AMD sent HD 4830 samples to reviewers with missing shaders

[W1zzard]

Introduction

Today AMD released their Radeon HD 4830 card which is based on the RV770 GPU which is also used on the HD 4850 and HD 4870. In order to properly position the HD 4830, AMD reduced the number of shaders in the RV770 GPU core from 800 to 640.

http://www.techpowerup.com/articles/155/images/slide.jpg

The locking mechanism relies on fuses inside the GPU, or on the substrate - a mechanism similar to Intel's multiplier locking. It is not reversible as far as I know. It is however possible to temporarily disable _additional_ fuses to reduce the shader count even further.

The Problem
When I added shader count reading support to GPU-Z yesterday, I noticed some contradicting numbers. I have two HD 4830 samples here: one card from AMD, and one card from PowerColor. Both cards are reference spec HD 4830s, both are supposed to have a shader count of 640 - just like all other HD 4830 cards.

The GPU-Z readings however showed that the AMD card had 560 shaders and the PowerColor card had 640. How is that possible? If 80 shading units were missing, then this could mean a performance hit of up to 10%.

In order to go from 800 shaders to 640 shaders, AMD disables two SIMD units.
The original RV770 has 10 SIMDs (10 [SIMDs] * 16 [Pipes] * 5 [ALUs] = 800 Shaders). With just eight SIMDs you will get 640 shaders (= 8 [SIMDs] * 16 [Pipes] * 5 [ALUs]). The information which SIMDs are disabled is stored inside a GPU register that can be read back, and is basically a bit mask of the disabled SIMDs.

For example, on the PowerColor HD 4830 that bit mask is 10 0000 0001. Ten bits, with two bits being 1 (=disabled). This leaves us eight SIMDs, resulting in 640 shaders. On the HD 4830 from AMD the bit mask is 00 0100 0011. But hey, that's only seven SIMDs which gives us 560 shaders. Truly strange.

So let's try to confirm that the two cards really have different shader counts, which should be visible in the performance of these cards.

Testing

I first tested if this is some kind of power management feature which disables shaders when the card is idle. This is not the case. The shader count is constant under idle and load using Catalyst 8.10.
Using 3DMark06 Perlin Noise (SM 3.0) I got the following results:
AMD 4830 (560 SP) = 219.8 FPS
PowerColor 4830 (640 SP) = 250.5 FPS
HD 4850 @ 4830 clocks (800 SP) = 309.8 FPS

http://www.techpowerup.com/articles/155/images/perlin1.jpg

It is clearly visible that there is some huge difference between all three cards, while there should only be a difference between the HD 4850 and the two HD 4830s.

In some further testing I took the reference design HD 4850 and gradually reduced the number of SIMDs while running 3DMark06 Perlin.

http://www.techpowerup.com/articles/155/images/perlin2.jpg

As you can see, the FPS changes linearly with the number of shaders in the GPU. Please note that this is data from some real empiric testing. All those ten tests were performed with a real running HD 4850 (running at HD 4830 clocks for better comparability), without reboot in between, shader count was changed on the fly.

If you look back up at the first graph, you can easily see that the AMD 4830 result matches the result I got with 560 shaders and the PowerColor 4830 matches 640 shaders.

This test confirms that 3DMark06 Perlin is a valid test to determine the relative shader performance of a RV770 card. It also confirms that the register contents reflect the active number of shaders and that AMD's sample has the shading power of a HD 4850 with 560 shaders and that the PowerColor card is equally as fast as a HD 4850 with 640 shaders.


Concluding, we can see that AMD's sample card does not feature the intended shader count of 640, but comes merely with 560 shaders.

What it means for you?
So what does that mean for you, the end user?

It is unknown how many of these cards AMD sent out to reviewers, so the reviews on other sites could be affected. I only talked to one other reviewer with a HD 4830 and he had also received a 560 shader card from AMD.

