News AMD Launches Radeon RX 5600 XT: Ultimate 1080p Gaming GPU at $279

[Admin]

The GeForce GTX 1660 Ti has met its match.

AMD Launches Radeon RX 5600 XT: Ultimate 1080p Gaming GPU at $279 : Read more

 

[alextheblue]

The Radeon RX 5600 XT holds its own agaisnt some of Nvidia's more recent offerings. The Radeon RX 5600 XT provides up to 20% and 10% higher performance than the GeForce GTX 1660 Ti in AAA and eSports titles, respectively.

Whem compared to the GeForce GTX 1660 Super OC, the Radeon RX 5600 XT still manages to come out on top. It's 15% and 5% faster than the GeForce GTX 1660 Super OC in AAA and eSports games, respectively.

This seems backwards to me. The 1660 Ti should edge out the 1660 Super, and thus the 5600 XT should be edging out the Ti by the smaller margin.

Anyway that's not a bad price. The 1660 Ti typically starts at $279, with many cards being closer to $300. Meanwhile the 1660 Super, while a bit slower, still represents a pretty darn good value.

 

[cryoburner]

The Radeon RX 5600 XT will be available on January 21 for $279, which is on par with the GeForce GTX 1660 Ti's pricing.

The 1660 SUPER starts at just $230 though, while performing within about 5% of a 1660 Ti, making it a better card to compare against. That makes the 5600 XT priced around 20% higher, for what sounds like maybe 15% more performance on average in demanding games. And it also remains to be seen how performance will compare against the next cards up, the 2060 SUPER and RX 5700, which can at times be found on sale for not a whole lot more than $300. I'm not exactly convinced that the pricing is all that attractive.

This seems backwards to me. The 1660 Ti should edge out the 1660 Super, and thus the 5600 XT should be edging out the Ti by the smaller margin.

While definitely an odd set of graphs, apparently those tests were performed on two different test systems. The comparison versus the 1660 Ti was performed on a Ryzen 3800X test system, while the comparison versus the 1660 SUPER was performed on an i9-9900K test system. Which begs the question, why did the system with the slightly faster gaming CPU create results that were closer to one another, rather than farther apart, as one might expect?

One possible explanation I can think of is that we may be seeing a repeat of the 5500 XT scenario. It may not be the 9900K holding the 5600 XT back, but rather PCIe 3.0, or more precisely, I suspect AMD might have once again gimped the card on PCIe 3.0 systems by giving it an x8 connection, which seems completely unacceptable for a nearly $300 graphics card. If that's true, then the performance on anything but an x570 motherboard might look more like the 1660 SUPER results, and as a result, the 5600 XT might only be around 10% faster than a 1660 Ti in demanding titles on most systems.

Or maybe it's simply a case of driver optimization, or someone messed up the graphs, or something else. It might be too early to jump to conclusions, but given that the 5500 XT sees similar performance anomalies due to its limited PCIe bandwidth on most systems, I wouldn't be surprised if that were the case.

 

[King_V]

I'm not sure if the PCIe 3.0 issue is it - but hard to say, and it would definitely be a mistake if AMD did this. Still, I suspect the performance would be worse, if that were the case. Don't hold me to that, though. EDIT: as was done for the 5500XT, I would definitely like to see that analyzed, though. I also wonder if it would gimp a 6GB card the way it gimped the 8GB 5500XT (relative to the 4GB version).

As for pricing, etc. The 1660 Super is, for all intents and purposes, about the same as the 1660Ti. Nothing about the Ti justifies paying that extra money over the Super.

So, saying that, I think the 5600XT is going to have to come down a little. At least, that would seem to make sense for me - edging the 1660 Super and Ti out both in performance (more), and in price (less), if AMD wants to hold on to owning the mainstream tier. Hopefully the prices dip down below the release price quickly.


On another note, I'm also really enthused about the idea that, while it's rated at a 150W TDP, it hits 120W in most gaming scenarios. If that's true, we are now officially looking at AMD matching (or slightly outdoing) Nvidia in terms of performance/power-draw ratio. After all, from the 1660 Super review:

Starting with the gaming power results, the EVGA 1660 Super SC Ultra averaged 123W during the Metro: Exodus test. The slightly faster GTX 1660 Ti SC Ultra used 128W, which fits in well with the expected power use.

