390X vs 290X with 15.7 drivers?

[shady28]

On a technical level, they are not rebrands. Pcper had a good write-up about it in their review:

http://www.pcper.com/reviews/Graphics-Cards/Sapphire-Nitro-Radeon-R9-390-8GB-Review


They are not 1:1 chips with an overclock. With that said, the changes are not big and as Simian mentioned, I think the combination of matured drivers and good cooling solutions out of the gate are what gave the 390(x) launch a much more favorable impression. I like the term refresh because it implies a tweaked chip rather than an improved architecture or a straight rebrand with higher clocks.

You forgot to read the rest of the article.

This is the next sentence :

Obviously we need to judge all of that for ourselves.

And here is the conclusion :

...yes the Sapphire Nitro R9 390 used less power in our testing than the ASUS Radeon R9 290X retail card, but the XFX retail R9 290 actually used about 20 watts less power than the Sapphire retail R9 390. While not what I was hoping to see, the results aren't damning; they just point to a less than advertised improvement in GPU efficiency. I guess I'm not that surprised.

So basically some marketing type made some claim... that didn't pan out. Most likely any power differences in the cards are from using more up-to-date VRAM with better process tech than what was available 2 years ago.

 

[Face2Face]

From Ryan's article -

http://www.anandtech.com/show/9387/amd-radeon-300-series/3

Last but certainly not least however, we want to talk a bit more about the performance optimizations AMD has been working on for the 390 series. While we’re still tracking down more details on just what changes AMD has made, AMD had told us that there are a number of small changes from the 290 series to the 390 series that should improve performance by several percent on a clock-for-clock, apples-to-apples basis. That means along with the 20% memory clockspeed increase and 5% GPU clockspeed increase, we should see further performance improvements from these lower-level changes, which is also why we can’t just overclock a 290X and call it a 390X.

So what are those changes? From our discussions with AMD, we have been told that the clock-for-clock performance gains comes from a multitude of small factors, things the company has learned from and been able to optimize for over the last 2 years. AMD did not name all of those factors, but there were a couple of optimizations in particular that were pointed out.

The first optimization is that AMD has gone back and refined their process for identifying the operating voltages of Hawaii chips, with the net outcome being that Hawaii voltages should be down a hair, reducing power and/or thermal throttling. The second optimization mentioned is that the 4Gb GDDR5 chips being used offer better timings than the 2Gb chips, which can depending on the timings improve various aspects of memory performance. Most likely AMD has reinvested these timing gains into improving the memory clockspeeds, but until we get our hands on a 390X card we won’t know for sure.

 

[werepossum]

While I lean toward calling them rebadges, isn't the base 390 running neck and neck with a similarly overclocked 8GB 290X?

 

[Hitman928]

You forgot to read the rest of the article.

This is the next sentence :




And here is the conclusion :



So basically some marketing type made some claim... that didn't pan out. Most likely any power differences in the cards are from using more up-to-date VRAM with better process tech than what was available 2 years ago.

They used a OC model which typically will sacrifice power draw for OC headroom and will draw more power than a stock bios. If you look at another model (HiS iceQ) which doesn't do this, the cards actually draw the same as the 200 models despite 7% higher clocks and 20% faster memory with twice the memory on card.

http://www.techspot.com/review/1019-radeon-r9-390x-390-380/page7.html

http://www.legionhardware.com/articles_pages/his_iceq_xsup2_oc_radeon_r9_390xr9_390_r9_380,10.html

Again, not a huge difference, but not the same exact die either. AMD said they made some tweaks and I believe them. Is it compelling enough for a "new gen" label (290->390) not to me, but that's the marketing.

 

[Hitman928]

While I lean toward calling them rebadges, isn't the base 390 running neck and neck with a similarly overclocked 8GB 290X?

The 290 I don't think should be throttling at reference clocks, even with the crappy cooler, could be wrong though. Again, I think matured drivers and good cooling explain most of it, but the faster memory and maybe 1-2 % performance tweaks help too.

 

[96Firebird]

I don't remember people asking sites to lower GTX 770 clocks to match GTX 680 and test them. Troll thread

Who asked sites to lower any clocks? Getting the same clocks can be as easy as overclocking the 290X to 390X speeds.

Call it a troll thread it if you want, simply because you don't like the results that could come from it. Better to hide your head in the sand, right?!

