AMD Radeon R9 290 Can Be Easily Transformed into R9 290X ? Report.

Ghstops6

Ghstops6

Admin
Admin
Joined
Sep 2, 2012
Messages
1,389
AMD Radeon R9 290 Can Be Easily Transformed into R9 290X

Just caught this little bit of interesting news in my news feed. So for those who don't want to pay a premium for the 290x, you can always BIOS flash the R9 290. According to the article, there is a 100% success rate. And it does unlock and enable all stream processors on the GPU.

Not sure if this is a mistake on AMD's part, or pure genius.
 
Update, TL/DR: Apparently, if you hope to have a CHANCE of doing this (I wouldn't count on it being 100% by any means, even with Powercolor - you never know when they can switch the chips), buy Powercolor: The R9 290 -> 290X Unlock Thread

Also, this might be "locked" in the future by AMD easily, and probably will. I'm thinking if I should get one right away, might if there's a good derp friday deal somewhere next week.
 
VTX3D-R9-290X-635x436.jpg

You can now buy 290 Pre-Flashed to the 290x with a 2 year warranty!!

Details Retailer Selling Pre-Flashed AMD R9 290 to R9 290X just for ?359.99 - Is AMD in Trouble?

Direct: VTX3D Radeon R9 290 Pre-Flashed to 290X, X-Edition 4096MB GDDR5 PCI-Express Graphics Card [VXR9 290 4GBD5-MDHX]
 
I'm not in England, and who the fuck is VTX?

It'd be kind of lulrarious if AMD implemented some kind of check in the drivers that fucks everyone who did the bios reflash. Unlikely though, but they'll probably physically shit on the newer chips sooner or later.
 
Soooooooooooooo newegg offered 5% off sitewide, and my credit card gives me another 1%, so I decided to bite my dick and order a XFX R9 290 because of the BF4 promo and reports that NONE of the new Powercolors from Newegg are unlocking (and some Sapphires are unlocking now? wtf.).

So, I'm a bit late for the unlock train, but wish me luck. I'll know on Wednesday. If it doesn't unlock, I'm almost tempted to return it to newegg for another try, but we'll see. I might not bother though, because the increase is about 7-8%, and while annoying, the card overclocked on water should be pretty close anyway :-S .

I guess it's official, I'm riding the AMD cock now. Can't beat price/performance, can't find reason to spend twice the amount for 10% more FPS with nvidia and 50% worse mining. Next upgrade will probably be for reasons of 4Gay(mmmK?).

I'll probably be mining on it for a while until the R9 290 waterblocks are back in stock (all I have to say about that is >.< ) . Hope it'll be all assembled and running by xmas.

Also having a good experience so far with XFX, returning a 7950 with two broken fanblades (don't ask lul), and they issued the RMA no questions asked. I just hope the replacement card isn't complete shit :-P . Oh, and as long as you let them know you're installing a waterblock, they actually honor your warranty.

In other news: WTS BF4 code. I hear they're going for $40ish? Anyone interested??? Will probably have it by Wed.
 
Fuck, FUCK yes. FUCKING yes.

Yes.

I scored this bitch hard.

oech.png


For the unenlightened, this means it's flashable to 290x and the shaders unlock. Basically, it's a 290x with the wrong bios ;-P .

Again, fucking. Yes.

XFX from Newegg ordered Sunday, arrived today. Good luck to everyone who's doing this. It should be a great card even without the extra Shad3rZ.

I'm not sure with what kind of mental fortitude I'll be installing the water block when it arrives on this priceless gem; I'll have to take some anxiety pills. Can't afford to fuck this one up :-) . When the waterblock arrives, me and this baby are in for some good times. My body is ready.

If someone wants to buy my BF4 code, let me know. I didn't get any leaflet with the card, I'm guessing I have to register it? I'll have to look into that.

In other rumors, there are derpings around the internets of a "Titan Black Edition." Basically a GTX 780 Ti with 12GB of memory (lulz?) and completely removed power and voltage restrictions.
 
Update:

This card is fucking weird. First of all, it wouldn't go above 50% without being forced to (by afterburner or cgminer). By default, it would go to 50% when pushed, and then start throttling the clocks. HARD. I'm talking about drops from 950 MHz core to 600 MHz. Not sure if that's a "feature" of the newest AMD drivers, or the bios of the card, but it's pretty fucking stupid... Then again, once you start raising the fan above 50%, it starts sounding really, REALLY loud. Like, REALLY loud. I'd forgotten what a ridiculously loud fan sounds like. It almost makes me scared that something will either burn out or blow up when it reaches about 70%. Up to 50% is OK.