Should those cards make it into retail channels you could potentially end up with a card that is quite a bit slower than intended (up to 10%). It is also unknown if only AMD's sampling is affected or if AMD board partners also sent out cards with 560 shaders.

That's why I ask all reviewers who tested the HD 4830 to download this (http://www.techpowerup.com/articles/155/images/GPU-Z.exe) build of GPU-Z and verify if their card really runs with 640 shaders. Until a statement is made by AMD (I notified them of course), buyers should also test their cards to verify they bought what was promised to them.

I am also 100% certain that shipping cards with 560 shaders was not intended by AMD at any time. Why should they make their own product slower?

Feel free to post the results of your own cards in the comments of this article so we can get an impression of how widespread this problem is.


Manufacturer Cards Tested 640 SP 560 SP
AMD 8 0 8
Sapphire 2 2 0
PowerColor 2 2 0


How to Fix
Added: Oct 23, 2008

Grab this (http://www.techpowerup.com/vgabios/25616/ATI.HD4830.512.081023.html) BIOS from our VGA BIOS Collection and flash it to the card using WinFlash (http://www.techpowerup.com/downloads/1122/Winflash_2.0.1.2.html) or ATIFlash (http://www.techpowerup.com/downloads/1123/ATIFlash_3.60.html).

This BIOS is designed for the AMD reference design samples that were sent out with 560 shaders, there is no reason to flash it on cards from other manufacturers that already have 640 shaders.

Statement by AMD
Added: Oct 24, 2008

"AMD has identified that a very limited number of ATI Radeon™ HD 4830 boards have shipped with an incorrect BIOS which may impact performance. This is in no way hardware related, and an updated BIOS fully resolves the performance limitation.
AMD has provided its board partners with the updated BIOS. Please contact your board manufacturer for instructions on updating the card BIOS."

[Solaris17]

Intriguing ill see if we have a sample.

Do you think their were 2 different batches? 560SP for internal testing until they finally decided on a shippable 640SP?

[wolf2009]

interesting .

How did you disable SIMD ?

[Steevo]

Pushing the big red button. Might get a fix for opening other doors and getting shaders back.

[eidairaman1]

well if its a disable bit, i wonder if its hard disabled or just bios disabled, because you could modify the disable bit to get the 640 shaders, but i think this was the bios encoders fault or something.

[btarunr]

well if its a disable bit, i wonder if its hard disabled or just bios disabled, because you could modify the disable bit to get the 640 shaders, but i think this was the bios encoders fault or something.

Did you read the article?

The locking mechanism relies on fuses inside the GPU, or on the substrate - a mechanism similar to Intel's multiplier locking. It is not reversible as far as I know. It is however possible to temporarily disable _additional_ fuses to reduce the shader count even further.

[ArkanHell]

Wiz, you are the man.

[newtekie1]

I am also 100% certain that shipping cards with 560 shaders was not intended by AMD at any time. Why should they make their own product slower?

They've done in the past.

[lemonadesoda]

The King of all Wizards.

THANKS FOR THE INFO!

[thebluebumblebee]

Did you pull the coolers to see if the chips are labeled the same?

[W1zzard]

Did you pull the coolers to see if the chips are labeled the same?

look at the two reviews. i took photos of both gpus

[thebluebumblebee]

oops

[suraswami]

[QUOTE=W1zzard;I am also 100% certain that shipping cards with 560 shaders was not intended by AMD at any time. Why should they make their own product slower?[/QUOTE]

Any chance Powercolor screwed it up with a wrong bios? If AMD intended to ship with 560 shaders with a reasonable reduced price over 4850 with that kind of performance hit, I think it makes sense. I believe the 4830 is going to be around $150 and 4850 can be got at that price after rebates. If the 560 shader one comes in around $110 then it will be a balanced card.

my 2 cents.

