Sunday, August 15th 2021
PSU Expert Aris Mpitziopoulos Responds to Gigabyte Statement on GP-P750/850 GM Design Flaws
Cybenetics certification program's chief engineer, a renowned expert with switching power supplies, and TechPowerUp contributor, Dr Aris Mpitziopoulos, released detailed commentary on the recent statement released by GIGABYTE on the GP-P750/850 GM series PSU design flaw reportage in the press. The earliest of this was broken on TechPowerUp, with our December 2020 review of the GP-P750GM, which led with the headline "With an Explosive Attitude."
Mpitziopoulos dives into the technical details of the design flaw being not just a badly programmed Over-Power Protection (OPP) safety mechanism, but a flawed hardware design overall. Since he reviewed PSUs from this series for TechPowerUp and Hardware Busters, he shares his perspective on how the reviews went, more importantly, the back-and-forth with GIGABYTE when the flaws were discovered.
The following is a statement by Aris Mpitziopoulos:
I didn't want to bother with this matter again initially, but something was troubling me. It can't be right, Gigabyte blaming us reviewers, or users, for abusing its power supplies and not its engineer. So after some thought, I decided to reply to their statement issued on 13th August 2021 about the GP-P750/850 GM PSUs. Below you will find Gigabyte's allegations, as they are written in its statement and my responses.
GIGABYTE: takes reports of this manner extremely seriously and therefore would like to address the reported potential issues as follows…
Aris response: I reported the issues I faced with the GP-P750GM power supply in late October 2020, before I post its review on TechPowerUp and the video review on the Hardware Busters YT channel, and GIGABYTE responded that its engineers tested five units and found no problems. They didn't ask for my bad sample back for failure analysis, which is the typical procedure, and they didn't offer a second sample to continue the review. I have kept all of my correspondence with GIGABYTE's respective team, in case there is any doubt.
GIGABYTE: The OPP safety feature is designed to shut down the unit when the power load exceeds the wattage the unit was designed to operate within. The OPP was set to 120% to 150%, 1020 W~1300 W for GP-P850GM, and 900 W~1125 W for the GP-P750GM.
Aris response: OPP is to protect the PSU from failures. That said, GIGABYTE's engineers should have configured it accordingly. Some platforms with top-notch and tolerant to stress components can have higher OPP settings than other lower-end platforms. It is up to the manufacturer to correctly set OPP to effectively protect the power supply under all conditions and the system that the PSU feeds with power. Lastly, OPP with a 30% range is too high. GIGABYTE should ask for a lower range.
GIGABYTE: We were made aware by third parties of concerns regarding potential issues of the GP-P850GM and GP-P750GM tripping at high wattages when tested via DC Electronic Load equipment for extended lengths of time repeatedly close to the 120% to 150% OPP trigger point. This level of extended testing could severely reduce the lifespan of the product and components of the GP-P850GM and GP-P750GM.
Aris response: First of all, there was no prolonged testing period under overloads since most samples died within a few minutes of testing as Steve (Gamer Nexus) mentions in his video. In my case, the GP-P750 GM sample that I tested shut down during a short period of OPP evaluation and exploded once I tried to start it again to continue testing, with no load on its rails. Even if extended testing at high loads was the case, the engineers should use lower OPP triggering points from the moment they are well aware that the platform can handle higher than normal loads. Finally, reducing the lifespan is an entirely different story from exploding parts, which clearly shows a problem with the OPP setting, which GIGABYTE believes is the culprit.
GIGABYTE: GIGABYTE has made adjustments and lowered the OPP on GP-P850GM and GP-P750GM…. from 120% ~ 150% to 110% ~ 120%
Aris response: GIGABYTE noticed, after our findings, that the platforms cannot withstand a 120%-150% OPP rating and decided to lower it. The problem is that 110%-120% OPP is impossible with analog controllers, which use resistors to adjust OPP. These resistors drift with temperature. In other words, their resistance changes according to the operating conditions, so it is impossible to achieve such a small OPP range under both cold and hot conditions. The only way to achieve a tightly set OPP is through digital circuits, an MCU. Finally, GIGABYTE doesn't mention the operating conditions under which the new OPP settings apply.
