Overclock.net banner

1 - 20 of 38 Posts

·
Registered
Joined
·
1,762 Posts
Discussion Starter #1
So, there's been absolutely no news articles about this, and I figured I'd start this now so someone reputable can investigate this and write an actual news article about it. It's been going on for over a year.

Right now, EVGA's website is down, but they have a list of graphics card models that are incompatible with SLI. https://www.evga.com/support/faq/afmviewfaq.aspx?faqid=59534

There's been no update on the GeForce FAQ about this:
https://forums.geforce.com/default/topic/469402/
Quote:
- Manufacturer
-- You can mix and match any manufacturer of nvidia GPUs. EVGA works fine with Asus etc.

- Clock speeds
-- You are able to use different clock speeds on each GPU working in SLI. An "overclocked" card will work with a "plain" card just fine.

- BIOS Version
-- You can use as many different BIOS versions between the cards you have in SLI as you like.
I could not find any FAQ about incompatible SLI models on Zotac's site.

A Reddit thread started 9 months ago mentions the issue:
https://www.reddit.com/r/buildapc/comments/30lw5z/psa_not_all_models_of_evga_gtx_970s_are_sli/

As well as this thread in the GeForce forums:
https://forums.geforce.com/default/topic/781585/can-t-enable-sli-with-2x-gtx-970-updated-/
 

·
Registered
Joined
·
2,816 Posts
Another reason DirectX12 and Vulkan games with their own multi-GPU support need to come out ASAP.
 

·
Premium Member
Joined
·
7,165 Posts
AFAIK this was more/less a NVIDIA issue in regards to reference 970's not being able to SLI...

Per my understanding it was a side-effect related to gimping the core...Thus, the SSC was introduced with the lanes being nixed as opposed to gimping the core.

While it is an EVGA Product, it's important to note that this occurred on a reference product...Which really muddies who carries the onus.
 

·
You broke it!
Joined
·
14,286 Posts
The reason this isn't "news" is because it isn't news....

Since the dawn of things, it has never been 100% guaranteed two differing card revisions/flavors/versions would work with each other. Most the time you can get two "different" cards, still same GPU, to work. However, that isn't always guaranteed.

Everything linked here just shows "I can't get Brand A to play nice with Brand B on my SLi". Yea, that happens.
 

·
Registered
Joined
·
1,762 Posts
Discussion Starter #5
Well, it first came to my attention because I sold both my MSI GTX 970 Gaming 4G cards. I had them in SLI with no issues, gave them to a friend, and he had no issues with SLI, either. He upgraded, so I got them back. One of the people I sold one of my cards to gave it to his son, who had an EVGA GTX 970 card, and he hasn't been able to SLI the cards. We both investigated a bit, and he even tried to flashing both cards with the same BIOS, but that didn't work. That's when I came across this issue, and the EVGA FAQ. I tried to find news articles regarding this, but I've yet to find anything.

http://forums.overclockersclub.com/topic/203639-sli-will-not-work/page-1
 

·
You broke it!
Joined
·
14,286 Posts
Quote:
Originally Posted by El_Capitan View Post

Well, it first came to my attention because I sold both my MSI GTX 970 Gaming 4G cards. I had them in SLI with no issues, gave them to a friend, and he had no issues with SLI, either. He upgraded, so I got them back. One of the people I sold one of my cards to gave it to his son, who had an EVGA GTX 970 card, and he hasn't been able to SLI the cards. We both investigated a bit, and he even tried to flashing both cards with the same BIOS, but that didn't work. That's when I came across this issue, and the EVGA FAQ. I tried to find news articles regarding this, but I've yet to find anything.

http://forums.overclockersclub.com/topic/203639-sli-will-not-work/page-1
SLi and Crossfire, always best to not mix and match if you can avoid it. Similar situation back on the ATi 4000 series for some people.
 

·
Premium Member
Joined
·
7,165 Posts
Quote:
Originally Posted by El_Capitan View Post

Well, it first came to my attention because I sold both my MSI GTX 970 Gaming 4G cards. I had them in SLI with no issues, gave them to a friend, and he had no issues with SLI, either. He upgraded, so I got them back. One of the people I sold one of my cards to gave it to his son, who had an EVGA GTX 970 card, and he hasn't been able to SLI the cards. We both investigated a bit, and he even tried to flashing both cards with the same BIOS, but that didn't work. That's when I came across this issue, and the EVGA FAQ. I tried to find news articles regarding this, but I've yet to find anything.

http://forums.overclockersclub.com/topic/203639-sli-will-not-work/page-1
MSI's weren't reference.

Again, it's important to note reference.

SSC was not reference.

It's not a bios issue...It's a chip issue. Per reference 970's the chip is gimped...Unfortunately because of how it was done, it disabled SLI.

With the SSC, they simply disabled the lanes thus, allowing for SLI.