I'm mostly curious if AMD pulled a fast one and withheld the performance-increasing drivers for the 200-series to make the 300-series look better for launch reviews, but also would like to know if there is a performance difference large enough to recommend the 300-series over comparable 200-series...

 

[Silverforce11]

It doesn't matter because 290/X supplies are drying up. So while there's time now to compare 290/X vs 390/X and which to buy (I would recommend 290/X!)... its harder to find them nowadays.

In the end, it's just a refresh, nothing exciting*. I don't call it a rebadge because it improves up reference 290/X specs. A rebadge is taking the exact same SKU, slap a new label on it.

* But interesting to see GCN still getting major perf boosts and now 390 > 970 and 390X = 980 (1440/1600p). They can also put up a decent OC with vcore so its competitive.

Check out the OC v OC comparison vs 390X/970/980 (I take the most recent on Guru3d):

970: http://www.guru3d.com/articles_pages/gigabyte_geforce_gtx_970_oc_mini_itx_review,28.html

http://www.guru3d.com/articles_pages/asus_geforce_gtx_980_matrix_platinum_review,26.html

980: http://www.guru3d.com/articles_pages/asus_geforce_gtx_980_matrix_platinum_review,26.html

390X: http://www.guru3d.com/articles_pages/msi_radeon_r9_390x_gaming_8g_oc_review,26.html

The 390X has excellent scaling for a 100mhz OC, 1.2ghz from 1.1ghz.

390: http://www.guru3d.com/articles_pages/powercolor_radeon_r9_390_pcs_8gb_review,26.html

Basically 290 ~= 970, 290X ~10% behind 980. AMD made up that perf gap with the 390/X and manage to keep power usage similar while having extra vram. It's a refresh.

On this GCN OC, every iteration seems to peak around 1.2ghz (my R290s did 1.2ghz but elpida vram basically meant could not even go 25mhz over base mem clocks), with golden samples hitting 1.25ghz. Average samples hit 1.15ghz. Thus I would expect similar for Fury/X one vcore control is available.

 

[Headfoot]

If 680 -> 770 = refresh, 290x -> 390x = refresh

 

[ocre]

What the heck?

No one did a review of a gtx680 at gtx770 clocks specifically to compare because everyone knew the results would be the same. They new it was a gk104 rehash. With a slightly more mature node, you can usually increase MHz at the same power consumption.

Everyone knew that the 770 was a rehash, respin, whatever you want to call it.

There was no special driver at launch and I don't remember anyone trying to promote it differently.

But with the 390(x), wow!

First AMD calls it Grenada instead of Hawaii. Then they have a special 300 series driver that wouldn't work for 200 series cards. And now lastly, we have threads like this where people are debating against something so absolutely obvious. And everyone calls nvidia a spin machine.

AMD has put a lot of effort in...........

Well, I guess that is enough. People will get mad

 

[shady28]

What the heck?

No one did a review of a gtx680 at gtx770 clocks specifically to compare because everyone knew the results would be the same. They new it was a gk104 rehash. With a slightly more mature node, you can usually increase MHz at the same power consumption.

Everyone knew that the 770 was a rehash, respin, whatever you want to call it.

There was no special driver at launch and I don't remember anyone trying to promote it differently.

But with the 390(x), wow!

First AMD calls it Grenada instead of Hawaii. Then they have a special 300 series driver that wouldn't work for 200 series cards. And now lastly, we have threads like this where people are debating against something so absolutely obvious. And everyone calls nvidia a spin machine.

AMD has put a lot of effort in...........

Well, I guess that is enough. People will get mad

Exactly.

The 770 was absolutely a rebadge of the 680, and that was clear when it was launched :

"NVIDIA's GeForce GTX 680 Gets Tuned, Rebranded as $399 GTX 770

http://www.dailytech.com/NVIDIAs+Ge...ned+Rebranded+as+399+GTX+770/article31660.htm


"Nvidia launches the Geforce GTX 770, a rebranded GTX 680"

http://vr-zone.com/articles/nvidia-...-a-rebranded-gtx-680/33153.html#ixzz3gAY9Dopv


THIS is a refresh :

http://www.anandtech.com/show/6973/nvidia-geforce-gtx-780-review

"The GeForce GTX 780 is the follow-up to last year’s GeForce GTX 680, and is a prime example of refreshing a product line by bringing in a larger, more powerful GPU that was previously relegated to a higher tier product. Whereas GTX 680 was based on a fully-enabled GK104 GPU, GTX 780 is based on a cut-down GK110 GPU, NVIDIA’s monster GPU first launched into the prosumer space with GTX Titan earlier this year. "


Other refreshes would include things like the 960 - basically a v2 Maxwell + HEVC decoding.