[W1zzard]

Any chance Powercolor screwed it up with a wrong bios? If AMD intended to ship with 560 shaders with a reasonable reduced price over 4850 with that kind of performance hit, I think it makes sense. I believe the 4830 is going to be around $150 and 4850 can be got at that price after rebates. If the 560 shader one comes in around $110 then it will be a balanced card.

my 2 cents.

why would amd state 640 shaders in their product brief then and risk to be exposed?

[suraswami]

why would amd state 640 shaders in their product brief then and risk to be exposed?

that part is bit shady. Had they said 560 and kept the price down with excellent media center capabilities I will jump on this.

[newtekie1]

I'm sure it wasn't intentional, really. The 560 SP cards were more than likely early samples, that probably just got mixed up when sending them out to reviewers. They certainly wouldn't intentionally send out the lower shader cards to the reviewers. I just hope none of them got sent out to retail channels.

Any chance Powercolor screwed it up with a wrong bios? If AMD intended to ship with 560 shaders with a reasonable reduced price over 4850 with that kind of performance hit, I think it makes sense. I believe the 4830 is going to be around $150 and 4850 can be got at that price after rebates. If the 560 shader one comes in around $110 then it will be a balanced card.

my 2 cents.

The HD4830 is already available on newegg for $119 after rebate, or at least it was yesterday.

Edit: The PowerColor one is actually available for $114.99 after rebate.

[Kreij]

why would amd state 640 shaders in their product brief then and risk to be exposed?

When I added shader count reading support to GPU-Z yesterday

They didn't think you would get GPU-Z to figure it out so quick?

[Polaris573]

They didn't think you would get GPU-Z to figure it out so quick?

When any product leaves a factory, in this case GPUs, an in depth analysis in not performed on every single unit rather a single unit is selected from that batch (production run of indefinite length) and only that unit is tested. This is generally enough to insure the quality of a batch of product, but if the manufacturing process is for some reason incorrectly producing 1 out of every 4 units a tester can easily miss that when only testing one unit from a batch. These are common quality control procedures, and while I do not know AMD's precise QC procedure I'm sure a scenario like similar to what was mentioned above is conceivable. When dealing with corporations there is a tendency for many people, without thinking about the situation, to jump on the "guilty until proven innocent" bandwagon. While it is entirely possible that it is conspiracy, it is more likely that when producing thousands of units a bad batch escaped.

[Kreij]

My comment was meant to be tongue-in-cheek ;)

I do not think AMD is trying to pull a fast one.

[suraswami]

I'm sure it wasn't intentional, really. The 560 SP cards were more than likely early samples, that probably just got mixed up when sending them out to reviewers. They certainly wouldn't intentionally send out the lower shader cards to the reviewers. I just hope none of them got sent out to retail channels.



The HD4830 is already available on newegg for $119 after rebate, or at least it was yesterday.

Edit: The PowerColor one is actually available for $114.99 after rebate.

oops I didn't see that. hmm I checked yesterday for 4670 but don't remember seeing this. oh well.

[dwarf]

Hey, this is Kevin from Neoseeker. I wanted to report that my ATI review sample HD 4830 is missing the shaders.

When I started work today I was sort of surprised when glancing over some other reviews stating that the HD 4830 kicks the 9800 GT's ass. That wasn't what I found. Now it makes a bit more sense.

Sort of nullifies my review, which sucks. Ah well; so it goes.

Great work W1zzard, investigating this. You are the man.

[davidletterboyz]

I noticed that all the early reviews (that shows GPU-Z screenie) that received HD4830s from AMD have 560sp. If u google a bit for reviews with GPU-Z screenies, those HIS and Powercolor samples have 640sp. I guess it's a human mistake from AMD's side, packing the wrong samples? LOL

[W1zzard]

bump for updated bios in article

[Wile E]

Was it AMD that released the BIOS? Or did a third party mod the BIOS? Just curious.

[alexp999]

So if I'm reading this right. All HD4830 GPUs are hard locked to 640SP, but for some reason the ati cards shipped with a bios locked to 560SP?
So a simple flash sets it back at 640SP?