With water, this will not be a concern for me, but I would seriously consider either waiting for absolutely massive aftermarket air coolers or buying nvidia if you are sensitive to fan noise. It's pretty insane. That, or expect massive downclocking when the card is at 100% "furmark-type" load.

In other news, 830KH/s scrypt hashing at 950 Core / 1500 Mem. This will do for a few days while the waterblock arrives :-P . Then I'll tweak more.
 
HeatSurge said:
This card is fucking weird. First of all, it wouldn't go above 50% without being forced to (by afterburner or cgminer). By default, it would go to 50% when pushed, and then start throttling the clocks. HARD. I'm talking about drops from 950 MHz core to 600 MHz. Not sure if that's a "feature" of the newest AMD drivers, or the bios of the card, but it's pretty fucking stupid... Then again, once you start raising the fan above 50%, it starts sounding really, REALLY loud. Like, REALLY loud. I'd forgotten what a ridiculously loud fan sounds like. It almost makes me scared that something will either burn out or blow up when it reaches about 70%. Up to 50% is OK.
Click to expand...

....HA! Anyways yeah can't wait to see some water sprinkled on a bitch.

Once you get it on some proper cooling be sure to do some endurance error checking. It is possible some of these cores were not disabled because they were bad or just to fill orders. Some of the cores might not have performed well enough to pass. Under water you will be much less likely to incur problems testing incurred on air so might be no issues but you don't know until you know and you won't sustain 1Ghz on air so.

Oh and as far as the 50% cap is there anything in there for the Silent mode (dunno if the 290 had it) for the 290X as that would be my first suspect for cause.
 
Ohhhhhhhhhh you're right, I'm retarded. I think there might be a physical switch for "silent/uber" mode in the back, durrrrrrr. I'll have to check :-d ... maybe that'll solve the "50% fan" issue, but from what I just read it lifts the limit from 40% (it's supposed to be 40%? LOL?) to 55%... so not quite a "solution," but then again I think you can make your own fan profile into a bios and flash it... as I said not really a concern for me with water, but just saying - it's pretty LOL to hard-limit the fan in a video card to 55%.

As far as the cores being disabled because they didn't perform, I doubt it. Reports are that they were short on chips around release so they basically put perfectly good 290X chips to meet supply for the demand on release and disabled them in the bios instead. The "non - Rx1/2 F8000" cards actually have a shader cluster physically laser cut to be made impossible to be "activated" through the bios.

I've always hated that shit - it's exactly like how I've read that railroad companies in the early days spent EXTRA to tear down perfectly good passenger cars so they could sell the "good" ones at a "premium" around the introduction of the class-tiered system. So AMD spends extra time and effort (incurs extra cost) to physically cut perfectly good chips so they can sell them for less, and the "good ones" for $150 more.
 
Waterblock installed, fuck everything about those fucking thermal pads that aren't pre-cut and arranging them individually over the memory chips.

After an annoying few hours, it's in and seems furmark stable at 1200MHz core / 1400 Mem... nice round numbers :-) . 60C furmark load, even hotter than my 480 lul (which was about 56-57 IIRC). Probably because I'm pushing it pretty hard in comparison, even small increases make it hard crash. Not flashed to 290X yet, annoying DOS boot that I don't want to deal with tonight, and I want to wait for the next MSI AB version anyway just in case... maybe I'll run it like this for a week and then flash it for extra lulz.

If anyone has any questions or wants anything run at 1080p or less at particular settings to ballpark performance, let me know. I don't have BF4 and give zero fucks for it though...
 
Have you watched the clocks (msi afterburner gives a pretty good graph) is it staying pegged or is it ever slightly dropping due to current? I imagine at 1200Mhz and combined with furmark its hitting its overcurrent protection. Which for furmark is normal its a ridiculous beast for sure.
 
It's stable, I'm watching AB of course. The only throttle that I know is temperature... and it's staying at 1200. But I'm getting some really weird/inconsistent results... will update soon. For now I'm trying out 1150/1500 instead.
 
