherwise is really great. I'm seeing 66W power use while idle, and that's with a 4 yr old Enermax 20pin PSU that's probably running 70-75% efficiency at those levels. May look into getting a 80+ PSU that does well in the 50-75W range.....

I've built a couple of systems around this motherboard now, I'm pretty impressed particularly with its ability to play back HD material without even bothering the CPU. Idle power draw is about 20W, but that's with a single 2.5" H/D.

My big gripes so far are:

- wake on USB doesn't seem to want to work, despite applying the registry hack, etc.
- the network interface doesn't work after waking from S3

both of which are problems which should have been consigned to history in 2005.

 

[swaaye]

Can't say I have to be honest, I'll try to pay more attention next time!

Scrolling up and down on Tech Report is particularly slow, for ex. Heck, here on the forum, scrolling with the scrollbar will snag it up. Nothing like TR though.

Just for comparison, Intel Extreme 2 in an i855 doesn't do this. So I definitely expect more from something with supposedly a RV610 inside. Not on Vista here. All it has to do is perform a 2D window redraw that a G400 or TNT2 from 1999 can do better, apparently.

 

[swaaye]

can you try with Opera also? preferably the latest 9.50 beta from http://my.opera.com/desktopteam/blog/
, there were significant webpage redrawing speedups, so if you get same problem there, it will be 99.9% prrof to be a driver bug imho.

Wow. Opera is very fast. This almost gets rid of Tech Report's sluggishness on here, but I can tell it's not quite as quick as a discrete vid card. It is still a vast improvement over Firefox 3 and (less so) IE7.