 

[cryoburner]

I'm not sure if the PCIe 3.0 issue is it - but hard to say, and it would definitely be a mistake if AMD did this. Still, I suspect the performance would be worse, if that were the case.

In the case of the 5500XT, the limited PCIe bandwidth only tends to cause major performance hits in some games with the 4GB version, since the VRAM can fill, at which point the card needs to access data from system memory over the PCIe bus. The 8GB version generally sees much better performance in those titles, since it has VRAM to spare, though smaller performance hits can still occur. If the 5600XT does the same, I suspect the performance hit on PCIe 3.0 might similarly be relatively small in most current titles, since 6GB of VRAM should still be enough for 1080p gaming at ultra settings. However, the card is notably faster than a 5500XT, which might potentially exaggerate any differences that do appear. Plus, while the VRAM may be adequate today, games coming out a couple years from now will likely want more, at which point the card could potentially fall behind. Again, that's assuming they went with an x8 interface, which might not actually be the case. We should know in a couple weeks. : P

On another note, I'm also really enthused about the idea that, while it's rated at a 150W TDP, it hits 120W in most gaming scenarios. If that's true, we are now officially looking at AMD matching (or slightly outdoing) Nvidia in terms of performance/power-draw ratio.

One thing helping efficiency here is that the clock rates are being kept relatively low, while the graphics core count remains the same as an RX 5700. The RX 5500XT, by comparison, has about 40% fewer cores, but the clock rates are cranked up higher than those of a 5700, to around 5700XT levels. So it wouldn't surprise me if the 5600XT's power draw isn't all that much different from that of the 5500XT.

 

[JonDol]

ROFL: 10-20 % better performance than GTX 1660 Ti while eating 25% more energy (120 W vs 150 W) ? Performance per watt is worse so thank you but not, I'll still stick to the greener card. One could always overclock the GTX to get slightly similar performance and power consumption...

 

[King_V]

ROFL: 10-20 % better performance than GTX 1660 Ti while eating 25% more energy (120 W vs 150 W) ? Performance per watt is worse so thank you but not, I'll still stick to the greener card. One could always overclock the GTX to get slightly similar performance and power consumption...

It appears that reading isn't your strong suit. If I may re-quote the bit from the 1660 Super review:

Starting with the gaming power results, the EVGA 1660 Super SC Ultra averaged 123W during the Metro: Exodus test. The slightly faster GTX 1660 Ti SC Ultra used 128W, which fits in well with the expected power use.

Now, again, because reading the article is important, from the very same article that this thread linked to in the original post:

The Radeon RX 5600 XT comes with a 150W Total Board Power (TBP) rating. However, the graphics card hits 120W in most gaming scenarios.

Please think carefully before you start spouting outright nonsense that is explicitly contradicted by documented facts.

 

[alextheblue]

I suspect AMD might have once again gimped the card on PCIe 3.0 systems by giving it an x8 connection, which seems completely unacceptable for a nearly $300 graphics card.

I'm quite positive all Navi 10 units have a 16 lane connection. Meanwhile the smaller Navi 14 only has 8 lanes to begin with.

Or maybe it's simply a case of driver optimization, or someone messed up the graphs, or something else.

Those first two do seem like the most likely culprits.

 

[joeblowsmynose]

I'm quite positive all Navi 10 units have a 16 lane connection. Meanwhile the smaller Navi 14 only has 8 lanes to begin with.
Those first two do seem like the most likely culprits.

I don't see these graphs you guys are talking about ... but system ram speed can make a decent difference with Ryzen in gaming. Just a thought; I always try to rule out "nefariousness" first, when trying to conclude on a consideration.

 

[TJ Hooker]

I don't see these graphs you guys are talking about

About half way through the article there's an image: it's actually a set of 4 images. If you go through them it includes some graphs comparing the 5600 XT to a 1060 6GB, a 1660 Super, and a 1660 Ti in 1080p gaming. It says the 5600 XT is 20% better than a 1660 Ti and 15% better than a 1660 Super, which doesn't make sense given that the 1660 Ti should be a bit better than the Super.

 

[joeblowsmynose]

About half way through the article there's an image: it's actually a set of 4 images. If you go through them it includes some graphs comparing the 5600 XT to a 1060 6GB, a 1660 Super, and a 1660 Ti in 1080p gaming. It says the 5600 XT is 20% better than a 1660 Ti and 15% better than a 1660 Super, which doesn't make sense given that the 1660 Ti should be a bit better than the Super.