HeatSurge said:
It's stable, I'm watching AB of course. The only throttle that I know is temperature... and it's staying at 1200. But I'm getting some really weird/inconsistent results... will update soon. For now I'm trying out 1150/1500 instead.
Click to expand...

According to the GPU rep and I quote off the top of my head "It will increase clocks until it reaches some other limit such as voltage or current".

Anyways furmark really isn't a sign of DX11 stable. Ive had clocks go all day in furmark and not error in basic error analsis and then crash 10 seconds into any decent DX 11 game. Then again this card is a couple generations newer not sure if there is much that hasn't been moved to the unified cores such as the tesselation units AMD used to have.
 
Actually you're right, when I was playing around, I noticed when I take "power limit" down to below +25-ish (+50 is max) the core clock throttles to 850-ish from 1200. So I took it back to the max.

As far as stability, yeah I discovered that pretty quickly. 1200/1400 seemed to be pretty "crashy" even though I could play things for a while and not crash sometimes?? At least ATI's driver seems to be pretty good at recovering from crashes which is amazing... saved me a lot of hard restart grief so far.

So, I took it down to 1150 and I could increase the memory more (1500, even more but I haven't tried I think 1500 is enough for this card :-D )... also about memory I think I'm going to start running into slowdowns related to error correction (if I'm not already, as far as I can tell I'm not though). So for now it seems to be quite stable in everything at 1150/1500.

As far as "inconsistency" goes, I was playing around with LTC mining yesterday, and that is just... weird. I'm not sure if it's how the miner is coded or the process itself, but I get ~860KH/s with 1000/1500 and when I increase the core to 1150/1500 it goes down to 730. :-D So, I think it's the weirdness of scrypt though, games seem to increase when you increase the clock from my benches, but I'm out to test more and for stability :-) .

I was hoping for more than 1150 on water... but no. Oh well, at least the unlock will 95% work and give me a bit extra :-) .

Edit: Flashed to 290X with atiwinflash. Worked just fine, no dos bullshit. Another free 3%-ish of performance in the bench that I ran. Pretty happy with this ^_^ .
 
Update: Scrypt (LTC+others) mining is really really weird, apparently. I didn't realize it made that huge of a difference, but there's a very specific clock-to-mem ratio that you have to run to get the most out of your card with the SAME power consumption! I had to LOWER my core clock when I flashed my card to 290X (probably has to do with the extra shaders) to get MUCH higher KH/s.

I'm not sure the technical reasons behind this, but these are the settings that worked "best" for me:
R9 290 - 1000 core / 1500 mem (wasn't extensively tested, about 830KH/s - might get a bit more try core up/down)
R9 290x - 920 core / 1500 mem (fairly well-tested, about 870KH/s).

As a reference of how much of a difference it makes, lowering OR increasing the core by 20MHz makes the speed go to 800-ish. Oh, and increasing core from 920 to 1150 (keeping everything else the same) drops it down to about 650-ish... lol.

So there you have it.

For games, extra clocks give more performance... not always with scrypt :-) .
 
I know when I was folding the memory speed made Way more difference than core clock. Never was a bad side effect though. I could under clock the core though and make up the difference in extreme memory clocks.

Sent from my SPH-D710 using Tapatalk
 
Yeah, usually more is always better, but this seems to be like a parabolic curve... Any lower OR higher than 920 results in about the same magnitude of decrease in hashing speed. Might have to do with the way algorithm works, but it's definitely the first time I've seen such a peculiar case.

I think it's similar to the 7950/7970s but getting this "ratio" wrong there didn't decrease speed by nearly as much.
 
Well I'm probably belching out my ass on this one but it reminds me of early PS2 emulation because the EEs two threads had to be in absolute sync.

Seems like the software is optimized to the point where every so many operations we have a fixed number of memory operations and your getting thread stalls when the clocks are out of that particular sync. Whereas if GCN had been optimized to it or it to GCN maybe not such a problem. I'd have no clue this is exactly why I never even bother looking at this shit.

I mentored a programmer years ago who went on to work on breaking some of the PSPs encryption schemes. He wanted me to look into something once and I'm like dude I know I might have left some BS impression on you 10 years ago but I'd be lost in your work.

Sent from my SPH-D710 using Tapatalk
 
You must log in or register to reply here.
Back
Top