[eidairaman1]

no, possible etching of the PCB was done either on purpose or accidentally, that or the cards were locked further because of Shader Flaws in the earliest batches, just managed to slip thru the cracks, but no worries, if the card doesnt meet the specs advertised you can take the part back and ask for your money back, but with this set of cards, i say until AMD gets the bugs out, i say go with the 4850 or 4870.

[W1zzard]

So if I'm reading this right. All HD4830 GPUs are hard locked to 640SP, but for some reason the ati cards shipped with a bios locked to 560SP?
So a simple flash sets it back at 640SP?

Yes. Whether some kind of BIOS mod can be used on HD 4830 to go from 640 -> 800 is unknown, but I highly doubt it.

[suraswami]

Yes. Whether some kind of BIOS mod can be used on HD 4830 to go from 640 -> 800 is unknown, but I highly doubt it.

But you are the Wiz, you can make it:D

[W1zzard]

bump for statement by amd

[wolf2009]

bump for statement by amd

pheew ... .

btw how do you disable SIMD ?

[lemonadesoda]

BUMP and thanks to w1zz for identifing the issue, informing AMD, getting a fix, and posting it!

:toast: :toast: :toast:

What a champ...

[newtekie1]

Ah, so it was just a faulty BIOS issue.

Does that mean a re-review of the ATi card is coming? Or did you already have to send the card back? Any chance boards slipped into the retail market with the defect?

I wonder if they flashed the BIOS with the locked pipes to possible test an even lower end configuration for the 4800 series. Maybe we will see an HD4810?

[wolf2009]

I wonder if they flashed the BIOS with the locked pipes to possible test an even lower end configuration for the 4800 series. Maybe we will see an HD4810?

and flash it to HD4830, or even 4850 !! Yay !!! :D

[W1zzard]

Does that mean a re-review of the ATi card is coming? Or did you already have to send the card back? Any chance boards slipped into the retail market with the defect?

nah already had enough troubles with the damn hd 4830 .. look at the powercolor review for the performance, look at the amd review for the rest

[newtekie1]

nah already had enough troubles with the damn hd 4830 .. look at the powercolor review for the performance, look at the amd review for the rest

Cool, thats what I figured.

Off Topic: Any tips on overclocking an HD4670? I bought the same one you reviewed from HIS, but I can't overclock it at all via software. I tried using AMD's GPU Clock Tool (0.9.8), but any time I adjust the core frequency at all the computer just locks up.

[wolf]

As you can see, the FPS changes linearly with the number of shaders in the GPU. Please note that this is data from some real empiric testing. All those ten tests were performed with a real running HD 4850 (running at HD 4830 clocks for better comparability), without reboot in between, shader count was changed on the fly.

It is however possible to temporarily disable _additional_ fuses to reduce the shader count even further.


how did you manage to alter the usable shader count without rebooting W1z?

[rui]

Hi W1zzard,

Where can I download the "560" shaders 4830 BIOS, or can you share this BIOS with us? Thanks!

[Chicco85]

I have a 4830 ES that has only 560 stream processor enabled. Flashed with the bios, i went up to 640 without problem (perlin noise test from 180 to 250 fps...).

Thanks a lot W1zzard.

http://forum.nexthardware.com/gallery/images/5226/medium/1_4830_775-1125.JPG (http://forum.nexthardware.com/gallery/file.php?n=8609&w=o)

Yes. Whether some kind of BIOS mod can be used on HD 4830 to go from 640 -> 800 is unknown, but I highly doubt it.

AMD said to me that the alu are physically disable. ;)

[hayder.master]

Intriguing ill see if we have a sample.

Do you think their were 2 different batches? 560SP for internal testing until they finally decided on a shippable 640SP?
that is really good point

[hayder.master]

ohh , w1zzard really nice work , ati now very thankful to you im sure

[Betty (Kung Pow)]

BUMP

Sry for it but this one catched my interest.