Ok I did see those but I was looking for the part that indicated they were testing on different CPUs ... or was there just a whole lot of speculation about why it might be the case? Sorry I thought that there was confirmation on the two different platforms used. Maybe I need to read posts more carefully.

 

[TJ Hooker]

Oops, yeah I have no idea where the info on there being two different systems came from.

 

[alextheblue]

I don't see these graphs you guys are talking about ... but system ram speed can make a decent difference with Ryzen in gaming. Just a thought; I always try to rule out "nefariousness" first, when trying to conclude on a consideration.

I never meant to suggest or imply it was nefarious. Rather, that someone made a mistake, or that there were differences in software or hardware between test systems that weren't factored in - such as a newer driver revision. The 1660 Ti is getting somewhat outpaced by the same-price-or-less 5600 XT. The cheaper Super should be getting outpaced by a slightly larger margin. That's pretty straightforward.

I think at ~$230 or so the Super still represents the best value if you can't step up to the ~$280+ range, but you have enough cash to get something above the ~$180 range (and under) cards.

Side note: I really want to see a 5600 XT overclocked to 5700 speeds, to see how much the narrower bus holds it back (or doesn't). I'm sure consumption will rise accordingly, so if the memory bandwidth holds it back it may not be worth it... but I am still curious.

 

[cryoburner]

Ok I did see those but I was looking for the part that indicated they were testing on different CPUs ... or was there just a whole lot of speculation about why it might be the case? Sorry I thought that there was confirmation on the two different platforms used. Maybe I need to read posts more carefully.

I don't think it was shown in this article, but there was a slide listing details of the test systems used to acquire the benchmark data used for the presentation. HardwareUnboxed mentioned the different systems used in their video, and you can see the relevant slide there, though they didn't really speculate on how exactly that would have affected the performance numbers. Skip to the 13 minute mark if you want the relevant section discussing performance relative to the 1660 SUPER and Ti...

View: https://www.youtube.com/watch?v=wEo4VuvzVks

I'm quite positive all Navi 10 units have a 16 lane connection. Meanwhile the smaller Navi 14 only has 8 lanes to begin with.

Has that actually been confirmed anywhere though? Sure, the chip supports it, but it's possible that AMD could have not used all traces in an effort to further differentiate the product stack, and maybe attempt to make PCIe 4.0 look better. After all, aside from the underclock, the main thing separating a 5600 XT's performance from that of a 5700 in today's games is going to be it's 192-bit memory interface on a GPU that would otherwise support 256-bits, cutting VRAM performance by 25% even before you get to the lower memory clocks. Sure, that may be related to the reduced VRAM compared to a 5700, but since they put up to 8GB on even the 5500 XT, it seems like reducing memory performance was their primary goal there. There's the possibility that they could have done the same to PCIe lanes as well.

As far as the card's value is concerned, there are also actually hints that Nvidia might be giving the RTX 2060 a price cut soon. Namely, there was a new card being shown off at CES (I believe from EVGA) with a $299 list price, which would probably be indicative of that. Considering Nvidia has launched SUPER cards around each of AMD's product launches in an attempt to undercut the competition, it seems likely that they will do the same for the 5600 XT as well. A $280 5600 XT might be priced "alright" relative to a $280 1660 Ti or a $350 2060, but those are year-old cards at this point. The value might not look so great against a 2060 for just $20 more.

 

[alextheblue]

Has that actually been confirmed anywhere though?

There's no official word, but yeah everything I'd heard seems to indicate 16 lanes. Again the only reason the 5500 doesn't have those 8 lanes isn't because they're disabled... they don't exist. If you can think of a case where AMD disabled working PCIe lanes in a GPU, let me know. I can't recall one.

The massively reduced memory bandwidth (clock and bus width) are more than enough to differentiate it from the 5700. Even with the obvious overclocking headroom, I don't think the 5600 series will crowd its 256-bit bus stablemate.

As far as the card's value is concerned, there are also actually hints that Nvidia might be giving the RTX 2060 a price cut soon.

Yep, good to have competition. Wonder if they'll cut prices on the 1660 Ti as well, or drop it entirely. I'm curious to see if they release the rumored 5600 non-XT with slightly reduced CU count. That one could be a great value, and probably a good overclocker.