|
Skuto posted:I'm sure most published benchmarks just outright disagree with you. Sounds interesting, find them and post the relevant graphs then.
|
#
?
Aug 29, 2015 21:15
|
|
|
|
| # ? Apr 28, 2024 05:50 |
|
|
Got my HOF into my system on my lunch break. Got a 16,470 on FireStrike 1.1. http://www.3dmark.com/3dm/8387532 My fans are doing a weird thing where they will rev up to speed and then rev back down like 400 RPM every few seconds. Using any kind of set fan speed control.  I don't know what those high spikes are, but those down spikes are the fan actually revving down from 3k RPM to 2.6k RPM.
|
|
#
?
Aug 29, 2015 22:32
|
|
|
Truga posted:Ouch. Did you not cool them enough, or did you just set a too high voltage? Cooling problem. I think the cooling pads for the VRMs were crap because the night before they died they hit almost 120 C while the GPU was only at 75.
|
|
#
?
Aug 29, 2015 22:36
|
|
|
Kazinsal posted:Cooling problem. I think the cooling pads for the VRMs were crap because the night before they died they hit almost 120 C while the GPU was only at 75. Appears to be a common issue with reference PCB for the 290/X. Thinking of moving to watercooling, but then the RX 400 series is rather close and I want to see if it's worth it just to jump ship to a new process.
|
|
#
?
Aug 29, 2015 22:56
|
|
|
Decided to do some light overclocking on my 670mx, and shocker, +135 was the most it even let's me do to the card but 135/500 gave me a fairly modest 5 fps boost, hardly worth much but did make me feel infinitely better about overclocking my 980 when I get it. Dispelled some things just doing this, thought I was getting tears but it was just vsync fuckery. Temps never broke 71 degrees Celsius. Feels good man.
|
|
#
?
Aug 29, 2015 23:31
|
|
|
My max boost so far on my 980 Ti HOF 1474 core. I can do 120% power target with Afterburner, but it starts limiting itself before it even hits temp or power limits.
|
|
#
?
Aug 29, 2015 23:35
|
|
|
sauer kraut posted:Of course it was that lovely Gigabyte Mini that's perpetually on super sale.  I have a Gigabyte GTX 680 Windforce 3X I got used and am super happy with, but they definitely cut any corner they can get away with. Anandtech caught them faking the voltage readings on their motherboards to make it look like the voltages sagged less and were more stable than they really were. Zotac's motherboards also tend to burn out their power delivery components just outside the warranty period.
|
|
#
?
Aug 29, 2015 23:38
|
|
|
Alereon posted:I thought nVidia Greenlight was supposed to protect us from Gigabyte and Zotac? Wow. Got a link to that AT article?
|
|
#
?
Aug 30, 2015 00:07
|
|
|
xthetenth posted:Wow. Got a link to that AT article?
|
|
#
?
Aug 30, 2015 00:34
|
|
|
Kazinsal posted:An R9 290 with aftermarket cooling has decent headroom. Unless you're me. Then you think it has more headroom than it has and end up frying the VRMs. Wonder if this happened to my overclocked 6970 a few years back - not a crazy OC but it died & shut my PC down mid game. Other cards worked fine, rest of the system tested 100% stable, no issues. Got it replaced under warranty but never had that happen before.
|
|
#
?
Aug 30, 2015 01:05
|
|
|
Skuto posted:Whoa, I don't think you can just equate the R9 290 and R9 390 like that. Most obvious difference is the latter having 20% higher memory bandwidth outright. There's process advances/better binning too. Because of the price hike the R9 290 is still in the picture for price/perf but it's just an inferior card compared to R9 390s. Clock for clock the 290 and 390 are completely identical. There has been nothing to show that the "process/binning" has changed. A 390 is a 290 with higher stock clocks and 8GB of slightly faster VRAM (that you can run the 290's VRAM at), nothing more.
|
|
#
?