When you found this out, did you see if there were any diffrence in heat as well between the cards?

Or even better, if there were any differ in current consumption?

:D

[W1zzard]

technically there is less heat output and lower power draw. i seriously doubt you will be able to measure the temperature difference with today's in-gpu sensors. the lower power draw should show up on any half decent watt meter

[Betty (Kung Pow)]

technically there is less heat output and lower power draw. i seriously doubt you will be able to measure the temperature difference with today's in-gpu sensors. the lower power draw should show up on any half decent watt meter

Thx.

I was just thinking if this could be done on laptops to reduce power consumption, or if its already done by the power saving mechanics?

[in_awe]

W1zzard,

How did you disable SIMD units on the fly? You mention a register, so I assume you edited that somehow, but what program did you use? I found it very interesting that the AMD 4830 used significantly less power at idle than the PowerColor 4830. Only about 25% of my time on my PC is spent on gaming. My biggest gripe with the 48xx cards (and now the 47xx cards too by the looks of the initial reviews) is their idle power usage in comparison with a 3850 (my current card) or a 4670. If SIMD units could be turned off on the fly on a 4850 or 4870, say down to 4670 levels, and there was a corresponding idle power usage decrease, that would be perfect, even if a restart was required to re-enable them. The power of a high end card when needed, the low power usage of a low end card when not.

Thanks in advance for any information you can provide.

[btarunr]

W1zzard,

How did you disable SIMD units on the fly?

He didn't disable it, he received a sample that came with a defective BIOS that used only 560 SPs.

[eidairaman1]

wow bring back a topic that has been dead and over with for some time now. Attn All 4830s have the 640 Shaders. May we please close this topic?

[in_awe]

He didn't disable it, he received a sample that came with a defective BIOS that used only 560 SPs.

Wow... super-fast reply! Yes, his test unit did come with only 560 SPs enabled, but he then went on to gradually disable SIMD units on a 4850 to see if there was a linear decrease in performance. I'm wondering how he disabled them on the 4850.

[[I.R.A]_FBi]

W1zzard,

How did you disable SIMD units on the fly? You mention a register, so I assume you edited that somehow, but what program did you use? I found it very interesting that the AMD 4830 used significantly less power at idle than the PowerColor 4830. Only about 25% of my time on my PC is spent on gaming. My biggest gripe with the 48xx cards (and now the 47xx cards too by the looks of the initial reviews) is their idle power usage in comparison with a 3850 (my current card) or a 4670. If SIMD units could be turned off on the fly on a 4850 or 4870, say down to 4670 levels, and there was a corresponding idle power usage decrease, that would be perfect, even if a restart was required to re-enable them. The power of a high end card when needed, the low power usage of a low end card when not.

Thanks in advance for any information you can provide.


Couldnt you edit the bios so the idle clocks are lower?

[btarunr]

Wow... super-fast reply! Yes, his test unit did come with only 560 SPs enabled, but he then went on to gradually disable SIMD units on a 4850 to see if there was a linear decrease in performance. I'm wondering how he disabled them on the 4850.

No, a comparison between HD 4830 (with the old defective BIOS), the HD 4830 (with its proper BIOS) and HD 4850 was made. End result effectively is the comparison between RV770LE 560, 640, and RV770Pro 800 SP.

To make things ideal, a reference AMD HD 4830 board was compared to the reference HD 4850 board (same PCB layout and components).

[in_awe]

eidairaman1,

Sorry to have bumped an old post. I just stumbled on it... late to the party. I tried to find some discussion of this in other posts or on other sites, but it looks like W1zzard is the only person who took his testing to this level. I'm not actually interested in how many shaders a 4830 has, but rather in being able to disable shader units on any 48xx card in order to reduce idle power consumption. W1zzard was able to disable them on a 4850 without even rebooting... I'm just hoping to benefit from his advanced knowledge.