I'm almost 99% sure, MSI's came AFTER both EVGA and ZOTAC thus, were able to modify their design, to circumvent this issue.

Who's responsible is more/less Nvidia tbh...Nvidia designed the reference card - Did the gimping themselves...That being said they are branded EVGA...So who actually takes ultimate responsibility is in the air.
 

·
Registered
Joined
·
1,762 Posts
Discussion Starter #8
Quote:
Originally Posted by PostalTwinkie View Post

The reason this isn't "news" is because it isn't news....

Since the dawn of things, it has never been 100% guaranteed two differing card revisions/flavors/versions would work with each other. Most the time you can get two "different" cards, still same GPU, to work. However, that isn't always guaranteed.

Everything linked here just shows "I can't get Brand A to play nice with Brand B on my SLi". Yea, that happens.
Quote:
Originally Posted by PostalTwinkie View Post

SLi and Crossfire, always best to not mix and match if you can avoid it. Similar situation back on the ATi 4000 series for some people.
I've never had any issues with SLI between brands. I've had GTX 460's (ASUS, MSI, Gigabyte, EVGA, PNY), GTX 560 Ti's (ASUS, MSI), GTX 580's (ASUS, MSI, Gigabyte, EVGA, Zotac, PNY), GTX 680's (ASUS, MSI, Gigabyte, EVGA, PNY), GTX 780's (ASUS, MSI, Gigabyte, EVGA, PNY), and I never had any issues with SLI between different models/brands. I haven't done SLI between two different models of GTX 970's (ASUS, MSI), 980's (ASUS), or 980 Ti's (MSI), yet, otherwise I would have probably written about this sooner.

Just sayin'.

Edit: Forgot about my little PNY!
 

·
Registered
Joined
·
1,762 Posts
Discussion Starter #9
Quote:
Originally Posted by Masked View Post

MSI's weren't reference.

Again, it's important to note reference.

SSC was not reference.

It's not a bios issue...It's a chip issue. Per reference 970's the chip is gimped...Unfortunately because of how it was done, it disabled SLI.

With the SSC, they simply disabled the lanes thus, allowing for SLI.

I'm almost 99% sure, MSI's came AFTER both EVGA and ZOTAC thus, were able to modify their design, to circumvent this issue.

Who's responsible is more/less Nvidia tbh...Nvidia designed the reference card - Did the gimping themselves...That being said they are branded EVGA...So who actually takes ultimate responsibility is in the air.
Thanks, this is the kind of information I was looking for.

MSI GTX 970 Gaming 4G and EVGA GeForce GTX 970 4GB GAMING ACX 2.0 have the issue. Just curious, is the 04G-P4-2972-KR a reference card? I wish EVGA's site was up...
 

·
Super Moderator
Joined
·
14,127 Posts
Not news, moved to nVidia
 

·
You broke it!
Joined
·
14,286 Posts
Quote:
Originally Posted by El_Capitan View Post

I've never had any issues with SLI between brands. I've had GTX 460's (ASUS, MSI, Gigabyte, EVGA), GTX 560 Ti's (ASUS, MSI), GTX 580's (ASUS, MSI, Gigabyte, EVGA, Zotac), GTX 680's (ASUS, MSI, Gigabyte, EVGA), GTX 780's (ASUS, MSI, Gigabyte, EVGA), and I never had any issues with SLI between different models/brands. I haven't done SLI between two different models of GTX 970's (ASUS, MSI), 980's (ASUS), or 980 Ti's (MSI), yet, otherwise I would have probably written about this sooner.

Just sayin'.
Your awareness of the possibility of two different GPUs not wanting to work in SLi, or Crossfire, has nothing to do with the fact that is the case, to be blunt with you.

This "news" that you started is not news, at all. It is one of the most basic understandings when dealing with a PC build. There are no guarantees differing hardware will function together. In your own "source" links, a Zotac representative even explicitly stated this.

In other words; the "problem" in this exists between the keyboard and chair, and not in the chassis.
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by Masked View Post

MSI's weren't reference.

Again, it's important to note reference.

SSC was not reference.

It's not a bios issue...It's a chip issue. Per reference 970's the chip is gimped...Unfortunately because of how it was done, it disabled SLI.

With the SSC, they simply disabled the lanes thus, allowing for SLI.

I'm almost 99% sure, MSI's came AFTER both EVGA and ZOTAC thus, were able to modify their design, to circumvent this issue.

Who's responsible is more/less Nvidia tbh...Nvidia designed the reference card - Did the gimping themselves...That being said they are branded EVGA...So who actually takes ultimate responsibility is in the air.
What do you mean by this?
 

·
Premium Member
Joined
·
7,165 Posts
Quote:
Originally Posted by El_Capitan View Post

Thanks, this is the kind of information I was looking for.

