Zotac GeForce 9300 (MCP7a) motherboard

Looks like a fantastic HTPC board. I really like the colour combo of orange and blue. I may use that in a future mod lol.

Once again a very nice review.

In the last section:

Zotac's board isn't the fastest or most capable out there, but for just being stable and for HTPC playback it's the best Intel solution currently available. The Nvidia GeForce 9300 ...

Shouldn't that be nvidia ??

Great review !

Best integrated graphics solution for Intel processors

What? No overclocking section with your cascade phase change setup? You are losing your touch.

Originally Posted by Jojii
What? No overclocking section with your cascade phase change setup? You are losing your touch.

Cant do jack all overclocking cause there's no CPU voltage change anyway :p. I tried, got about 80MHz FSB which is pants.

GPU overclocks like a rocket though - we maxed it out on the Nitro :D

I know I keep repeating myself but...

An HTPC board without "analogue" TV out?

S-Video...Component?

Xir

How does it do on games? A quick benchmark on something not too demanding would be nice to see!
I do like the colour scheme and it seems a good board so long as your aren't trying to overclock it, which you aren't going to in an HTPC. I suppose you might want to underclock though.

Originally Posted by Cupboard
How does it do on games? A quick benchmark on something not too demanding would be nice to see!
I do like the colour scheme and it seems a good board so long as your aren't trying to overclock it, which you aren't going to in an HTPC. I suppose you might want to underclock though.

I didn't have time or space in the office right now for games as I'd have to retest ALL the other AMD boards again. We reviewed it as a HTPC and general productivity board since that's what most people will be using it for. Not many really use GeForce Boost or the mGPU to play 800x600 with everything turned off.

Xir - only very, very few boards offer(ed) component without an adapter and none these days. It's all HDMI, VGA and DVI!

I've got some plans to include the GeForce 9300 board in my 4550 review that's currently being worked on... :)

Originally Posted by Tim S
I've got some plans to include the GeForce 9300 board in my 4550 review that's currently being worked on... :)

There, that's the reason I didn't of course Tim had it covered..

first i thought. aah! a micro ATX HTPC board!. And stopped reading when I saw the IDE and FDD cables and most importantly the Active cooling solution.

From the other geforce 9 series article:

Nvidia also becomes the second company to drop IDE from its chipset line-up too, after Intel did way back with ICH8

According to the nvidia website that is not the case

Nvidia's GeForce MCP7a does include 7.1 channel LPCM HD audio support over HDMI 1.3 without any need for an extra TDMS like Intel's G45, so this could certainly perk the interest of some HTPC enthusiasts. Coupled with dual digital outputs - a first for integrated video

I thought one of the changes in the G45 relative to previous intel IGPs was that it had a built in TMDS (not TDMS) making it easier for motherboard manufacturers to implement the digital video outputs. Is that not the case? Also the 690G supported dual digital outputs so the 9300 is not the first.

Instead we matched another low power CPU, Intel's previously inexpensive E6400 at 2.13GHz against AMD's recent BE-2350 CPU also at 2.3GHz

Believe that should be "also at 2.1GHz".

Memory latency is also incredibly high for the Zotac board too, due to the fact that when set to 1,333MHz FSB - 800MHz memory (not an uncommon setting by any stretch) this means the memory controller runs in unlinked mode, rather than the more efficient linked setting.

According to this that's not necessarily the prime reason for the lack of performance.

On page 10

we recorded the average and minimum CPU usage during this section of the film.

Think you mean maximum not minimum.

The performance of the G45 with the Casino Royale BD-DVD looks suspicious to me. The cpu usage seems so high that it might be a software issue somewhere.

Both the AMD and Nvidia chipsets excel here with very little power use because they have ample shader power to hand.

With a more efficient CPU than AMD, the Core 2 Duo/GeForce 9300 combination offers a lower CPU use than the GeForce 8200 with the same shader power.

What role do the shaders have in video decoding? Don't they have dedicated logic for decoding video e.g. UVD?

However, the difference between the Zotac GeForce 9300 and Jetway GeForce 8200 with the same shader horsepower is quite large in the 720p playback test.

There have been clear improvements with the GeForce 9300 and PureVideo HD, and the last GeForce 8200 we looked at, despite the fact the core shader count and core graphics performance should be the same.

Correct me if I'm wrong but doesn't the 9300 have 16 shaders to the 8200/8300s 8?

we left the sharpness at an Nvidia recommended 65 percent in the Nvidia Control Panel

Is that referring to the "edge enhancement" slider?

I have to say that the HDHQV score for the 9300 was very impressive if not all that relevant for blu-ray movies as you pointed out.

