Discussion in 'News' started by btarunr, Oct 16, 2013.
Anyway, here's an compiled comparison of another NDA-breaking run...
I don't show PCIe 3.0 either unless I force enable it using the file on the following link, regardless of what driver I have installed, you might wanna give it a try David:
Dunno if it would have made much of a difference on a single card setup though, I can tell you this from personal experience, jumping from PCIe 3.0 in my Haswell rig to PCIe 2.0 on my SB-E setup didn't make that much of a difference, yes I'm comparing rendering a little over 6MP per frame on my setup vs over 10MP on a 4K monitor, but what I'm trying to say is even at such high resolutions, it doesn't seem like dual Titans are constrained by PCIe ver 2.0.
And yes, IB-E natively supports PCIe 3.0 on X79 , dunno if hardware sites who published these benchmarks were using that or IB-E, not much information has been made public by them...
Well, sometimes full 16x per card in dual setup does wonders (I've just recently seen a R9 280X crossfire review w/ 8x and 16x results and... it's there). But I do agree that on single card setups, it means next to squat. And I also know that most people claiming it would make a difference are sore GTX 780 users...
Wadda ya think?
So, if I understand the graphic correctly, running a XFire 290x gets you between 1.8 times and 2 times the performance of running a single. Is that right? And did they do anything with something like Skyrim with tons of VRAM requirement to see if XFire handles well over a 3G point?
You mean a unified memory pool between cards ? If so then no. AFR doesn't support it at present. The little used SuperTiling implementation I think did support unified memory - but since SuperTiling has its own issues (geometry, overdraw overhead, no OGL support) I think it may have gone the way of the dodo.
Like it's been said... we're still far off from unified VRAM on multi-GPU setups.
But on the bright side, I think AMD mentioned something about 6GB versions of R9 290X.
Which begs the question, have they "fixed" the uneven VRAM placement on their cards? Look at GTX 660 Ti, nVidia uses 2GB as default... and that doesn't exactly fit with a 192bit bus (there's also 3GB version). AMD never (/could do) did this before, it was either double or nothing.
Detail information about XFX R9-290x
with reference cooler ? not bad..
The final runs were done with 100% Fan Speed. Furmark and 3DMark 11 are fine
He try'n to find the OC limit in 3DMark 11 it seams and those runs are fine.
It looks like the core will be similar to the OC of Tahiti on reference but then again if you look at the settings you want someone better at running benchmarks to release #'s and screenshot settings
Some other leaked review scores
source (higher resolution benches on there)
It's looking very, very strong.
R9 290X Lightning, anyone?
I'll wait for first hand review from Wizzard before i make anymore comments
Like the 98% of this forums users Me included.
No thanks.. If it is anything like the GTX 780 Lightning anyway...
Hmm... Those charts show it weaker in 2 of the three compared to Titan. No comment until I see official reviews....
Apart from saying I would have no interest in a 290X Lightning as MSI have been doing sub par on that front. 7970 with poor connectivity - read W1zz's review and the 780 Lightning with shit voltage locks to keep NV happy and no workaround. I think the Matrix cards are looking like better bets these days.
Wizz's review, with respect, missed out on A LOT for that card simply due to the canned nature of his review process. It is honestly not made for these types of cards so it misses out on those angles.
The LN2 bios I don't think he even tried to use and ships with the same power limit as the stock bios rendering it useless. When you use the 300% bios MSI released, there are power use problems that, when cranked to 300% show ~250% power use out of the gate (though power meters show same consumption). That said without a third part bios, its less useful than the Classified or HOF.
Here is our review on it. We are only one of two sites that actually mentioned the power use issue...(and the other one blamed it on MSI AB, not the bios) and quite frankly the only site that gave it the rating it deserved.
I have been working with MSI for almost 8 weeks now trying to get a proper bios out but nada. For that card, look for 3rd party bios for sure...
Friday or next Monday?
As you said we have to look for 3rd party bioses, I reckon Skynet's bios gives full voltage and power limit control, right?
He's good at doing that, I'm using his 400%+ power limit bios for my GPUs.
SKynet's are hit and miss. I went through some testing with the guy...
The last one I tried with him had the 300% limit (= 640W) but the voltage only allowed it to go to 1.21v. Does he have newer ones out that allow you to use at least 1.25v (what you can sort of use from the factory)?
I don't know much about the actual physical components but so far from all I've seen (mostly over at OCN) the 1.21v is a limit that must be addressed via software that reads from the NCP4026 thingy.
The actual drivers are the problem. The Afterburner soft mod bypasses the driver and reads from the NCP2046 magic box. It's the direct reading from there that allows the voltage change - that's why it doesn't work on EVGA's Precision X (apparently). Although Precision is based on AB it is not the same.
I don't think a BIOS can change it past 1.21 unless the drivers are modded too? But I could be totally wrong as I technically know nothing about electronics
All I've found is up to 1.212v, thought for a moment he had it fully unlocked but the post specified unlocked "up to 1.212v".
Have you tried asking MSI the usual unlocked (non public) version of afterburner?
TBH a part of this big fail is to be given to nvidia and its useless limits, they should just stick to limit reference cards, annoying pricks they are.
I think there's an MSI AB command that allows 1.325v, check OCN forum
(Er, if that was about 780 / Titan.)
Yes, that's what I was talking about above. If you have the NCP4206 chip (which can be found through a DOS command) you can insert two lines in the AB profile under settings. But that allows up to 1.3v afaik. A LLC mod allows an extra .025v.
But we're hijacking a thread here - sorry folks.
But ,here's hoping no such hijinks are required to get the best out of the 290X but my gut is saying that Uber mode means they have some TDP shit going down. They all seem to brag these days about being power efficient. Irrelevant for enthusiast cards tbh.
Right, which, does not work with Lightning? I recall trying this on the reference 780 and it didn't work. Not sure on the 780 Lightning though. To be honest, not too worried about it as I have a version of MSI AB that goes up a lot higher than that (@ Radrok).
My goal is to get the 300% bios working with the +100mv that MSI gives you out of the gate. I don't care if it comes from MSI (though it damn well better!!!) or 3rd party, but I haven't seen anything over 1.21v myself from Skyn3t without adding those registry entries.
Its not the driver it seems. As I can, with the right bios and MSI AB, run WELL past 1.35v.
Separate names with a comma.