RX480 re-badged as RX 580 in Dell Inspiron 5675

imrazor

Distinguished
I recently found a deal on a Dell Inspiron Gaming 5675. To avoid the hassle and delays inherent in building your own system, I thought I'd pick it up and save some time and money. I still got a good deal, but there are a few gotchas.

When I first got the PC I set about testing it and was puzzled by some of the results and benchmarks. I know the 580 is generally superior to my old GTX 970, but I was surprised when the 970 (admittedly overclocked) was tying or beating the 580 in benchmarks and games.

The system was advertised as having an AMD Radeon RX 580. However, this 580 is clocked at 480 speeds and is even identified as a 480 by 3DMark Fire Strike. The BIOS identifies the card as a 580, but the card only has a 75 watt 6-pin PCIe slot, whereas most 580s I'm familiar with require 150 watts (in addition to 75w from the motherboard.) Apparently this also affects the RX 570s sold with this system, see here: https://www.dell.com/community/Desktops-General/DELL-Inspirion-5675-Radeon-RX-570-or-RX-470/m-p/5187371#M1032138 .

From reading the posts on the Dell forums, it seems that Dell can get away with this because a) the RX 480 and 580 essentially have the same GPU, and a "real 580" is just clocked higher and b) AMD just sets generally guidelines for clock speed and power consumption. OEMs can implement whatever clock speed they want to, even if it's no faster than the older generation. Still seems shady to me...

There are also a couple of other issues with this system. The PCIe x16 slots only function at x8. While this doesn't impair current gen GPUs in any noticeable way, it may be a concern for the future (see notes here: https://www.dell.com/community/Desktops-General/Inspiron-5675-motherboard-PCIe-slots/m-p/5171757#M1029192 ) As well, the m.2 NVME slot functions at Gen2 speeds, effectively halving the available bandwidth ( https://www.dell.com/community/Desktops-General/Inspiron-5675-Desktop-NVME-M-2-slot-speed/td-p/5172495v )

So if you're considering one of these systems, buyer beware. The CPU performance is great, but don't be surprised if the GPU leaves you underwhelmed. Overall, it's still a good deal.
 
You could not get full RX 580 performance without the two PCIe connectors. As the RX 580 is more or less an overclocked RX 480, programs will often give you mixed messages on what the card is when you are dealing with rebadged cards. I believe that it is actually a RX 580 which was modified and downclocked to meet power usage restraints but is still technically an RX 580. I find this dishonest personally as when you buy a PC with an RX 580 you expect full RX 580 performance. I'd actually return the PC if this happened to me if it wasn't bought on sale. I will have to watch for this now as there are many very appealing RX 580 prebuilds available right now.

Out of curiosity, could you post a userbenchmark test? I'd like to see this card's performance relative to other RX 580 cards.

PCIe x8 speed slots should not bottleneck a gaming graphics card even if you had a GTX 1080. There is still enough bandwidth. If the slot bottlenecked the card then Dell probably would not have opted for that configuration.
 

Sure you can, there are a number of RX 580s with only one 8 pin connector.

That is kind of annoying about the x8 PCIe slot and the gen 2 PCIe lanes for the M.2 slot. I'm guessing Dell decided to use the same mobo for both Ryzen and AM4 APU systems. Since APUs only have 8 PCIe 3.0 lanes and 4 PCIe 2.0 lanes (usually used for NMVe storage), they may have designed it to cater to the lowest common denominator as a cost saving measure.
 
@TJ Hooker This thing has one 6-pin PCIe power socket, not 8-pin. That means that there's a ceiling of 150 watts this thing can pull. An 8-pin would add 150w to the mix, giving the card a maximum of 225 watts. I think @jr9 is probably right.

@jr9 I'm interested in comparing benchmarks too. At present I have the "580" swapped out for a GTX 970, and I've got another project on my plate this evening (replacing the heatsink & heatpipe in a Macbook Pro.) I'll try to see if I can benchmark it later this week. In the meantime, you can check out the Fire Strike score here:
https://www.3dmark.com/fs/15186411
and Time Spy here:
https://www.3dmark.com/spy/3501266
I ran these when I first got the PC, before I swapped out the 580 for my old 970. Note that 3dmark identifies it as an RX 480 even though Dell markets as a 580.
 