Overall a promising chipset which will only get better with driver and bios revisions.

Nice review Bindi

Originally Posted by Renoir
From the other geforce 9 series article:According to the nvidia website that is not the caseI thought one of the changes in the G45 relative to previous intel IGPs was that it had a built in TMDS (not TDMS) making it easier for motherboard manufacturers to implement the digital video outputs. Is that not the case? Also the 690G supported dual digital outputs so the 9300 is not the first.

The 690G didn't and nor does the 780G - both have the option of either DVI or HDMI - not both at once - there's a digital switch behind the rear I/O that selects them.

Intel G45 still doesn't have in built TMDS (consistent typo - D is closer than M :p) - the mini-ITX board has two. It's cheaper for Intel not to have to pay HDCP and HDMI licenses and also because they have massive OEM business most of these customers still want VGA only. This is also probably why G45 has a HDCP repeater issue perhaps.

On the website it's referring to the motherboard above I think - it says DDR3-1333 without mentioning DDR2. Nvidia were clear about it not including IDE and the Zotac motherboard would not include a JMB386 controller if it didn't need it :)

Yea it's missing advanced path which there is a setting of in the drivers and I was told it was disabled but forgot to put it in - I was waiting for more BIOS details about the other two settings as well. But either way, unlinked mode is still **** and the memory settings are completely FUBAR for DDR2 right now.

Believe that should be "also at 2.1GHz".

Yep, it was 2.1GHz and 2.13GHz

On page 10 Think you mean maximum not minimum.

They are? (Unless Tim changed it..) I started recording max (should have done in the first place in hindsight) to see if it was always smooth and to show the overhead.

The performance of the G45 with the Casino Royale BD-DVD looks suspicious to me. The cpu usage seems so high that it might be a software issue somewhere.

It could well be - I'll revisit it when I come to try the E5200 in the G45 review I've got coming.

What role do the shaders have in video decoding? Don't they have dedicated logic for decoding video e.g. UVD?
Correct me if I'm wrong but doesn't the 9300 have 16 shaders to the 8200/8300s 8?
Is that referring to the "edge enhancement" slider?

No 8200/8300 has 16 according to the GPU-Z I shot in the Jetway 8200 article.

Decoding has its own specific hardware, but the shaders do the deinterlacing and some other processing - resizing too iirc. But it's all down to software to use it efficiently - Intel's ClearVideo deinterlacing codec is fantastic compared to AMD's and Nvidia's.

I have to say that the HDHQV score for the 9300 was very impressive if not all that relevant for blu-ray movies as you pointed out.

Indeed it is, and it's an arbitrary number but I try to include as much info as possible for comparison because it's so personal POV.

Nice review Bindi

Cheers dude :D I always find my stress level increases when I see your posts though :P:P:P

The 690G didn't and nor does the 780G - both have the option of either DVI or HDMI - not both at once - there's a digital switch behind the rear I/O that selects them.

The info I was going off was from our discussion of the topic here (1st post) that was based on the info here

Intel G45 still doesn't have in built TMDS (consistent typo - D is closer than M ) - the mini-ITX board has two. It's cheaper for Intel not to have to pay HDCP and HDMI licenses and also because they have massive OEM business most of these customers still want VGA only.

Fair enough.

This is also probably why G45 has a HDCP repeater issue perhaps.

hmmmm interesting theory I like it.

On the website it's referring to the motherboard above I think - it says DDR3-1333 without mentioning DDR2. Nvidia were clear about it not including IDE and the Zotac motherboard would not include a JMB386 controller if it didn't need it

I agree that only mentioning ddr3 is suspicious but there is nothing on the page to suggest that it's associated with any particular motherboard. The inclusion of the JMB386 does suggest it's dropped support for pata but it's not conclusive given that intel support gigabit ethernet on the ICH10 but a lot of boards still opt for an auxiliary controller instead so there is precedence.

Yea it's missing advanced path which there is a setting of in the drivers and I was told it was disabled but forgot to put it in - I was waiting for more BIOS details about the other two settings as well. But either way, unlinked mode is still **** and the memory settings are completely FUBAR for DDR2 right now.

No worries for forgetting. I agree that the memory situation is crap and certainly suggests the product launch was rushed.

They are? (Unless Tim changed it..) I started recording max (should have done in the first place in hindsight) to see if it was always smooth and to show the overhead.

Sorry for not being clear. The graphs are correct but the preceding sentence "we recorded the average and MINIMUM CPU usage during this section of the film" is what I was referring to.

It could well be - I'll revisit it when I come to try the E5200 in the G45 review I've got coming.

