290x vs 390x

[Edgar R. Zakarian]

Is it the same Card just overclocked?

 

If I buy both Cards from Gigabyte (or other, just same both 8GB models)

 

Will overclocking the 290x to the 390x speeds produce same FPS / Thermal results as the 390x?

Or are the Cards fundamentally different?

[Mew]

AMD claims the 390x is an improved 290x with better ipc. Some people call it gimping in drivers as they are architecturally the same. 

[Suffokation]

The 390X has an extra 4GB frame buffer.

[Curufinwe_wins]

AMD claims the 390x is an improved 290x with better ipc. Some people call it gimping in drivers as they are architecturally the same. 

They did pass most of the driver differences on to the 290x a month or so ago (although the bios setup is slightly different, and accounts for literally all of the remaining differences in the cards outside of better cooling/components).

 

The 390x is overpriced relative to both the 390 and the 290x (assuming you aren't paying more than the 390 for one), and the extra vram (assuming you are not looking at a 8GB 290x) is nice, but not important basically ever...yet.

[manikyath]

someone on the forum has put a 390 bios on a 290, they can give you more info on this.

 

very paraphrased explanation:

they "refined" their process a bit, basicly creating "higher binned" chips more reliably, so they can clock them higher out the factory.

 

that, as well as being 8GB VRAM by default, which is quite nice for the modded skyrim folks

 

and for the people who use that one second life client that can cripple a titan X...

[Sakkura]

Will overclocking the 290x to the 390x speeds produce same FPS / Thermal results as the 390x?

 

If you set a 290X and 390X to the same clocks, they will perform identically. But the 390X does come with better memory chips that should hit higher clocks in general, as well as 8GB standard where the 290X is 4GB standard (but with more expensive 8GB versions around). The 390X GPUs may also on average be binned a little higher, thus giving you better chances in the silicon lottery for overclocking. Shouldn't be a big difference though.

[Sir Asvald]

[Nacho Marco Segui]

Please get the Sapphire/MSI version,the gigabyte one is loud,hot and voltage locked

[WereCat]

R9 390x has slightly lower TDP resulting in slightly lower temps. Is better bined resulting in higher average clocks thus better performance on average. Has a lot faster memory resulting in better performance.

390x is overall a better card by about 10% in performance.

[Kinda Bottlenecked]

The 390x scale better with less voltage. The 290x scale better with more voltage. If you're an overclocking enthusiast the 290x is the better choice. If you want better everyday overclocks, the 390x is the better option. Either way, avoid Gigabyte all together for AMD as well. 

[Coaxialgamer]

I suppose they would perform the same at the same speeds , as both cards are based on the gcn architecture . The grenada chip in the 390x has a lower power consumption and less heat output compared to the 290x . This is probably due to the fact that the chip requires less voltage to run .

But to answer your questions , both cards have the same "IPC" and should perform the same at equal speeds, except when you are limited by your frame buffer . (>4gb usage).

[Curufinwe_wins]

I suppose they would perform the same at the same speeds , as both cards are based on the gcn architecture . The grenada chip in the 390x has a lower power consumption and less heat output compared to the 290x . This is probably due to the fact that the chip requires less voltage to run .

But to answer your questions , both cards have the same "IPC" and should perform the same at equal speeds, except when you are limited by your frame buffer . (>4gb usage).

90% of the improvement (which is minimal at best to be honest) is just newer binned chips and newer power delivery components (just like how later batches of intel chips tend to OC better).

 

The extra 10% being bios changes that caused more common midrange overclocks (to allow higher end factory base clocks) but that appeared to reduce their max overclocks when not thermally limited (voltage modded 390(x) bios's from everyone but MSI are basically still never reaching 1200 core clock which is bizarre, but many of the MSI 390(x) can reach over 1200 without bios hacking). (It's a pretty interesting topic at the  moment actually.)

 

Rest is true.

[Hiebly]

The firmware and drivers slightly improve energy management allowing the core clock to be a bit higher, and the 390/390X have better VRAM than the 290/290X, so there is a pretty big bump in memory clock speed. Otherwise yes, the cards are architecturally the same and the 390X is more or less just an overclocked 290X.