Or v2 maxwell - which is v1 Maxwell expanded (new die) + HDMI 2.0 and Delta Color Compression.

Those are refreshes to the GPU. Changing the type of VRAM you use and making some BIOS tweaks, that's not a refresh that's something every AIB does.

Honestly AMD has 3 architectures and that's it. GCN 1.0 1.1 and 1.2. Using the same measuring stick as has been used against Nvidia (and the one that makes sense), Fiji is a major refresh and expansion of Tonga, both of which are GCN 1.2 arch. Using that same measuring stick against 390 / 380 / 370 - they are all rebadge / rebrands. That's just a fact.

 

[tweakboy]

I would say its 30 to 50 percent faster the 390x,, for example..,, etc

 

[werepossum]

What the heck?

No one did a review of a gtx680 at gtx770 clocks specifically to compare because everyone knew the results would be the same. They new it was a gk104 rehash. With a slightly more mature node, you can usually increase MHz at the same power consumption.

Everyone knew that the 770 was a rehash, respin, whatever you want to call it.

There was no special driver at launch and I don't remember anyone trying to promote it differently.

But with the 390(x), wow!

First AMD calls it Grenada instead of Hawaii. Then they have a special 300 series driver that wouldn't work for 200 series cards. And now lastly, we have threads like this where people are debating against something so absolutely obvious. And everyone calls nvidia a spin machine.

AMD has put a lot of effort in...........

Well, I guess that is enough. People will get mad

I think the driver doesn't work with the 290 series because some parts of the chip (power management and memory controller IIRC) were actually changed. Question is whether that (together with faster and more efficient VRAM) is enough to make the 300 series a refresh. I lean toward calling it a rebadge, but I am impressed with it and plan on buying one*. And I promise not to get mad whether people call it a rebadge or a refresh.

* I am torn between the R9 390 and the GTX970 but heavily leaning toward the former. I like the idea of more VRAM for megatextures but I dislike the idea of my office getting hotter when I game.

 

[Gikaseixas]

Who asked sites to lower any clocks? Getting the same clocks can be as easy as overclocking the 290X to 390X speeds.

Call it a troll thread it if you want, simply because you don't like the results that could come from it. Better to hide your head in the sand, right?!

I'm mostly curious if AMD pulled a fast one and withheld the performance-increasing drivers for the 200-series to make the 300-series look better for launch reviews, but also would like to know if there is a performance difference large enough to recommend the 300-series over comparable 200-series...

No need to get all emotional kid, I'm not concerned if AMD gets exposed or Nvidia for that matter, I just don't understand what drives some here to purposely look for faults here and there on stuff they themselves would never purchase. That applies to people in both camps

 

[omek]

The 290 I don't think should be throttling at reference clocks, even with the crappy cooler, could be wrong though. Again, I think matured drivers and good cooling explain most of it, but the faster memory and maybe 1-2 % performance tweaks help too.

Yeah, that's completely in-line with the 10% gain that the 390/x has over the 290x. We all know that Hawaii's core clock yields very linear results - add 5% to the core, 20% on the memory and compare that with the 10% gain. The 20% on the memory most likely translates to 3-5% real world. 5% on the core, 3-5% on the memory is 8-10% which is tagging the 10% gain that the 390x has.