Look forward to it.

No 8200/8300 has 16 according to the GPU-Z I shot in the Jetway 8200 article.

The article here says it's 8 which is backed up by benchmarks here & here showing the 9300 significantly outperforming the 8300. This would also make sense given that the 9300 is built on a 65nm process while the 8300 is built on a 80nm process (I realise some of the "extra" die space is taken up by the memory controller that isn't present on the 8300).

Decoding has its own specific hardware, but the shaders do the deinterlacing and some other processing - resizing too iirc. But it's all down to software to use it efficiently - Intel's ClearVideo deinterlacing codec is fantastic compared to AMD's and Nvidia's.

I would have thought the post processing would take roughly the same cpu usage on each chipset just that one might do it better than the other if it has more shader power. I agree that intel have really nailed SD DVD playback which is still the most important aspect for me personally.

Indeed it is, and it's an arbitrary number but I try to include as much info as possible for comparison because it's so personal POV.

As I'm sure is clear by now the more info the better as far as I'm concerned so don't change a thing ;)

Cheers dude I always find my stress level increases when I see your posts though :P:P:P

LOOOOOOOOOL Breaking news forum poster causes author blood pressure issues (the bit community would never forgive me :D)

Originally Posted by Bindibadgi

Xir - only very, very few boards offer(ed) component without an adapter and none these days. It's all HDMI, VGA and DVI!

Adapter is fine...is ther an Adapter from "HDMI, VGA and DVI" to SVideo?

I guess as long as Philips manages to sell DVD-Players that only have a Scart and a Component out (so you can hook it up to your beamer in the worst possible quality) there's not enough of a problem.

Of course this problem will fade over the next few years...but then again PC's are just used a few years.

Flatscreens really started selling around here 3-4 years ago...HD-ready (720) without HDMI or DVI or VGA in.
For about 1-2 years now, the mainstream models (still HD-ready 720) have been equipped with VGA in woohoo!

Xir

Any comments on my post Bindi especially related to my comments on the issues of pata support & 16vs8 shaders?

Also in case you're not already reading this I recommend the following blogger for information on intel integrated graphics. In particular his post on the HDCP repeater issue is very interesting and suggests that your theory above on the issue may not be correct.

Each TPC (thread processing cluster) in G8x/G9x derived GPUs has 16 shaders, split into two blocks of eight SPs - all GPUs after G80 have eight texture address/filtering units per TPC, while G80 had four texture address/eight texture filtering units per TPC. These are shared amongst the two blocks of eight SPs and each texture unit is accessible by ANY of the 16 stream processors in that TPC.

I can't see how Nvidia could make a part with less than 16 stream processors if it's derived from G8x/G9x and less than 24 stream processors if it's derived from GT200, as there would have to be some fairly substantial changes to the transistor layout.

Pretty much all of the GPUs Nvidia launches follow on from when the new architecture is launched (OK, GT200 is hardly 'new', but it is tweaked) and the number of TPCs and ROPs are just scaled to suit the needs of the price point Nvidia is trying to hit with a particular GPU. The actual TPCs and ROPs remain the same - they're self-sufficient and the number of them is scalable. If GeForce 8200 was an eight SP part, it would be as much of a new architecture as GT200 is when compared to G80/G92 - it's 24 SPs and 8 TMUs per TPC. :)

Further to support this was the fact that Nvidia could not do mismatched SLI due to what I've been told are scheduling issues and massive driver overhead. To the best of my knowledge, the only cards that support GeForce Boost are cards with 16 stream processors and it's because of this 'problem' Nvidia has had with getting mismatched SLI to scale properly.

Thanks for the reply Tim! I've just read another couple of reviews which also say it's 16 shaders so I suppose we just have to put it down to an error on anandtech's side. In that case what do you attribute the significantly increased performance to?

Also do you have any idea why most motherboard manufacturer's appear to be releasing mobos with only single-link dvi when the chipset supports dual-link?

Originally Posted by Xir
Adapter is fine...is ther an Adapter from "HDMI, VGA and DVI" to SVideo?

I guess as long as Philips manages to sell DVD-Players that only have a Scart and a Component out (so you can hook it up to your beamer in the worst possible quality) there's not enough of a problem.

Of course this problem will fade over the next few years...but then again PC's are just used a few years.

Flatscreens really started selling around here 3-4 years ago...HD-ready (720) without HDMI or DVI or VGA in.
For about 1-2 years now, the mainstream models (still HD-ready 720) have been equipped with VGA in woohoo!