GIGABYTE: GIGABYTE would like to stress the potential issues that were reported, only seemed to occur after very long time periods of extreme load testing via DC Electronic load equipment and would not be typical of any real world usage.
Aris response: Gamer Nexus' samples died in a matter of minutes, and my sample died after a short period of OPP evaluation. Moreover, a quick look at the user reports (Newegg, forums, etc.) shows that most of these PSUs died under normal conditions. With so many failures reported on Newegg reviews, it cannot just be a coincidence.
GIGABYTE: GIGABYTE GP-P850GM and GP-P750GM PSU's included industry standard power protection designs OCP, OTP, OVP, OPP, UVP, and SCP.
Safety certification from various countries to ensure safe and stable operation of your system.
Aris response: From the moment these units have a CE certification, I would love to check the corresponding CE reports including protection features evaluation. To the best of my knowledge, no safety certification evaluates the PSU's protection features.
GIGABYTE: Despite the fact that both before & after OPP adjustment versions are reliable for real world usage
Aris response: Many users that bought these products have a different opinion and experience. Also, OPP is not there for us reviewers only, but it should protect the PSU under all conditions. Otherwise, there is no point in having this protection feature when it doesn't save the PSU. And also, who and what defines real-world usage? For me, typical use can be having my PC idle most of the time while other users play games most of the time, stressing the entire system. Other users can run tests with Furmark and Prime95 at the same time. My point is, real-world usage varies from user to user.
GIGABYTE: Serial Number below can apply for Return and Exchange service
Aris response: GIGABYTE had made OPP changes to some production batches but didn't inform the people that bought units with high OPP about this or even give them the chance to replace their units, just to be on the safe side. They applied a silent fix, and this means that they were troubled enough to do it.
Conclusion: This is not just a badly set OPP since many units died under moderate loads and within short periods. I strongly believe that this is just a bad design, the FETs are not driven correctly, and although in quick pre-checks and normal conditions, the PSU can be ok, there are cases (not only under stress) where they fail. The timing of the FETs is not correct, and this is due to a lousy gate driver or a lousy implementation. Of course, I cannot be dead sure without any samples, older and newer generations, in my hands to test and break apart for failure analysis.
This article explains the problem I described above in detail.
Mpitziopoulos dives into the technical details of the design flaw being not just a badly programmed Over-Power Protection (OPP) safety mechanism, but a flawed hardware design overall. Since he reviewed PSUs from this series for TechPowerUp and Hardware Busters, he shares his perspective on how the reviews went, more importantly, the back-and-forth with GIGABYTE when the flaws were discovered.
I didn't want to bother with this matter again initially, but something was troubling me. It can't be right, Gigabyte blaming us reviewers, or users, for abusing its power supplies and not its engineer. So after some thought, I decided to reply to their statement issued on 13th August 2021 about the GP-P750/850 GM PSUs. Below you will find Gigabyte's allegations, as they are written in its statement and my responses.
GIGABYTE: takes reports of this manner extremely seriously and therefore would like to address the reported potential issues as follows…
Aris response: I reported the issues I faced with the GP-P750GM power supply in late October 2020, before I post its review on TechPowerUp and the video review on the Hardware Busters YT channel, and GIGABYTE responded that its engineers tested five units and found no problems. They didn't ask for my bad sample back for failure analysis, which is the typical procedure, and they didn't offer a second sample to continue the review. I have kept all of my correspondence with GIGABYTE's respective team, in case there is any doubt.
GIGABYTE: The OPP safety feature is designed to shut down the unit when the power load exceeds the wattage the unit was designed to operate within. The OPP was set to 120% to 150%, 1020 W~1300 W for GP-P850GM, and 900 W~1125 W for the GP-P750GM.
Aris response: OPP is to protect the PSU from failures. That said, GIGABYTE's engineers should have configured it accordingly. Some platforms with top-notch and tolerant to stress components can have higher OPP settings than other lower-end platforms. It is up to the manufacturer to correctly set OPP to effectively protect the power supply under all conditions and the system that the PSU feeds with power. Lastly, OPP with a 30% range is too high. GIGABYTE should ask for a lower range.
