AMD 7900XT and 7900 XTX Reviews

[erek]

Here's some more "updates" on the thermal controversies / silicon revisions

 

[Gulkor]

I went to AMD.com just a bit ago and was placed in queue for an XTX so they might still be available for someone that's interested.

I thought about it but nah. $1000 to play games that aren't even limited by my 2080ti is just dumb in my case. I'd be better off buying a 5800x3d or adding some more RGB.

I was thinking getting rtx 4080 or rx 7900 xtx, i found evga rtx 3080 $525 shipped. games i play i wont have problem getting this card last 3-4 years. im on 27" 1440p 165hz

 

[Bankie]

I was thinking getting rtx 4080 or rx 7900 xtx, i found evga rtx 3080 $525 shipped. games i play i wont have problem getting this card last 3-4 years. im on 27" 1440p 165hz

Yeah I have some friends with spare 3090s that would sell one to me but I'm not dropping $800-900; at that point I might as well just bite the bullet and grab a 7900xtx.

 

[Gulkor]

Yeah I have some friends with spare 3090s that would sell one to me but I'm not dropping $800-900; at that point I might as well just bite the bullet and grab a 7900xtx.

Now i might do is wait for rtx 5000 series come out, and picked up good priced used rtx 4080. i found out long time ago, id rather save extra money and buy used parts on forums here.

hell i picked up asus tuf z690 wifi + i5 12600k for $325 shipped.

I love make my $1 work for me, i try make my dollar go twice as far

 

[Schro]

Schro said this:

Use of an AMD platform for testing (with Smart Access Memory enabled) makes a fairly significant performance difference - north of 10% in some games and especially at lower resolutions. Thus, if you're comparing a reviewer's numbers with an Intel platform vs AMD, you're going to see wildly different results.

I quoted them and asked them to "prove it" (prove that SAM on an AMD platform, performs better than REBAR on an Intel platform, with an AMD GPU).

So far, I haven't seen any evidence showing that. The article you linked, does not show that with numbers. Even though they do otherwise seem to say that.

When searching on this topic, there's a LOT of inaccurate information that surfaces about the technology that has been written over the past few years.

This response has taken longer than it should. Here's what I have so far -

1. I met with AMD's Radeon team at CES today and articulated the question clearly and received the following - SAM is proprietary to AMD hardware and operates on top of REBAR. Therefore, if you put the AMD card in an Intel system, you will get _some_ performance increase between REBAR on/off, however, it will not be as much as compared to SAM.
2. I asked if there was a way to test this on an AMD system (so we can remove the Intel variable from the comparison) so that we could test 1. No REBAR/SAM vs 2. REBAR only vs 3. SAM. Right now, the steps to enable SAM is to basically turn on REBAR, therefore, that test is not possible (though, they have done it in their labs).

They should be getting back to me soon with what I hope is a way that I can directly test the 3 scenarios above - I'll keep following up on it.

At the beginning of the article, they ask the question Are reviewers using Intel systems, hurting Radeon Radeon RX 7900 Series GPU performance?
But, they do not answer that question, with numbers.

Then in the conclusion, they imply a few times, that using an AMD plaform, has an advantage over using an Intel platform, for resizeable BAR, with a 7900 series GPU. But they do not show any proof. Here is one example:
If you are on an Intel platform, you won’t get AMD Smart Access Memory, but you can still enable Resizable BAR on them as well, so at least do that for the best possible experience.

It's a very difficult question to answer with numbers as changing from an AMD system to an Intel system itself is a variable in performance, thus, even if tested directly doesn't really provide much in the way of answers. The point is - AMD states that the best performance for the card is using SAM and in order to meet all of the requirements for SAM, you must also use a compatible AMD rig. Therefore, reviewers using Intel rigs to test the RX 7900 series cards have potentially limited the GPU performance due to their test bench hardware selection (which, to be fair, as long as it's disclosed and the pros/cons are discussed for such a decision is completely legit).

Now, it's entirely possible if we ever get to run the 3 test scenario using identical hardware that we find that SAM performance is actually 90% REBAR and 10% SAM - it's just something we do not currently know.

 

[chameleoneel]

Schro said this:


When searching on this topic, there's a LOT of inaccurate information that surfaces about the technology that has been written over the past few years.

This response has taken longer than it should. Here's what I have so far -

1. I met with AMD's Radeon team at CES today and articulated the question clearly and received the following - SAM is proprietary to AMD hardware and operates on top of REBAR. Therefore, if you put the AMD card in an Intel system, you will get _some_ performance increase between REBAR on/off, however, it will not be as much as compared to SAM.
2. I asked if there was a way to test this on an AMD system (so we can remove the Intel variable from the comparison) so that we could test 1. No REBAR/SAM vs 2. REBAR only vs 3. SAM. Right now, the steps to enable SAM is to basically turn on REBAR, therefore, that test is not possible (though, they have done it in their labs).