Xir

No. S-Video, like composite requires a TV encoder like the old ATI Rage Theatre or Philips SAE chips in early hardware. At some point for some GPUs this was incorporated onto the silicon or was done in drivers (I'm not sure), but either way it's all about VGA, DVI and HDMI these days and occasionally component. It's simply a cost choice for motherboards - they've never really offered it with commitment. Even graphics cards are dropping the 7-pin support occasionally.

Ren - I'll read the rest of your post in a minute and edit this, but I had Nvidia sitting there telling me the 8200 was 8-shader when I questioned it was 16, yet, unless GPU-Z is consistently wrong I can only assume 16. GPU-Z currently doesn't read 9300 so I can't check the difference.

Iirc Tech Report confirmed our finding of 16 for 8200 too originally, maybe I'm wrong. I remember reading it though.

It can mix-match clocks between the 8400 GS (16 shader part) fine, but mis-matching shader quantities is far more difficult.

EDIT: IDE - maybe Nvidia originally included it but it's currently broken in the final silicon? The chipset is heavily delayed after all. The difference with Intel GbE is that the Intel GbE solution costs more than a marvell or realtek chipset who companies already have contracts with (they buy huuuuge amounts for every board), whereas using native IDE requires no extra chipset = saved $$$.

690G - I've no board to check any more and I can't remember for the life of me. AMD also said to me that the 780G PCI-E x16 lane could be split into two x8s, but in actual fact, it couldn't. They also said the SB700+ would have HyperFlash support... Our discussion about LVDS was different - the LVDS hardware is (usually) external to the GPU and powers a display on its own. I'm not doubting the slide but my question is: why would AMD go backwards from dual support on the 690G to single digital output on 780G?

The 9300 is faster simply because it uses an Intel CPU and has a direct access to main memory without going through the CPU. I would guess it's probably mostly CPU driven at very low resolution and graphical features - the 9300 is faster than the 790GX even and it has a lower clock.

As for shader processing - I think driver optimisation is a bigger limitation and access to performance than raw shader power. It's also dependent on the software playing it too - PowerDVD/Archos/WinDVD/MPC-HC all handle the playback %age differently.

Iirc Tech Report confirmed our finding of 16 for 8200 too originally, maybe I'm wrong. I remember reading it though.

You are correct in that the tech report says the 8200/8300 has 16 shaders in their 8300 review but in their 9300 review they appear to have changed their minds to 8 shaders. This sort of inconsistency drives me mad!

EDIT: IDE - maybe Nvidia originally included it but it's currently broken in the final silicon? The chipset is heavily delayed after all.

This is an interesting theory and you may well be right. This chipset has been launched in such a way that many things may not be simple or as they appear to be. At the end of the day a products capabilities at launch should be accurately reflected on all promotional and informational material provided by the manufacturer. For example if the 9300 does not support IDE in any way then it shouldn't have it listed as a feature on the web page I linked to.

The difference with Intel GbE is that the Intel GbE solution costs more than a marvell or realtek chipset who companies already have contracts with (they buy huuuuge amounts for every board), whereas using native IDE requires no extra chipset = saved $$$.

You're quite right, on reflection it's not quite the same situation. Coincidently now that I think about it, when intel dropped support for IDE didn't they still have to leave something in the south bridge for auxiliary IDE chips to "hook" into? Perhaps that's why the web page states support for one IDE channel. Support being used in that oh so annoying sense like with early gpus having "support" for HDCP!

690G - I've no board to check any more and I can't remember for the life of me. AMD also said to me that the 780G PCI-E x16 lane could be split into two x8s, but in actual fact, it couldn't. They also said the SB700+ would have HyperFlash support... Our discussion about LVDS was different - the LVDS hardware is (usually) external to the GPU and powers a display on its own. I'm not doubting the slide but my question is: why would AMD go backwards from dual support on the 690G to single digital output on 780G?

Why AMD would go backwards was much discussed by myself & others in your 780G review with no conclusive outcome. I doubt I'd purchase a 690G now anyway so I guess the debate is pretty academic so I think we'll just leave it.

The 9300 is faster simply because it uses an Intel CPU and has a direct access to main memory without going through the CPU. I would guess it's probably mostly CPU driven at very low resolution and graphical features - the 9300 is faster than the 790GX even and it has a lower clock.

I realise that lower resolutions tend to be more cpu bound but I wouldn't have thought that would be the case with a "weak" gpu. Even at low resolutions I would have expected the gpu in this case to be the bottleneck. If you're right in that the 8300 & 9300 both have 16 shaders (most of the evidence is leaning that way) then you must be right and so what I would have expected above is clearly wrong. As for the direct access to main memory I asked a question that was discussed on the latest tech report podcast where Scott mentioned that this may not be such a big issue for gaming due to bandwidth being far more important than latency for gaming.