Aug 30, 2015 01:26
|
|
|
BurritoJustice posted:Clock for clock the 290 and 390 are completely identical. There has been nothing to show that the "process/binning" has changed. A 390 is a 290 with higher stock clocks and 8GB of slightly faster VRAM (that you can run the 290's VRAM at), nothing more. And a stock HSF that doesn't choke it to death running benchmarks (though of course that was only really an issue if you were unfortunate enough to buy a 290(X) with a stock HSF, rather than the very capable AIB ones). When you can get a EVGA b-stock 980 for only $40 more than a 390 (and $40 LESS than a 390X), I have a hard time understanding why you'd even consider AMD's offerings.
|
|
#
?
Aug 30, 2015 01:40
|
|
|
I think the removal of the stock model has made a huge difference to public perceptions. Now when people are using a "reference 390" in their benchmarks, it is a MSI card with a fantastic cooler and a hefty stock clock boost. Imagine if every benchmark comparing against a 970 used a MSI card at 1400MHz.
|
|
#
?
Aug 30, 2015 04:07
|
|
|
So what's the verdict on those DX12 benchmarks... Is AMD really that much faster with DX12? Or are these purely synthetics and will never translate to in game FPS improvements?
|
|
#
?
Aug 30, 2015 05:23
|
|
|
Deathreaper posted:So what's the verdict on those DX12 benchmarks... Is AMD really that much faster with DX12? Or are these purely synthetics and will never translate to in game FPS improvements? There's no verdict. It is one game. Edit: Okay, maybe that's a bit terse, the AMD stuff almost certainly get a good boost since DX12 is essentially an even more refined Mantle, but still, no real verdict until we start seeing real games. Zero VGS fucked around with this message at 05:37 on Aug 30, 2015 |
|
#
?
Aug 30, 2015 05:34
|
|
|
Deathreaper posted:So what's the verdict on those DX12 benchmarks... Is AMD really that much faster with DX12? Or are these purely synthetics and will never translate to in game FPS improvements? AMD's DX11 performance is terrible in CPU heavy situations compared to Nvidias. If you look at the DX12 results for say, the 980ti and the FuryX they are basically neck and neck in a super CPU heavy scenario. If you look at the DX11 performance for a FuryX in the same situation it is loving terrible, the touted "80%+ gains!!" are just bringing it into line.
|
|
#
?
Aug 30, 2015 05:34
|
|
|
There's some mumbo jumbo about how radeon has quite a bit more compute power over a similar nvidia card (a 980Ti has 5.63 TFlops, same as a *reference* 290x), which translates to more async shader power, which means a decent dx12 engine will run better. Meanwhile, nvidia allegedly has an uarch that seems to be better suited to dx11 (more likely, their drivers just aren't trash), so they get the performance advantage there. The differences here are minimal, though, and a lot of the performance difference on dx11 is what ^^^^ is saying. Truga fucked around with this message at 06:00 on Aug 30, 2015 |
|
#
?
Aug 30, 2015 05:57
|
|
Deathreaper posted:So what's the verdict on those DX12 benchmarks... Is AMD really that much faster with DX12? Or are these purely synthetics and will never translate to in game FPS improvements? Those benchmarks were not comparing Nvidia to AMD directly, they were comparing Nvidia DX11 to Nvidia DX12 and AMD DX11 to AMD DX12. AMD cards gain a big jump in performance going from DX11 to DX12, Nvidia cards gain little but they were already ahead by miles so this just brings AMD into rough parity with Nvidia. On top of that you have to consider that it's going to take a few years for DX12 to become mainstream, we will be playing DX11 games for quite a while.
|
|
|
#
?
Aug 30, 2015 06:02
|
|
|
Truga posted:There's some mumbo jumbo about how radeon has quite a bit more compute power over a similar nvidia card, which translates to more async shader power, which means a decent dx12 engine will run better. Meanwhile, nvidia allegedly has an uarch that seems to be better suited to dx11 (more likely, their drivers just aren't trash), so they get the performance advantage there. if Pascal even has a 10% increase in DX12 over comparable maxwell cards, AMD has their work cut out for them.
|
|
#
?
