On Tuesday afternoon, the information embargo on RTX 4070 Ti Super test results at the price recommended by Nvidia, i.e. for non-overclocked models, ended. I was originally going to do a review, we’re going to look at something else instead – there are already three different BIOSes for this model with which to test it, so we’ll discuss how the results in the published reviews differ depending on which BIOS the reviewer is using he measured the card and we will release the test later with only the last one.
One card, three bios
In this article, we’ll try to clear up some of the uncertainty associated with published reviews, look behind the scenes, shed some light on what was actually going on, what the various BIOSes did to the card, allay any concerns, and look at what what is the condition of the Ventus 3X currently.
Similar to the original GeForce RTX 4070 Ti, Nvidia did not prepare its own Founders Edition card for this model either. Only cards from partners went to the reviewers. The MSI RTX 4070 Ti Super Ventus 3X is one of the basic models for which it was possible to publish test results with the end of the embargo, the embargo on the more expensive overclocked versions ends today.
The problem is that reviewers have already found this model to be performing less than expected over the past week and have taken it up with Nvidia and MSI. The measured results were several percent lower than for other similarly clocked models, and in some situations it was possible to measure even worse results than for the RTX 4070 Ti. And it also seems that the performance differences against other models are not consistent, but vary slightly from piece to piece.
At that moment, I put the testing on hold and put another card in the computer, the review of which is still waiting for you. But it is very likely that if you buy an MSI RTX 4070 Ti Super Ventus 3X from the first deliveries after the launch, it will have the original, older or slightly newer BIOS on it (the card goes to reviewers by a different route than to stores).
However, it’s virtually impossible that a card you get in the store now will have a BIOS with performance-enhancing updates from Sunday or Tuesday noon.
MSI has been trying to figure out what the problem is since last week. On Sunday evening, another BIOS arrived, with which the operating clocks of the chip were raised by about 30-60 MHz and the performance of the card increased by units of percent, but the results were still several percent behind other models – even if they were real working clocks of the chip feel the same as other cards. We, who postponed the testing due to the announced BIOS update, started testing on Monday to at least somehow manage it, someone probably just gave up saying that the difference will not be big in the end.
And in the end, it doesn’t even matter that much anymore, because two hours before the end of the embargo, another e-mail landed in our e-mail boxes with an apology and a BIOS that increases the performance a bit again. In other words, the MSI RTX 4070 Ti Super Ventus 3X test results you can find on the Internet are different from what you get if you load the card with the latest firmware (and after the few tests I’ve done on it, I wouldn’t be surprised if even this BIOS was not the last).
Same parameters, different behavior and performance
So the RTX 4070 Ti Super Ventus 3X received two new BIOSes within three days, each increasing (usually) performance by some 1-3%. And I’m a little afraid that it’s not final either, because even with the current BIOS, the performance is still a bit lower than the other card I’ve had here (and now it has lower chip clocks).
If you get your hands on an RTX 4070 Ti Super Ventus 3X, the easiest way to find out which BIOS version it has loaded is with GPU-Z:
It contains the following BIOS versions on the main tab in the BIOS column:
- 95.03.45.40.58 in the original BIOS
- 95.03.45.40.DC for BIOS compiled on January 20 (shipped on January 21)
- 95.03.45.40.F0 for BIOS dated January 23rd
And if you don’t want to study the annoying numbering, just look at the Advanced tab and click on the NVIDIA BIOS item in the drop-down menu – there you will see the date when the BIOS of the card was compiled. The old version had a build date of December 10, 2023, the second Sunday version of January 20, 2024, and the latest yesterday, January 23, 2024.
With yesterday’s BIOS, the clock/voltage curve that controls the clocks for automatic boost has probably changed slightly. I didn’t take a screenshot with the original BIOS, that would be good, I only thought before flashing between the first and second repair BIOS. The essential difference is at its end. It is in the order of low tens of megahertz – on the left is the curve from the BIOS sent out on Sunday, on the right from the one from Tuesday.
This is reflected in the performance when the consumption of the card is so low that the card does not reach the power limit. In that case, this curve will limit the clock rate a little lower than with the previous BIOS, and with the new BIOS the performance is a little lower than with Sunday’s – you can see it in the graphs from F1, where the chip clock rate is capped a little lower.
Paper parameters do not make performance
An unpleasant thing is what accompanies all graphics since Nvidia came up with automatic clocking – although on paper the card looks like the parameters such as clocks or power have not changed in any way, in reality the behavior of the card has changed. To a certain extent, the manufacturer can set what and how he wants and thereby shift the performance of the card and its consumption, without you having a chance to find out from the data in the BIOS that the card settings have changed in any way. In other words, the same card on paper behaves slightly differently in all three cases.
I’ve tested pretty much everything from the normal methodology on Sunday’s BIOS, and I’ve been debating whether to release a review, to publish a retest in a few days with different results, where the performance is different by a few percent, or to screw it up and not do it even more chaos than it is and release a review with the results that the card gives with the BIOS last.
I decided to give up on publishing a test with the results on the BIOS, which probably will never be made public, and I will measure the performance again only with the public BIOS, so that I don’t make more confusion and the measurement in the test corresponds to what you will be able to buy .
In the Ventus 3X reviews, keep an eye on what it was tested on
Why I bother – at this point it’s likely that the MSI GeForce RTX 4070 Ti Ventus 3X reviews published on Tuesday or a few hours later may have different performance and behavior depending on whether the reviewer tested the card with the original BIOS with lower performance, or was still retesting it with the BIOS updated on Sunday, or not. And the next shift awaits us with patch tests with the BIOS released de facto with the end of the information embargo.
Will you have Ventus 3X from the first deliveries? Flash it. Or not?
Not a critical issue purely from a performance perspective, the usual measured differences are somewhere between 3-10fps at common frame rates of 40-200fps, something you have no chance of knowing unless you have a second card to compare against performance on the same assembly. On the other hand, a four percent difference in performance is something for which there are usually four-hundred-post cuts in discussions about which card is better and more profitable, and also something for which factory overclocked models are paid a few hundred to a thousand more.
It is likely that because of those units of percent in performance, MSI will not stop sales and recall the cards for reflashing the BIOS, but they will go on sale normally and MSI will try to notify the owner of the firmware update.
I’m not saying it will, but it’s an expected scenario – MSI hasn’t revealed anything more about the matter so far, other than that they’ve “discovered more ways to improve the performance of RTX 4070 Ti Super Ventus 3X cards” and that they’re “releasing a new BIOS, with by which the card performs as expected.”
In this article, we’ll just take a quick look at how all three BIOSes differ in terms of performance, as that’s perhaps more interesting than the performance differences themselves.
The card behaves a bit non-standard with all three BIOSes – according to monitoring (e.g. in GPU-Z), the card’s consumption is lower than that of another 285W RTX 4070 Ti Super that I have tested.
On the socket wattmeter, however, the set with both cards has practically the same power consumption. Or, to be more precise, it should have – with the latest BIOS, the power consumption of the assembly with Ventus 3X is already slightly higher than that of a card from another manufacturer, despite the fact that, according to the diagnostics, it appears to consume 14 W less.
In short, it seems that MSI is trying to squeeze out some extra performance by tinkering with the boost regulation settings so that it achieves an increase in performance and does not affect the card’s other characteristics such as consumption, temperatures or noise.
2024-01-24 08:56:06
#hours #MSI #GeForce #RTX #Super #Ventus #tests #bad #firmware #boosts #performance