[xAcid9]

Clock per clock they will perform similar but good luck getting the memory on the 290x to 1500mhz without an increase in voltage and temp. Higher memory clock in 390/390x is pretty important because it slightly decrease latency thus improve some overhead.

 

According to The Stilt, memory on 290s with Hynix/Elpida mostly won't go past 1450mhz without producing ECC error and it's not something you can monitor.

Hawaii with AFR or Elpida doesn´t seem to do over 1450MHz or so without EDC (ECC errors).

Increasing VDDCI will help slightly, but the I´ve never seen an air cooled Hawaii card to do past 1480MHz without at least one of the eight sequencers producing errors. The EDC is slightly problematic as there is no way the end user can monitor those and they are not visible as artifacts (due correction) either. You just will end up basically wasting cycles when you push the memory too high. The Hynix AFR modules themselves can do >1700MHz easily just like on other cards, however the eight memory controllers cannot.

 

For Grenada, AMD most likely latched some of the timings affecting the latency between different controllers / memory channels to achieve higher memory clocking without errors. Since there are literally at least a hundred different memory timings, I neither have the time or will to find the correct one(s) and tune them for Hawaii.

 

Part of the difference might be caused by the newer GDDR5 (Hynix Polaris) die manufactured on smaller node.

 

At 1375MHz and with tuned latencies you should by no means be limited by the bandwidth anyway.

http://www.overclock.net/t/1561904/mlu-bios-builds-for-290x/10#post_24120051

 

AMD already merged their 390s and 290s driver so there's no exclusivity in driver improvement.

Not really worth it to flash 390x BIOS into 290x.

[Curufinwe_wins]

Clock per clock they will perform similar but good luck getting the memory on the 290x to 1500mhz without an increase in voltage and temp. Higher memory clock in 390/390x is pretty important because it slightly decrease latency thus improve some overhead.

 

According to The Stilt, memory on 290s with Hynix/Elpida mostly won't go past 1450mhz without producing ECC error and it's not something you can monitor.

 

AMD already merged their 390s and 290s driver so there's no exclusivity in driver improvement.

Not really worth it to flash 390x BIOS into 290x.

Indeed, although the wider bus than say nvidia's cards makes memory overclocking much less of an issue (on maxwell memory overclocking is actually a really big overall boost for the 970 and 980...

[WaitForMeImOnDialUp]

they are the same thing clock for clock. which is why i bought a 290x instead of a 390x. look at the new star wars benchmarks,  290x matching the gtx 980!!!!!!

 

 

 

 

that said its easier to reach higher overclocks on 390x than it is with 290x, if you liquid cool then the 290x can still beat it,even a 290 could. so far i pushed mine to 1150/1500 which is better than a 390x stock for sure.

 

a used reference 290x can be bought here in usa for around 200-220, or buy a new 390x for 400

[Kinda Bottlenecked]

they are the same thing clock for clock. which is why i bought a 290x instead of a 390x. look at the new star wars benchmarks,  290x matching the gtx 980!!!!!!

 

 

that said its easier to reach higher overclocks on 390x than it is with 290x, if you liquid cool then the 290x can still beat it,even a 290 could. so far i pushed mine to 1150/1500 which is better than a 390x stock for sure.

 

a used reference 290x can be bought here in usa for around 200-220, or buy a new 390x for 400

 

A 290x/390x matching a 980? Looks like Nvidia hasn't finalize the drivers for battlefront. 

[Curufinwe_wins]

A 290x/390x matching a 980? Looks like Nvidia hasn't finalize the drivers for battlefront. 

I mean it all depends some specifics. I am always most interested in post-oc numbers (with listed core/mem clocks ofc) as I feel overclocking is a must in the first place, and that often changes the hierarchy of graphics cards by a decent amount.

[lxyfutures]

That really depends on what resolution...If you play in 1080P,290X would be a great choice .In 2K I think 4GB vram already makes me feel not enough...You can buy 8GB version of 290X,but I think it will cost the same as a 390X...Not worth it. 

[Kinda Bottlenecked]

I mean it all depends some specifics. I am always most interested in post-oc numbers (with listed core/mem clocks ofc) as I feel overclocking is a must in the first place, and that often changes the hierarchy of graphics cards by a decent amount.