Aug 30, 2015 06:04
|
|
|
BurritoJustice posted:Clock for clock the 290 and 390 are completely identical. There has been nothing to show that the "process/binning" has changed. A 390 is a 290 with higher stock clocks and 8GB of slightly faster VRAM (that you can run the 290's VRAM at), nothing more. Silicon that has had process improvements (that allow for better binning) is going to be clock for clock identical. Not sure what else you expected?! The whole point is to up the clockspeed (or reduce the voltage).
|
|
#
?
Aug 30, 2015 07:02
|
|
|
Skuto posted:Silicon that has had process improvements (that allow for better binning) is going to be clock for clock identical. Not sure what else you expected?! The whole point is to up the clockspeed (or reduce the voltage). They are built on the exact same process. Anything else was just reddit speculation. There has been nothing showing consistently higher overclocks, the clocks are only higher comparing factory overclocks on the 390s versus the 290 reference clocks. They still all go to the same average 1100-1150 range, with 1200 for super good chips. You can take an 8GB 290x and flash a 390x vBIOS on it and you've got a 390x. AMD did do some funny stuff with drivers to make the 390 seem better clock for clock at launch by branching the drivers and only adding improvements when a 300 series card was detected. They've since merged the drivers.
|
|
#
?
Aug 30, 2015 07:33
|
|
|
So, originally backplates were designed to help absorb heat from the GPU/VRAM as sort of a giant, flat heatsink right? But now we just use them as structural support for the PCBs? The backplate on my 980Ti only touches the card where the screw holes are. Everything else is lifted away from the PCB by like 2-3mm.
|
|
#
?
Aug 30, 2015 07:36
|
|
|
BurritoJustice posted:They are built on the exact same process. Anything else was just reddit speculation. To be fair for example Anandtech.com claimed the same, e.g 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... " Which at the end refers to voltage binning specifically.
|
|
#
?
Aug 30, 2015 08:36
|
|
|
Binning silicon and fabbing on a modified process are entirely different things. AMD binning chips to support the higher clocks of the reference boards is reasonable and expected, considering that binning has always been used for non-reference 200 series cards. Every GPU you buy has been binned to some capacity, so tightening the requirements for lower power draw (clocks at volts) chips is easy, it just means that AMD is throwing more out. There has been no evidence of a changed process tech at all.
|
|
#
?
Aug 30, 2015 09:00
|
|
|
Why is the AMD reference blower design so much worse than Nvidia's?
|
|
#
?
Aug 30, 2015 13:16
|
|
|
I've always been of the opinion that AMD is actually choking the poo poo out of their blower-style fans. Look at the r9 290X's bracket and compare it to a Titan X's. It's like trying to push a bouncy house's worth of air through a straw. There's a reason why nature favors hexagonal shapes. Hexagonal packing is the most-resource-efficient shape to cover a given area. That they've only subdivided it further into triangles is of little consequence. (I can't explain why the original Titan and Black also had identical brackets to the 290X but still weren't as poo poo. This is only a conclusion I came to fairly recently. But you can see the progression between the Black and the X in the Z's triangular-shaped holes.) SwissArmyDruid fucked around with this message at 13:35 on Aug 30, 2015 |
|
#
?
Aug 30, 2015 13:28
|
|
|
Does anyone know if a Radeon R9 390 has HDMI 2.0? I have a 290 and just bought a 4K TV but can't output 4K higher than 30FPS and I believe it's due to this card only having HDMI 1.4. Having a surprising amount of trouble finding the technical specs for the ports on each card.
|
|
#
?
Aug 30, 2015 13:39
|
|
|
You need a nvidia 900 series to get hdmi 2.0 at the moment.
|
|
#
?
Aug 30, 2015 13:50
|
|
|
Josh Lyman posted:Why is the AMD reference blower design so much worse than Nvidia's? http://www.anandtech.com/show/6760/nvidias-geforce-gtx-titan-part-1/3 The Titan cooler is quite expensive, the standard budget black box blowers from the partners are not better than AMD's.
|
|
#
?
Aug 30, 2015 13:50
|
|
|
Anime Schoolgirl posted:pre-maxwell 2 is also exclusively serial with maxwell 2 having the barest support for parallel processing, which is something Pascal may change since it's a **brand new!!** architecture No. Pascal is going to be on a ~1.5 node shrink with a new arch and some other shiny toys. Arctic Islands is going to be on a similar shrink with hopefully a non-lopsided architecture that can take advantage of its toys. Anything less than a 50% increase will be shameful. Granted they will likely do the same half-generation bullshit again so that'll be relative to the 980, but a 10% increase would be calamitous.
|
|
#
?