GIGABYTE: We were made aware by third parties of concerns regarding potential issues of the GP-P850GM and GP-P750GM tripping at high wattages when tested via DC Electronic Load equipment for extended lengths of time repeatedly close to the 120% to 150% OPP trigger point. This level of extended testing could severely reduce the lifespan of the product and components of the GP-P850GM and GP-P750GM.
Aris response: First of all, there was no prolonged testing period under overloads since most samples died within a few minutes of testing as Steve (Gamer Nexus) mentions in his video. In my case, the GP-P750 GM sample that I tested shut down during a short period of OPP evaluation and exploded once I tried to start it again to continue testing, with no load on its rails. Even if extended testing at high loads was the case, the engineers should use lower OPP triggering points from the moment they are well aware that the platform can handle higher than normal loads. Finally, reducing the lifespan is an entirely different story from exploding parts, which clearly shows a problem with the OPP setting, which GIGABYTE believes is the culprit.
GIGABYTE: GIGABYTE has made adjustments and lowered the OPP on GP-P850GM and GP-P750GM…. from 120% ~ 150% to 110% ~ 120%
Aris response: GIGABYTE noticed, after our findings, that the platforms cannot withstand a 120%-150% OPP rating and decided to lower it. The problem is that 110%-120% OPP is impossible with analog controllers, which use resistors to adjust OPP. These resistors drift with temperature. In other words, their resistance changes according to the operating conditions, so it is impossible to achieve such a small OPP range under both cold and hot conditions. The only way to achieve a tightly set OPP is through digital circuits, an MCU. Finally, GIGABYTE doesn't mention the operating conditions under which the new OPP settings apply.
GIGABYTE: GIGABYTE would like to stress the potential issues that were reported, only seemed to occur after very long time periods of extreme load testing via DC Electronic load equipment and would not be typical of any real world usage.
Aris response: Gamer Nexus' samples died in a matter of minutes, and my sample died after a short period of OPP evaluation. Moreover, a quick look at the user reports (Newegg, forums, etc.) shows that most of these PSUs died under normal conditions. With so many failures reported on Newegg reviews, it cannot just be a coincidence.
GIGABYTE: GIGABYTE GP-P850GM and GP-P750GM PSU's included industry standard power protection designs OCP, OTP, OVP, OPP, UVP, and SCP.
Safety certification from various countries to ensure safe and stable operation of your system.
Aris response: From the moment these units have a CE certification, I would love to check the corresponding CE reports including protection features evaluation. To the best of my knowledge, no safety certification evaluates the PSU's protection features.
GIGABYTE: Despite the fact that both before & after OPP adjustment versions are reliable for real world usage
Aris response: Many users that bought these products have a different opinion and experience. Also, OPP is not there for us reviewers only, but it should protect the PSU under all conditions. Otherwise, there is no point in having this protection feature when it doesn't save the PSU. And also, who and what defines real-world usage? For me, typical use can be having my PC idle most of the time while other users play games most of the time, stressing the entire system. Other users can run tests with Furmark and Prime95 at the same time. My point is, real-world usage varies from user to user.
GIGABYTE: Serial Number below can apply for Return and Exchange service
Aris response: GIGABYTE had made OPP changes to some production batches but didn't inform the people that bought units with high OPP about this or even give them the chance to replace their units, just to be on the safe side. They applied a silent fix, and this means that they were troubled enough to do it.
Conclusion: This is not just a badly set OPP since many units died under moderate loads and within short periods. I strongly believe that this is just a bad design, the FETs are not driven correctly, and although in quick pre-checks and normal conditions, the PSU can be ok, there are cases (not only under stress) where they fail. The timing of the FETs is not correct, and this is due to a lousy gate driver or a lousy implementation. Of course, I cannot be dead sure without any samples, older and newer generations, in my hands to test and break apart for failure analysis.
This article explains the problem I described above in detail.