MSI GTX 970 Gaming 4G and EVGA GeForce GTX 970 4GB GAMING ACX 2.0 have the issue. Just curious, is the 04G-P4-2972-KR a reference card? I wish EVGA's site was up...
I only know that the gimping of the chip allowed for an exception...That exception was the "loss" of SLI...That's the end of my knowledge.

MSI came late to the 970 party thus, they would've been able to make a change or two.

Interestingly enough, if you look at the boxes, they don't advertise the card as actually being SLI...Neither EVGA nor Zotac, do...

I agree with Postal, really. It doesn't say SLI on the box, itself so, there's no expectation of it being able to SLI.
Quote:
Originally Posted by xxdarkreap3rxx View Post

What do you mean by this?
So the core itself wasn't "gimped", lanes to the core were disabled that, ultimately cost the reference design the ability to SLI...MSI AFAIK, was able to swap the lanes that were disabled thus, enabling SLI.

I could be wrong...This has just been a rumor of sorts for a few months and there's some clear evidence available that supports it.
 

·
Registered
Joined
·
1,762 Posts
Discussion Starter #14
Quote:
Originally Posted by PostalTwinkie View Post

Your awareness of the possibility of two different GPUs not wanting to work in SLi, or Crossfire, has nothing to do with the fact that is the case, to be blunt with you.

This "news" that you started is not news, at all. It is one of the most basic understandings when dealing with a PC build. There are no guarantees differing hardware will function together. In your own "source" links, a Zotac representative even explicitly stated this.

In other words; the "problem" in this exists between the keyboard and chair, and not in the chassis.
I think GeForce should at least update their FAQ, then. I didn't know where to put this because THERE IS NO NEWS about this, like I stated in the OP and Title. I'm hoping someone could shed more details on this, as Masked has.
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by Masked View Post

So the core itself wasn't "gimped", lanes to the core were disabled that, ultimately cost the reference design the ability to SLI...MSI AFAIK, was able to swap the lanes that were disabled thus, enabling SLI.

I could be wrong...This has just been a rumor of sorts for a few months and there's some clear evidence available that supports it.
I think you're misunderstanding it: http://www.guru3d.com/news-story/middle-earth-shadow-of-mordor-geforce-gtx-970-vram-stress-test.html
Quote:
You will notice that grayed-out right-hand L2 for this GPU right ? That is a disabled L2 block and each L2 block is tied to ROPs, GTX 970 does not have 2,048KB but instead has 1,792KB of L2 cache. Disabling ROPs and thus L2 like that is actually new and Maxwell exclusive, on Kepler disabling a L2/ROP segment would disable the entire section including a memory controller. So while the L2/ROP unit is disabled, that 8th memory controller to the right still is active and in use. Now that we know that Maxwell can disable smaller segments and keep the rest activated, we just learned that we can still use the 64-bit memory controllers and associated DRAM, but the final 1/8th L2 cache is missing/disabled. As you can see the DRAM controller actually need to buddy up into the 7th L2 unit, that it the root cause of a big performance issue. The GeForce GTX 970 has a 256-bit bus over a 4GB framebuffer, the memory controllers are all active and in use, but disabling that L2 segment tied to the 8th memory controller will result in the fact that overall L2 performance would operate at half of its normal performance.
So the issue is with the die config.

Full GM204 Block Diagram:


http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Discloses-Full-Memory-Structure-and-Limitations-GTX-970
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by El_Capitan View Post

Just curious, is the 04G-P4-2972-KR a reference card? I wish EVGA's site was up...
There wasn't really any "reference" PCB for the GTX 970. For Nvidia's own branded GTX 970, they used their GTX 980 PCB (which is considered reference).

Nvidia GTX 970: http://www.ekwb.com/configurator/step1_complist?gpu_gpus=1632
Your card: http://www.ekwb.com/configurator/step1_complist?gpu_gpus=1511

Compatibility for the reference 980 block shows: http://www.ekwb.com/configurator/waterblock/3831109869239

  • Inno3D Inno3D iChill GeForce GTX 970 HerculeZ X4 Air Boss 4GB GDDR5 (C97V-1SDN-M5DNX)
  • Inno3D Inno3D iChill GeForce GTX 970 HerculeZ X4 Air Boss Ultra 4GB GDDR5 (C97U-1SDN-M5DNX)
  • Leadtek Leadtek WinFast GeForce GTX 970 4GB GDDR5
  • MSI MSI GTX 970 4GD5 OC GeForce GTX 970 4GB GDDR5 (V317-001R)
  • Zotac Zotac GeForce GTX 970 AMP! Extreme Core Edition 4GB GDDR5 (ZT-90107-10P)
  • Zotac Zotac GeForce GTX 970 AMP! Omega Core Edition 4GB GDDR5 (ZT-90106-10P)
Some manufacturers used reference GTX 670 PCBs for the GTX 970: http://www.ekwb.com/configurator/waterblock/3831109856482