There are a load of anomalies (for lack of a better word) that make the 390/x look much faster at times

1. The drivers have matured since the release of the 290/x which leads to the question:
_a. Are they recycling old 290/x benchmarks? 14.12 was extremely popular.
_b. What driver version are they using for the 290/x? It's unknown most of the time because the focal point is on the 390/x (remembering that 15.15 was not able to run on a 290/x)
2. 15.15 was a largely improved driver showing a large increase in DX11 draw call performance and either tessellation profiles per game or a tessellation enhancement across the board; no one is quite sure but it may tie into the draw call increase or be misrepresented by it and seen as a driver level tessellation improvement. It also was the driver that the 390/x shipped with and only the 390/x was "able" to use it leaving the 290x without any of the 15.15 enhancements which was seen in the reviews. Of course the driver was modded to allow the 290/x to use it but your not going to see that comparison on any large review site.
3. Thermal throttling. Review sites continue to use the reference 290/x and throttling (even if it's unoften) most likely still happens skewing the comparison slightly. There's no throttling at all with the 390/x, just as there wouldn't be with an aftermarket 290/x

With that said the 390/x are great cards, just as great as the 290x is without the perceptual baggage. I'd have to argue against it being a rebrand because there are memory differences, from size to timings. I'd be more inclined to call it a refresh while being very close to being a rebrand.

 

[96Firebird]

You think my post was emotional? Wait until you get out into the real world, you'll see what real emotions look like at every turn...

This thread isn't about definitions or anything like that, I am looking for data to draw facts from.

 

[Abwx]

Grenada has better frequency scaling than Hawai, this suggest that there was changes at very low level with latencies within the GPU staying constant at extreme frequencies. This has nothing to do with drivers, it s the GPU that can sustain its IPC better as frequency is increased.

 

[96Firebird]

Grenada has better frequency scaling than Hawai, this suggest that there was changes at very low level with latencies within the GPU staying constant at extreme frequencies. This has nothing to do with drivers, it s the GPU that can sustain its IPC better as frequency is increased.

Great, it shouldn't be hard to back that up with data!

 

[Abwx]

Great, it shouldn't be hard to back that up with data!

Of course, just a few post before yours...

http://forums.anandtech.com/showpost.php?p=37564917&postcount=32

 

[96Firebird]

Hawaii not found.

 

[soccerballtux]

I've been trying to make this decision myself. I don't game much anymore, but when I do I want it to be decent. my 7850 is basically 80% acceptable, which is a frustrating place to be (80/20 rule split). Additionally I'll probably upgrade for when UT4 comes out so...tough place to be.

 

[soccerballtux]

WHat driver? If you take a 369$ 8gb 290x with a aftermarket cooler, overclock it to 1050, it is a 429$ 390x. Can I make it any more simple than that?

Why would I not trust (H)? They are [an honest] site that doesn't care about the whining fanboys. I like em.

how high does the 390x clock?

how high does the 290x clock?

These will help me with my decision

 

[soccerballtux]

One thing that a lot of people fail to mention or take into account is that not every 290/290X will do 1500 memory out of the box, especially not without a voltage bump. Cards with Elpida memory especially had trouble running at those speeds. A lot of the new 390's also have significantly bigger coolers than the non-reference 290's had (for example see MSI's R9 390X vs their 290X gaming cards). Regardless of what you want to call it, rebrand or refresh - the better coolers on launch, mature drivers, higher out of the box clock speeds, etc make them much more appealing than the 290/290X at launch. The 290/290X are a great purchase if you can still find them for cheap, but I don't think you can go wrong with a 390/390X either - they are quite competitive with NVIDIA's offerings. In fact - the 390X might be so competitive that it eats into AMD's own higher end offerings (the Fury non-x for example). It's hard not to see the value in the 390/390X - 8GB of VRAM at a fairly affordable price.

what do you consider 'cheap'?

how well do you think 290x 4GB will handle UE4/UT4 at 1440p?

 

[Hitman928]

FWIW, from Harocop:

http://www.hardocp.com/article/2014...0x_gaming_4g_video_card_review/3#.Vamf6LaVukA

First we raise the PowerLimit to +50 to allow the highest overclock and performance. We managed to push the Core Clock speed from 1100MHz to 1150MHz without any voltage or fan changes. . . this overclock is at the maximum levels we got out of R9 290X WITH voltage control. In fact, even some cards with voltage would not overclock this high.

Their final OC on the 390x was 1180 MHz. A cursory survey of HardOCP, Guru3d and a couple others seem to confirm the sentiment. 290X topped out ~1100-1130 MHz while the 390x seems to top out about 1170 - 1200 MHz.

 

[soccerballtux]

ok, that's good to know. 1100 vs 1170 isn't worth the extra $60 (or $200 if bought 290x used) to me.

 

[Hitman928]

Yeah, the difference between the 290/390 generations aren't really enough for that price bump unless the 8GB of memory is something you feel is important.