[I.R.A]_FBi,

Lowering the idle clocks does help, but not enough. I bought a 4850 when they first came out and tried super-low settings. That was when the BIOSes were shipping with the wrong settings in them. The idle power usage never came anywhere near 3850 levels though. W1zzards testing potentially opens up a new way to lower power usage.

[in_awe]

No, a comparison between HD 4830 (with the old defective BIOS), the HD 4830 (with its proper BIOS) and HD 4850 was made. End result effectively is the comparison between RV770LE 560, 640, and RV770Pro 800 SP.

To make things ideal, a reference AMD HD 4830 board was compared to the reference HD 4850 board (same PCB layout and components).

Check out the second half of the Testing section in the first post of this thread. W1zzard shows the results of gradually disabling the SIMD units on a 4850.

[W1zzard]

i still have the card. it doesnt matter for this part of the conversation anyway.

at this time i'm not willing to disclose how to reduce the number of shaders

edit: thread has been cleaned up

edit2: there is no "program" to do it

[lemonadesoda]

W1zzard,

How did you disable SIMD units on the fly? You mention a register, so I assume you edited that somehow, but what program did you use? I found it very interesting that the AMD 4830 used significantly less power at idle than the PowerColor 4830. Only about 25% of my time on my PC is spent on gaming. My biggest gripe with the 48xx cards (and now the 47xx cards too by the looks of the initial reviews) is their idle power usage in comparison with a 3850 (my current card) or a 4670. If SIMD units could be turned off on the fly on a 4850 or 4870, say down to 4670 levels, and there was a corresponding idle power usage decrease, that would be perfect, even if a restart was required to re-enable them. The power of a high end card when needed, the low power usage of a low end card when not.

Thanks in advance for any information you can provide.

I welcome your question. Never mind about it being an old thread. It is relevant and topical, and your question is better placed here than starting a new thread and having to do a whole lot of explaining. I hope w1z can share some insights...

And welcome to TPU. :pimp:


i still have the card. it doesnt matter for this part of the conversation anyway.

at this time i'm not willing to disclose how to reduce the number of shaders
Shady dealer

[in_awe]

i still have the card. it doesnt matter for this part of the conversation anyway.

at this time i'm not willing to disclose how to reduce the number of shaders

edit: thread has been cleaned up

edit2: there is no "program" to do it

Thank you for the reply W1zzard (and for cleaning up the thread). Do you plan to release that information at some point in the future, or perhaps include it as part of a tweaking program? The low-power freaks like me would greatly appreciate it. Or are you planning on keeping it to yourself forever? (cue evil laugh)

Of course, I just realized it may be a hardware hack, which would exclude me from using it. I'm completely inept when it comes to using a soldering iron.

[W1zzard]

Thank you for the reply W1zzard (and for cleaning up the thread). Do you plan to release that information at some point in the future, or perhaps include it as part of a tweaking program? The low-power freaks like me would greatly appreciate it. Or are you planning on keeping it to yourself forever? (cue evil laugh)

Of course, I just realized it may be a hardware hack, which would exclude me from using it. I'm completely inept when it comes to using a soldering iron.

as tweaking program might be possible in the future.

it's not a hardware hack, it can be done in realtime, even while the card is rendering.

i doubt there is that much to be gained power consumption wise. it is definitely not 50% for going to half the shaders. educated guess might be 5%

[in_awe]

as tweaking program might be possible in the future.

it's not a hardware hack, it can be done in realtime, even while the card is rendering.

i doubt there is that much to be gained power consumption wise. it is definitely not 50% for going to half the shaders. educated guess might be 5%

OK, thanks. I was just guessing on the idle power usage based on your original AMD 4830 review (with only 560 shaders) versus your PowerColor 4830 review. The AMD 4830's idle usage was close to 4670 levels. Achieving 4670 levels with a 4850 or one of the new 4770s would be awesome.