56 Comments on PSU Expert Aris Mpitziopoulos Responds to Gigabyte Statement on GP-P750/850 GM Design Flaws
-Gigabyte
don't assume that because you see a label with some alphabet (IEEEC/CE ) that it's magically safer. sorry but i have argued this for years now.
as the old saying goes, you can lead a horse to water but you can't make him drink it.
Gigabyte could have owned the problem and worked to make it right. However, this pathetic, shameful reaction destroys trust and brand faith which was already less-than-great.
@ GigaByte
What the actual hell are you thinking?!? Get your damned act together, get your head out of your collective bums and make this situation right with people before you get returned to bottom tier status!
As far as what I saw, the Gigabyte PSU was sold with other brands of graphics cards as a bundle as well, not just Gigabyte cards.
These psus are everywhere, newegg happens to be the one making the worst bundles trying to push it onto consumers.
Microcenter has shelves of these trash psu's but aren't forcing bundles.
Yes, it's a crap PSU, but Gigabyte clearly didn't force these bundles based on your own answer here, as if that was the case, then all retailers would be selling the same bundles.
I think microcenter having them on shelves is a horrible thing given they are bombs waiting to go off. Newegg is just worse by forwarding the bundle forced on them.
The GP--GM series PSU label has UL (which is not required), CCC, TÜV, KCC (which are required in those regions).
Patrick noted that different samples have different parts used for the primary PWM switchers.
I have to wonder if the decision was made to use multiple sources was made initially, or because so many suppliers simply have no parts (I can't get an Infineon MOSFET for an HX's PWM circuit to save my life right now!). If you second and third source after you've already labeled and launched the product, you have to provide samples, pay and wait each time you do so. Thousands of dollars and several weeks. Part of me wonders if the logos on the label even actually represent the product shipping.
Gigabyte, EVGA and probably other GPU vendors that also happen to have PSUs are doing the same thing. They're telling retailers that they can have graphics cards if they also buy PSUs. In some cases it's one to one. The PSUs are often heavily discounted too as they apply Nvidia MDF to the cost of the PSU. Since there's desperation in getting graphics cards, many vendors are taking on excessive PSU inventory to maintain their level of GPU stock.
This isn't an uncommon practice. This is why you'll see a lot of SIs using the same brand GPU and PSU. When I worked for BFG, we played this game too. If Best Buy wanted BFG graphics cards, they'd have to also stock BFG power supplies exclusively. Once BFG lost got screwed over by Nvidia when they started selling it's own branded reference cards directly to Best Buy, Best Buy stopped buying BFG cards, and therefore PSUs. The end result was, of course, BFG went out of business because Best Buy was more than 60% of their sales.IEC-62368-1 (the most current standard) ensures that a PSU is "safe" to the user. Unfortunately, testing to this is not required.... until you get busted and almost (or do) kill someone.
It's not possible to "force" any company to make a bundle like that.
Bundles are usually done in a way that the seller gets a kickback in some form from the manufacturer, it's the only way bundles happen. There's no other incentive that could enable bundles like this.
You're free to dive into your conspiracy theories here, just remember to wear a tinfoil hat.
Also, I doubt this was "unwanted" gear, as I'm sure Gigabyte thought it was a perfectly fine product.The FCC doesn't certify power related products, that's UL, but I guess you knew that. However, FCC is required to sell a device with a radio transmitter in it of any kind in the US market.
CE is required to sell products in Europe, but yes, it's self certification as you point out. However, if you get reported, you'll have to pull the products from the market.
CCC is pointless, that's a PRC certification that you can get with a red envelope.
TÜV can be anything, as you pay them to test what you want to have tested.
KCC is for Korea only.
Also, I thought UL was required these days, but I could be wrong.
I don't think any decision was made, this is what happens when you don't have QC/QA staff on site and manufacture things in the PRC.
I've had similar experiences, both at companies I've worked for and at my own business.
I made a very simple power board for a customer, the company that manufactured it for me couldn't get hold of the right part, so they used a pair of slightly lower rated capacitors and it ended up being a problem, since in certain circumstances the products that were powered didn't get enough power, so they wouldn't start. Do you think the manufacturer told me about the change? :shadedshu:
It caused quite a bit of a headache and I can tell you that I didn't use that manufacturer again.