As for shader processing - I think driver optimisation is a bigger limitation and access to performance than raw shader power. It's also dependent on the software playing it too - PowerDVD/Archos/WinDVD/MPC-HC all handle the playback %age differently.

To be honest I think my initial question on your comments in this regard in the article was just me being nit picky with the way you worded it so I think I'll just forget about it. I can't even remember what my initial gripe was about anyway.

Intel G45 still doesn't have in built TMDS (consistent typo - D is closer than M ) - the mini-ITX board has two. It's cheaper for Intel not to have to pay HDCP and HDMI licenses and also because they have massive OEM business most of these customers still want VGA only.

The datasheet here states that the G45 supports DVI-HDMI-DisplayPort "Natively" and has "Integrated" HDCP keys. It has 2 digital "ports" each of which can be configured to either DVI, HDMI or Displayport. Both ports also support HDCP but not at the same time. They support resolutions up to 2560x1600@60Hz. If I've read the datasheet correctly then the G45's display output capabilities are pretty darn good!

This is also probably why G45 has a HDCP repeater issue perhaps.

On further reading this is definitely not the cause. The cause as explained here appears to be that:

the standard COPP protocols that all of these players use do not support the use of repeaters (even though our competition appears to be ignoring these strictures), and the software players don't yet use OPM (which does support repeaters).

ArcSoft and Corel have since updated their software and now have HDCP repeater support for the G45 but AFAIK PowerDVD hasn't yet been updated

It could well be - I'll revisit it when I come to try the E5200 in the G45 review I've got coming.

The cpu usage numbers in this review are definitely not right and now I know why. I've read the better part of ten reviews on the G45 and noticed that all the ones which didn't appear to have HA working were using intel motherboards. Turns out that you need to change the PAVP setting in the BIOS to Lite as stated here

Originally Posted by Renoir
You are correct in that the tech report says the 8200/8300 has 16 shaders in their 8300 review but in their 9300 review they appear to have changed their minds to 8 shaders. This sort of inconsistency drives me mad!

That's because NV went around telling everyone the 8200/8300 had only 8

The cpu usage numbers in this review are definitely not right and now I know why. I've read the better part of ten reviews on the G45 and noticed that all the ones which didn't appear to have HA working were using intel motherboards. Turns out that you need to change the PAVP setting in the BIOS to Lite as stated here

Cheers! ;)

That's because NV went around telling everyone the 8200/8300 had only 8

That would make sense chronologically if they had said 8 in the 8300 review and 16 in the 9300 review but they did it the other way around which doesn't make sense to me. If Nvidia were telling people it was 8 when the 8300 first came out then I would expect it to have been 8 in the earlier review and 16 in the latter once they had had time to figure out that Nvidia had some incompetent people giving out false information.


Any comments on what I found in the G45 datasheet as regards display output capabilities? It seems pretty clear to me that the G45 is now like competing chipsets in that it has native support for digital outputs so I would be interested to know why you feel the G45 still operates similar to previous intel IGPs in so far as requiring 3rd party chips to enable digital outputs?

No they said 16 earlier and 8 later for 8x00, even though GPU-Z reads the 8x00 as 16 all the time.

I haven't even read the G45 stuff tbh, brain is elsewhere and I'll forget it by the time I actually test G45 properly.

Why I feel that way: because I was told by several TWese mobo manuf. that it does and because it has two digital chips by the HDMI and DVI outputs (although I haven't cross-matched these online yet).

No they said 16 earlier and 8 later for 8x00, even though GPU-Z reads the 8x00 as 16 all the time.

Wow! In that case I give up as none of the sequence of events makes much sense. I don't envy your job of trying to get accurate and timely information from these people! Perhaps a column on the subject might be an idea (prob don't wanna p*** them off so I understand if you decline :D)?

Why I feel that way: because I was told by several TWese mobo manuf. that it does and because it has two digital chips by the HDMI and DVI outputs (although I haven't cross-matched these online yet).

I'm not sure how what you've just said and what's in the datasheet can be reconciled but I'm sure we'll get to the bottom of it. I guess we'll continue the discussion when you publish your G45 review (give you a break till then :)).

I don't know if you've been reading the IGP chronicles series over at anandtech but they highlight a VERY important issue with Cool 'n' quiet causing dropped frames when playing hd video and I wander whether this is something you plan to address in your G45 review given that it could be a deal breaker for those deciding between an intel and amd platform for a HTPC?

Any idea where one could buy this board from? I'm not having much luck with google.....