 

Doesn't change anything though. Overclock both and both will have a good amount of FPS increase. The strange thing is that 290x/390x edged out the 980 when it isn't supposed to on stock speeds. Might mean that drivers or the game itself needs a lil bit more work.

[Curufinwe_wins]

Doesn't change anything though. Overclock both and both will have a good amount of FPS increase. The strange thing is that 290x/390x edged out the 980 when it isn't supposed to on stock speeds. Might mean that drivers or the game itself needs a lil bit more work.

That isn't true because the amount of overclocking headroom between the cards is massively difference. Plus memory overclocking tends to remove vram bus limitations (and can slightly make up for a lower amount overall.)

[Kinda Bottlenecked]

That isn't true because the amount of overclocking headroom between the cards is massively difference. Plus memory overclocking tends to remove vram bus limitations (and can slightly make up for a lower amount overall.)

 

hmm? What is? If you're talking about clock speeds, that doesn't mean anything. Both cards overclock differently. 

[Curufinwe_wins]

hmm? What is? If you're talking about clock speeds, that doesn't mean anything. Both cards overclock differently. 

Overall performance headroom for overclocking is a bit higher on the 970/980 as it is on the 390/390x, and is massively higher for the 980ti than for any fiji chip. 

 

But that is already known. Obviously in some cases the 390(x) will still win (like so far is indicated by dx12), but in many current cases where they are nearly tied for "stock to stock" (even though there is no stock 390(x), and blower coolers are stupid unless in sli) they are not tied when overclocking is taken into account.

 

And I don't say that to be a nvidia or amd fanboy. It's just the truth at the moment. It should be obvious that different cards running different architectures have different overall performance headroom for overclocking, and personally I would never not-overclock a graphics card, so I see literally no use for me personally in seeing non-overclocked benchmarks. (hence why I like jayztwocents as a benchmark source.)

[Kinda Bottlenecked]

Overall performance headroom for overclocking is a bit higher on the 970/980 as it is on the 390/390x, and is massively higher for the 980ti than for any fiji chip. 

 

But that is already known. Obviously in some cases the 390(x) will still win (like so far is indicated by dx12), but in many current cases where they are nearly tied for "stock to stock" (even though there is no stock 390(x), and blower coolers are stupid unless in sli) they are not tied when overclocking is taken into account.

 

And I don't say that to be a nvidia or amd fanboy. It's just the truth at the moment. It should be obvious that different cards running different architectures have different overall performance headroom for overclocking, and personally I would never not-overclock a graphics card, so I see literally no use for me personally in seeing non-overclocked benchmarks. (hence why I like jayztwocents as a benchmark source.)

 

In terms of clock speeds yea. Not so much for in game performance. Hawaii and the other GCN cards still overclock alright. Though not so much for GCN 1.3. Now those are not good overclockers. 

 

It is fair to say that the 980 Ti overclocks better than the Fury X in general(Fury/X can't overclock for nuts, unless u do LN2). Sadly the same is not said for the others like the 980/390x/290x or  970/390x/290 and so on. Because those when you have a nice overclock will have a good amount of FPS increase for all of those cards. Like buying a GTX G1 Gaming or so for some nice overclocks. You gotta buy the right card on the other side as well. 

[don_svetlio]

Overall performance headroom for overclocking is a bit higher on the 970/980 as it is on the 390/390x, and is massively higher for the 980ti than for any fiji chip. 

 

But that is already known. Obviously in some cases the 390(x) will still win (like so far is indicated by dx12), but in many current cases where they are nearly tied for "stock to stock" (even though there is no stock 390(x), and blower coolers are stupid unless in sli) they are not tied when overclocking is taken into account.

 

And I don't say that to be a nvidia or amd fanboy. It's just the truth at the moment. It should be obvious that different cards running different architectures have different overall performance headroom for overclocking, and personally I would never not-overclock a graphics card, so I see literally no use for me personally in seeing non-overclocked benchmarks. (hence why I like jayztwocents as a benchmark source.)

I just want to point out - AMD have roughly 50% more instructions per clock so 1MHz on an AMD card is about 1.5MHz on an Nvidia competitor - roughly.