This is nothing new for Dell. I have a pile of 1792MB Dell GTX260 cards with only one 6-pin. They are only 192-shader (all of the other 55nm 260 cards were 216-shader and thus equivalent in performance to the HD4870) and the clocks were dropped from the stock 576/1242 to 518/1080, presumably all so they could run on just 150w.

Think of all the pennies Dell saved on cheaper PSUs with only one PCIe connector.
 
For me that's actually worst than the GPU rebadge. Not only that but they probably used cheap Kingston sticks and the Dell motherboard caps you at 2400MHz for RAM speed when you want higher than that on Ryzen. You might even be running at 2133MHz. If it was a non Dell motherboard you could figure out which motherboard they used for that PC then try to find RAM would run at 2666MHZ to 3200MHZ which is what you want with Ryzen.
 
well rebranding is very common with OEM. this happen with both AMD and nvidia. though realistically even true RX580 won't be much faster than 970. especially factory overclocked one. in one of hardware unboxed test overclocked 970 most often almost as fast as EVGA GTX1060 FTW they use in the test.
 
@jr9 I installed the 580 back in the Dell and ran userbenchmark on it.

Here are the results with my 970: http://www.userbenchmark.com/UserRun/8015390

And with the 580: http://www.userbenchmark.com/UserRun/8015959

It looks like my 'problem' is that I have a strong GTX 970, as it looks like it's in the 93rd percentile of all tested 970s. During the benchmarks it was hitting a steady 1450 Mhz core clock. In addition to the factory overclock, I also applied an extra 150Mhz to the core clock and 175Mhz to the memory clock. I didn't try very hard either. The 970 may be capable of even more. I have not done extensive stress testing, however - other than gaming for a few hours.

The 580 actually seems to benchmark decently compared to its peers, though it doesn't stand out from the crowd. Overall it seems pretty inferior to my particular 970. I have not tried overclocking it, though as someone pointed it that may be difficult with only 150 watts to work with.

EDIT: I may have to eat a bit of crow. I was confused about base clock vs boost clock, as well as the Dell forum post about the 470 vs 570 debacle. The base clock of the 480 is ~1120Mhz and the boost clock is 1266. When I looked at the *base* clock of my 580 in the Adrenalin utility it showed 1266MHz and I connected it immediately to the RX 480 *boost* speed. However, looking at the reference specs of the 580 it shows a *base* clock of 1266, exactly what my card shows. So despite this card being identified as a 480 by some utilities and only having one PCIe 6-pin connector, it may well be a 'real' 580.

Seems odd that the base clock of the 580 is exactly the same as the boost speed of the 480...
 

That's why I always trust GPU-Z. It shows the default clock, the OC clock, it has a "stress" mode that can show the real GPU clock under stress.
My RX580 had a default clock of 1350MHz, I raised to 1485MHz, and under 100% GPU load, it stays at 1485MHz, with a total power draw of 140W ( fan speed 3500rpm at 69-70C temperature) - from that total the PCI-E power is 13.6W, the rest is via power connector.
 

How are you measuring total power, as well as power through the PCIe slot?
 
@TJ Hooker I used userbenchmark because jr9 was curious about what results I was seeing. I suppose he wanted to compare my system to his, or one he was considering buying. My Fire Strike and Time Spy results are posted further up the thread. I do know that my AC Origins framerates *heavily* favor the 970 over the 580.
 


I don't. The software does measure the power trough the connector and adds the 13.6W max from the PCI-E (that's not measured).
 
I was just wanting to see the simple performance percentage of the "580" vs the 970 thanks. The percentiles for the RX 580 are very badly screwed downwards so they aren't any use. They do tell me though that between that the 3Dmark scores the card doesn't perform as something like an MSI RX 580 would. I will have to note this as I had my eye on an RX 580 prebuilt. Gaming benchmarks generally agree that a regular RX 580 should be slightly better than or significantly better than a regular GTX 970 not slower.
 
passmark_rx580_full.png