They should be getting back to me soon with what I hope is a way that I can directly test the 3 scenarios above - I'll keep following up on it.



It's a very difficult question to answer with numbers as changing from an AMD system to an Intel system itself is a variable in performance, thus, even if tested directly doesn't really provide much in the way of answers. The point is - AMD states that the best performance for the card is using SAM and in order to meet all of the requirements for SAM, you must also use a compatible AMD rig. Therefore, reviewers using Intel rigs to test the RX 7900 series cards have potentially limited the GPU performance due to their test bench hardware selection (which, to be fair, as long as it's disclosed and the pros/cons are discussed for such a decision is completely legit).

Now, it's entirely possible if we ever get to run the 3 test scenario using identical hardware that we find that SAM performance is actually 90% REBAR and 10% SAM - it's just something we do not currently know.

I don't doubt the possibility that AMD has some optimizations they do for their infinity fabric, inifinity cache, etc.
However, some Tech reviewers such as Hardware Unboxed have looked at SAM/REBAR performance with AMD cards on Zen 2, Zen 3, and Intel Z690 with a 12900k. And the performance boost on Intel was essentially the same, despite whatever AMD is eluding to.

And at the time of testing, their Zen 2 system didn't even have very good support for SAM. I already posted this video:


Its possible that AM5 and 7900 GPUs have even better SAM optimizations, which might actually eclipse REBAR on an Intel platform. Hopefully HU will do an updated test of Zen 4 SAM compared to Raptor Lake Rebar.

 

[Schro]

I don't doubt the possibility that AMD has some optimizations they do for their infinity fabric, inifinity cache, etc.
However, some Tech reviewers such as Hardware Unboxed have looked at SAM/REBAR performance with AMD cards on Zen 2, Zen 3, and Intel Z690 with a 12900k. And the performance boost on Intel was essentially the same, despite whatever AMD is eluding to.

And at the time of testing, their Zen 2 system didn't even have very good support for SAM. I already posted this video:


Its possible that AM5 and 7900 GPUs have even better SAM optimizations, which might actually eclipse REBAR on an Intel platform. Hopefully HU will do an updated test of Zen 4 SAM compared to Raptor Lake Rebar.

I suppose I have a couple of issues with this -

1. In the introduction to the video, he states that REBAR == SAM. AMD says they are different. Only one of the parties can be correct.
2. Testing in this video is not specific to the 7900 series GPUs, thus, not necessarily relevant to the topic at hand, especially if the 7900 series architecture is more reliant on SAM than the 6000 series.
3. Differences in gaming performance between Intel/AMD chips can influence the results, thus, making it difficult to draw a certain conclusion on the impact of the specific technology.

 

[Meeho]

It's a very difficult question to answer with numbers as changing from an AMD system to an Intel system itself is a variable in performance, thus, even if tested directly doesn't really provide much in the way of answers.

Not that difficult. Performance gain on/off for cards on different platforms, and relative performance difference comparison of AMD and Nvidia between platforms.

 

[chameleoneel]

I suppose I have a couple of issues with this -

1. In the introduction to the video, he states that REBAR == SAM. AMD says they are different. Only one of the parties can be correct.
2. Testing in this video is not specific to the 7900 series GPUs, thus, not necessarily relevant to the topic at hand, especially if the 7900 series architecture is more reliant on SAM than the 6000 series.
3. Differences in gaming performance between Intel/AMD chips can influence the results, thus, making it difficult to draw a certain conclusion on the impact of the specific technology.

Uh huh. But that video is a lot better than blindly saying stuff.

Indeed, Intel and AMD CPUs can have different performance in general, depending upon the game. But, if you compare Zen 3 to itself (SAM Vs. no SAM) and Intel 12th gen to itself (REBAR V.s no REBAR), the overall benefit is more/less the same. Which suggests that with RDNA2 and Zen 3------whatever SAM specific optmizations may be there: do not equal strictly better performance than using RDNA 2 on intel 12th gen.

And about your #2 there----well I already said this:

Its possible that AM5 and 7900 GPUs have even better SAM optimizations, which might actually eclipse REBAR on an Intel platform. Hopefully HU will do an updated test of Zen 4 SAM compared to Raptor Lake Rebar.

 

[mkppo]

Great article over at Chips & Cheese, the dual issue part corresponds with what I mentioned earlier. It'll be interesting to see the performance of the XTX with newer games.

7900 XTX deep dive