AMD Navi Product Reviews and Previews: (5500, 5600 XT, 5700, 5700 XT)

I think Steve is simply wrong here at least in the start of the video (didn't watch whole 30min thing), talking about big gap between different 5600 XT's. The gap between different 5600 XT's is still smaller than what we're used to seeing. Probably the biggest ones were in the Maxwell-range, where certain AIB-models were 20%+ faster than reference. No-one thought that difference would be a problem.
 
I think Steve is simply wrong here at least in the start of the video (didn't watch whole 30min thing), talking about big gap between different 5600 XT's.

Watch the whole video and your opinion should change as there will be up to 12% performance differences between 5600 XT's throughout the life cycle of this card.
 
Watch the whole video and your opinion should change as there will be up to 12% performance differences between 5600 XT's throughout the life cycle of this card.
Maxwell isn't that ancient history, and there were 980 Ti's that were about 20% faster than reference without even breaking the bank (Zotac Amp Extreme), no-one complained about the big gap between different 980 Ti's.
 
Maxwell isn't that ancient history, and there were 980 Ti's that were about 20% faster than reference without even breaking the bank (Zotac Amp Extreme), no-one complained about the big gap between different 980 Ti's.
Difference is:
-No 980Ti was sold below advertised performance or reviewed performance, you get the performance advertised or more, unlike the 5600XT, where reference models provide sub par performance, and where many AIB models also can't handle the new VBIOS, and thus also provide sub par performance ..

-No 980Ti required the user to manually apply a VBIOS and risk bricking the card if he is not experienced enough, unlike the 5600XT ..

-No 980Ti suddenly got it's clocks increased to new "untested" levels, thus putting the GPU in danger of being damaged and requiring RMA, unlike the 5600XT ..
 
Difference is:
-No 980Ti was sold below advertised performance or reviewed performance, you get the performance advertised or more, unlike the 5600XT, where reference models provide sub par performance, and where many AIB models also can't handle the new VBIOS, and thus also provide sub par performance ..
Neither has 5600 XT, these are all factory OC'd models, the OC limit is just higher than in original plans. AIB-partners like to send their fastest available models to reviewers, the review situation has repeated itself every single time there's a new model on market without AMD/NVIDIA reference model in sight - fastest AIB models get reviewed, most cards you can buy are slower. If you buy the reviewed card, you get what you see.
-No 980Ti required the user to manually apply a VBIOS and risk bricking the card if he is not experienced enough, unlike the 5600XT ..
Yes, this is a real thing for the first batches of cards but the problem will clear out quickly as new cards start shipping with pre-installed higher OC BIOS.
-No 980Ti suddenly got it's clocks increased to new "untested" levels, thus putting the GPU in danger of being damaged and requiring RMA, unlike the 5600XT ..
Indeed, "untested" rather than untested, since the same chips are known to run at much higher frequencies in 5700 series and the memory chips are all validated at 14 Gbps by memory manufacturer (even in those Asus & MSI cases where they released higher OC'ing BIOS with lower 12 Gbps memory clocks for specific models the chips themselves are 14 Gbps)
 
Last edited:
The biggest difference is that 980Ti was a top end card and making it faster meant higher priced and faster cards which weren't competing with anything above them. Well, apart from Titans but that's their destiny.
 
The biggest difference is that 980Ti was a top end card and making it faster meant higher priced and faster cards which weren't competing with anything above them. Well, apart from Titans but that's their destiny.
Fastest 5600 XT is Asus Strix TOP and 5700 reference beats it handily with 7 - 9 % gap depending on resolution.
 
+8% in performance for +25% in price.
Asus has actually priced their 5600 more expensive than the cheapest 5700s (and not just reference models, but custom AIB models too)
(their pricing on AMD cards is ridiculous, not sure if same applies to their NVIDIA cards)
 
Neither has 5600 XT, these are all factory OC'd models, the OC limit is just higher than in original plans. AIB-partners like to send their fastest available models to reviewers, the review situation has repeated itself every single time there's a new model on market without AMD/NVIDIA reference model in sight - fastest AIB models get reviewed, most cards you can buy are slower. If you buy the reviewed card, you get what you see.
And once again underscoring the problems with the GPU makers supplying factory OC cards. If AMD had sent out a version of the Pulse programmed to run at AMD reference clocks out of the box, there wouldn't have been any ambiguity. AMD could have done something about this in advance; they didn't.

In the meantime, if I may get a little self-righteous here, this is why undoing factory overclocks (be it via BIOS hacks or underclocking) is so important for these sorts of reviews. The only specs anyone should have used for measuring the performance of a generic 5600 XT are reference specs. Factory overclocked cards should be treated as separate products.
 
And once again underscoring the problems with the GPU makers supplying factory OC cards. If AMD had sent out a version of the Pulse programmed to run at AMD reference clocks out of the box, there wouldn't have been any ambiguity. AMD could have done something about this in advance; they didn't.

In the meantime, if I may get a little self-righteous here, this is why undoing factory overclocks (be it via BIOS hacks or underclocking) is so important for these sorts of reviews. The only specs anyone should have used for measuring the performance of a generic 5600 XT are reference specs. Factory overclocked cards should be treated as separate products.
Valid points, but not the first nor last time this happens. I don't remember many raising the same points on Turing launch though, where even NVIDIAs own highend FE's were factory-OC'd.
 