Aug 30, 2015 16:11
|
|
|
xthetenth posted:No. Pascal is going to be on a ~1.5 node shrink with a new arch and some other shiny toys. Arctic Islands is going to be on a similar shrink with hopefully a non-lopsided architecture that can take advantage of its toys. Anything less than a 50% increase will be shameful. Granted they will likely do the same half-generation bullshit again so that'll be relative to the 980, but a 10% increase would be calamitous. I don't remember 50% leaps between generations, unless we're counting Tesla > Fermi, usually 30% is considered a success.
|
|
#
?
Aug 30, 2015 18:17
|
|
|
780ti vs 980ti is close isn't it
|
|
#
?
Aug 30, 2015 20:31
|
|
|
Don Lapre posted:780ti vs 980ti is close isn't it
|
|
#
?
Aug 30, 2015 20:35
|
|
|
FaustianQ posted:I don't remember 50% leaps between generations, unless we're counting Tesla > Fermi, usually 30% is considered a success. TSMC has said their 16FF+ process has 40% better performance with power draw constant compared to their 28HP process. Remember the 20/22nm node didn't work out for non-mobile chips so the next generation of GPUs will benefit from two process shrinks. The hope is therefore very much for a large performance jump.
|
|
#
?
Aug 30, 2015 20:36
|
|
|
Anime Schoolgirl posted:More like 780ti vs 980, but the 980ti is not the same GPU as the 980, whereas the 780ti was a better-binned 780. No it wasn't, it was a full version of the chip in the Titan. FaustianQ posted:I don't remember 50% leaps between generations, unless we're counting Tesla > Fermi, usually 30% is considered a success. I don't remember many generations being the same time as going about one and a half nodes forward in process, so there's potential for something huge.
|
|
#
?
Aug 30, 2015 21:05
|
|
|
Anime Schoolgirl posted:More like 780ti vs 980, but the 980ti is not the same GPU as the 980, whereas the 780ti was a better-binned 780. 780ti was the full Kepler chip. It was more powerful than the titan. The titan black was the 780ti equal.
|
|
#
?
Aug 30, 2015 21:09
|
|
|
Don Lapre posted:780ti vs 980ti is close isn't it A fully overclocked 970 was better than my factory overclocked MSI 780ti ... subjectively speaking. Not sure about actual numbers. 980 surely beat it, and now the 980ti kind of left it in the dust. Unless you meant the 50% number then no I highly doubt it was that far apart, nor do I really expect the next gen to be for that matter in anything other than very specific benchmarks
|
|
#
?
Aug 30, 2015 21:32
|
|
|
I'm about two minds being on a Ti cycle, I've got a 780 right now and I'd rather go for the first Pascal release than wait longer.
|
|
#
?
Aug 30, 2015 22:16
|
|
|
Retrospect Kepler wasn't really that awesome, good cards but in the end AMDs offerings were better IMHO. Maxwell is also 8000 series good so it makes Kepler look even worse. Actually this is totally reminiscent of the HD 2000/3000 vs 8000 series, except Nvidia isn't going to sit on what is effectively a double refresh and allow AMD an HD 4000 coup.
|
|
#
?
Aug 30, 2015 22:27
|
|
|
|
| # ? Apr 28, 2024 05:50 |
|
|
ijyt posted:I'm about two minds being on a Ti cycle, I've got a 780 right now and I'd rather go for the first Pascal release than wait longer. To me it seems that with the split generations the Ti bracket makes no sense unless you absolutely have to have the best and are willing to work the used card market with 'old' cards. My 970 is less than a year old and sure I could get a bunch more performance, but it would come at twice the cost. If I'm going to upgrade every other half generation step, the price/performance is well on the side of the high midrange (I'm pissed that the midrange chips are now 50% more expensive though).
|
|
#
?
Aug 31, 2015 00:05
|
|