  • EVGA EVGA GeForce GTX 970 4GB GDDR5 (04G-P4-1970)
  • EVGA EVGA GeForce GTX 970 SuperClocked 4GB GDDR5 (04G-P4-1972)
  • Gainward Gainward GeForce GTX 970 4GB GDDR5 (3354)
  • Gainward Gainward GeForce GTX 970 Phantom 4GB GDDR5 (3361)
  • Inno3D Inno3D GeForce GTX 970 HerculeZ X2 4GB GDDR5 (N97V-1SDN-M5DSX)
  • Palit Palit GeForce GTX 970 4GB GDDR5 (NE5X970014G2F)
  • Palit Palit GeForce GTX 970 JetStream 4GB GDDR5 (NE5X970H14G2J)
  • PNY PNY GeForce GTX 970 4GB GDDR5 (GF970GTX4GEPB)
  • PNY PNY GeForce GTX 970 4GB GDDR5 (VCGGTX9704XPB)
 

·
Premium Member
Joined
·
7,165 Posts
Quote:
If the issue WAS with the die, itself than, it'd be present in the entire run. It's not.

The SLI issue only exists in first release "reference" product...And once again, SLI isn't present on the box.

With the SSC and everything thereafter (We can assume it's a Beta/Gamma run) the correction is made but, reference doesn't change.

Thus why the MSI's which are also "reference" work but, the EVGA/Zotac's do not.

A die change, would've been noted...It wasn't so, I'm led to believe otherwise...

Again, this was/is a rumor that's been around since samples went out so...I don't have any factual information above/beyond the fact that lane changes, don't actually have to be "noted"; a die change would have to be.
 

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by Masked View Post

If the issue WAS with the die, itself than, it'd be present in the entire run. It's not.

The SLI issue only exists in first release "reference" product...And once again, SLI isn't present on the box.

With the SSC and everything thereafter (We can assume it's a Beta/Gamma run) the correction is made but, reference doesn't change.

Thus why the MSI's which are also "reference" work but, the EVGA/Zotac's do not.

A die change, would've been noted...It wasn't so, I'm led to believe otherwise...

Again, this was/is a rumor that's been around since samples went out so...I don't have any factual information above/beyond the fact that lane changes, don't actually have to be "noted"; a die change would have to be.
EVGA doesn't advertise SLI on any of the GTX 970 boxes AFAIK. I even did a Google images search for other 970s and I don't see SLI being advertised on any of the boxes either (from ASUS, MSI, or Gigabyte).

Again you said:
Quote:
So the core itself wasn't "gimped", lanes to the core were disabled that, ultimately cost the reference design the ability to SLI...MSI AFAIK, was able to swap the lanes that were disabled thus, enabling SLI.
The gimping wasn't the disabling of lanes from the VRAM to die, it was physically on the die itself which AIBs have no control over:


Quote:
The NVIDIA GeForce GTX 980 video card uses the full GM204 Maxwell GPU that contains four graphics processing clusters (GPC). Each GPC houses four SMM units and the number of CUDA cores per SMM is 128, so there are 512 CUDA cores per GPC. That means the fully enabled GM204 GPU has four GPCs and 2,048 CUDA cores. The NVIDIA GeForce GTX 970 features a partially disabled GM204 GPU and as a result it has 13 SMM units instead of the 16 SMMs on the GeForce GTX 980. With just 13 SMMs enabled it means that there are 1,664 functional CUDA cores. The practice of cutting down a full GPU to be used in less power cards is a practice that AMD and NVIDIA have done for years and to be honest it usually doesn't have any unknown or hidden side effects.
http://www.legitreviews.com/nvidia-issues-statement-geforce-gtx-970-memory-allocation-issue_157869
Quote:
Let's start with a new diagram drawn by Alben [NVIDIA's Jonah Alben, SVP of GPU Engineering] specifically for this discussion.

http://www.pcper.com/reviews/Graphics-Cards/NVIDIA-Discloses-Full-Memory-Structure-and-Limitations-GTX-970
 

·
Premium Member
Joined
·
7,165 Posts

·
Registered
Joined
·
2,816 Posts
Quote:
Originally Posted by Masked View Post

Ohhhh. Okay. Gotcha.
I updated it with a good diagram from this guy: http://nvidianews.nvidia.com/bios/jonah-m-alben that shows it much better. From the quote above (guru3d): "Disabling ROPs and thus L2 like that is actually new and Maxwell exclusive, on Kepler disabling a L2/ROP segment would disable the entire section including a memory controller." Thus, if that were true with Maxwell, the GTX 970 would truly only have 3.5 GB of VRAM. What's better though? Only 3.5 GB or 3.5 GB + 512 MB that's not as fast?
biggrin.gif
 
1 - 20 of 38 Posts
Top