Valid points, but not the first nor last time this happens. I don't remember many raising the same points on Turing launch though, where even NVIDIAs own highend FE's were factory-OC'd.
Oh I think we're in full agreement. I'm just exasperated that my prediction from last week is coming true. And while I make it a policy to avoid punching down, I do blame the press for it as much as I blame AMD.

Ultimately there's definitely a lot more public attention this time around, that's for sure. Having such significantly overclocked cards sampled to the press doesn't help matters. But if that gets AMD/NV to stop sending factory overclocked cards (even if only for a time), that would be nice.
 
Doesn't this trends makes the amd oc become the new standard clock?

As 3rd party boards OC will be even higher from the new standard over clocked clock from AMD
 
Doesn't this trends makes the amd oc become the new standard clock?

As 3rd party boards OC will be even higher from the new standard over clocked clock from AMD
No. AMD never changed the 5600 XT reference spec. They only allowed factory OC'd models to be overclocked higher than originally planned (including memory 'overclock' too), but they were factory OC'd before the BIOS update too.
Currently only one card (that I'm aware of) is known to have reference clocks (MSI Mech), every single other one is factory OC'd with or without BIOS update. Highest Game/Boost-clocks are 1670 / 1750 MHz available on Asrock Phantom Gaming D3 and Asus Strix with 12 Gbps memory and on Asus Strix TOP with 14 Gbps memory.
 
Neither has 5600 XT, these are all factory OC'd models, the OC limit is just higher than in original plans. AIB-partners like to send their fastest available models to reviewers, the review situation has repeated itself every single time there's a new model on market without AMD/NVIDIA reference model in sight - fastest AIB models get reviewed, most cards you can buy are slower. If you buy the reviewed card, you get what you see.
Again, the 980Ti reviews didn't come with the 20% OC'ed 980Ti models, they came with the reference clocks, even the 2060 reviews came with reference clocks, Unlike the 5600XT.

Indeed, "untested" rather than untested, since the same chips are known to run at much higher frequencies in 5700 series and the memory chips are all validated at 14 Gbps by memory manufacturer (even in those Asus & MSI cases where they released higher OC'ing BIOS with lower 12 Gbps memory clocks for specific models the chips themselves are 14 Gbps)
Several AIBs have spoken against the memory being rated at 14Gbps, stating it's a target that is of questionable attainability for the current 5600XT models.

Yes, this is a real thing for the first batches of cards but the problem will clear out quickly as new cards start shipping with pre-installed higher OC BIOS.
Or it may take much longer to clear, considering the holidays that left many AIBs factories nonoperational.
 
Again, the 980Ti reviews didn't come with the 20% OC'ed 980Ti models, they came with the reference clocks, even the 2060 reviews came with reference clocks, Unlike the 5600XT.
980 Ti's didn't, but RTX 2080 and 2080 Ti did come to reviews with OC'd models. And so has every single other model out there from which AMD/NVIDIA didn't launch separate reference version.

Several AIBs have spoken against the memory being rated at 14Gbps, stating it's a target that is of questionable attainability for the current 5600XT models.
AFAIK So far every single card taken apart has had the very same 14 Gbps memory, even the MSI rep on their own video corrected the other guy who said something about not all cards having the chips to do 14 Gbps that every single 5600 XT does have 14 Gbps memory but apparently AMD hasn't separately validated them as such (even when the memory manufacturer has)
 
MSI Announce That Not All RX 5600 XTs Will Use New vBIOS
January 28, 2020
Announced during the most recent episode of the MSI Insider podcast, MSI revealed that not every AMD Radeon RX 5600 XT will use new vBIOS.

Eric Van Beurden, MSI’s marketing director, explained that the new vBIOS would increase a card’s clock speed. However, the memory increase will be a separate issue entirely. This means that some companies are boosting memory independently, but this won’t necessarily apply across the board as standard.
...
Some of the cards will be able to boost the memory, but some aren’t certified to work at frequencies of 14 Gbps, meaning they will lose the stability they have when they’re operating at a stock speed of 12 Gbps.

The reason for this was further explained by Pieter Arts, product marketing manager at MSI. He said that AMD supplies partners with silicon and memory chips, but there’s no guarantee that all of them will have the same overclocking abilities to run at the higher 14 Gbps.

This could be the reason behind MSI’s and Asus’ hesitancy to apply vBIOS to existing cards because there’s no guarantee it will be worth it, just to boost the clock speed.
....
It’s possible that many cards will support 14 Gbps, but there’s a risk attached to applying the new vBIOS to them.

Doing so could lead to some RMA problems for users, so it all depends on whether you want to risk it to achieve 14 Gbps or whether you’re happy with the standard 12 Gbps.

MSI is promising that most cards will receive the new BIOS by February, but in most cases, the memory will remain at 12 Gbps, so the improvement will be to the processor frequency only.

https://www.wepc.com/news/msi-rx-5600-xt-new-vbios/
 
Last edited:
Back
Top