AMD Vega 10, Vega 11, Vega 12 and Vega 20 Rumors and Discussion
Bondrewd
Veteran
It's both Vega and taming the uncore.
D
Deleted member 13524
Guest
The timing seems to be in tune with Q4 2017 availability, which is what has been promised for a while.
What does leave me depressed is that HP seems to be pairing up Ryzen Mobile with single-channel DDR4 by default, which will sabotage its performance greatly.
It's Carrizo all over again.
Bondrewd
Veteran
KBL-R model of the same laptop also has single-channel memory.
It's HP being HP and nothing else.
The problem is it's not just HP, it's pretty much all but maybe the "gaming laptops" that for some bizarre reason default to single channel RAM, be it on a stick or soldered on motherboard.
(or at least that's how it feels like. My Lenovo ThinkPad came with 1x4 GB RAM out of box, and due pricing it's now running 12GB with 1x4GB+1x8GB (are the memory controllers now good enough to run dual channel mode even if they don't have same amount of RAM?)
Those Frametimes do not look as if they were taken from even the same game, let alone supposedly comparable run/drive throughs. What's wrong with this game?
Please, don't troll the thread with needless pie throwing. There's nothing constructive to be found in that. Also, this is not a NV vs AMD discussion in the first place.
Anyway, Techpowerup mentions that ASUS Strix vega card now has clock speed information. Very slight bump over stock aircooled card on core, nothing on RAM.
Maybe this means launch is finally on the horizon? One can only hope! "Early september" feels like such a long time ago...
Anyway, Techpowerup mentions that ASUS Strix vega card now has clock speed information. Very slight bump over stock aircooled card on core, nothing on RAM.
Maybe this means launch is finally on the horizon? One can only hope! "Early september" feels like such a long time ago...
I am aware of present mon and it's capturing of each frame presented. It's the same values that you can get with Fraps where it's working. Yet, those curves look very irregular to me, as I said. Not even between 1080 and 1080 Ti there seems to be a pattern even when they share they same inherent strengths and weaknesses.
Anarchist4000
Veteran
No 1080p results shown in those runs either. Which was where Vega was really out in front originally and should still be competitive as I pointed out before. This doesn't change anything, besides presenting some really strange results. Nor does the game seem to bringing any of the new features into play.
I asked Kyle about those results on the first benchmark because the 1080ti specifically had really weird frametimes. No correlation to 1080 and highly variable. Possible the game is just unstable at higher FPS on any card or was close to a CPU bottleneck and the Ti crossed that threshold.
Doesn't look like CPU bottleneck IMO. That weird up-and-down zigzag the pastel blue line is doing towards the end of the graph, it bears no resemblance whatsoever to what the other lines are doing. It looks more like massive CPU contention, like windows decided to run some disk indexing or malware scanning in the background or whatever right there. It should have triggered any person running that benchmark, like, can this be right? Does this make sense?! Because, no. No it doesn't.
So the "argument" is "founded" on Vega being "competitive" at one resolution in one game that could be "unstable at higher FPS" while performing like a regular 1080 at higher resolutions with updated drivers. Great argument.
Anarchist4000
Veteran
No, the argument is that once DSBR and some primitive work on the front end kick in they should be competitive. Allowing for the usual back and forth on some titles. Difference being culling mechanisms reducing the effective workload. In this case, in at least one test and likely without those features, it is already competitive. Following the release of Nvidia's magic driver with "up to 25%" improvement. And these tests are only looking at averages which seem unstable at best. The Vega results seem much more stable, but around frame 8800 for example even Vega gets a huge uplift that wouldn't seem to align with other cards. The Ti taking a dive towards the end, and nothing in general lining up well. The initial benchmarks had really tight frame times for Vega. In the original benchmarks, specifically at 4k, Vega had lower averages and higher minimums than the Ti. So there would be another argument for subjectively competitive. It's possible to tune for FPS over latency and that may be the case here.
DavidGraham
Veteran
If you are referencing the initial ComputerBase testing, then again these were with old bugged drivers, so please just STOP referencing them to prove your shaky points, you are dealing with a new reality right now. If your argument that Vega will catch Pascal when drivers are bugged then there is no need to continue the discussion.
And If you are referencing the original HOCP test, that was done on a different set of drivers for both vendors, and with a different game version, as well as with a DYNAMIC resolution setting: ON, so the comparison is not valid due to cards variably dropping resolution. The new HOCP results are the latest in everything: latest drivers from both teams, latest patch for the game, and with a fixed resolution setting. The new ones are the most valid even compared to ComputerBase.
Vague statements with vague names and numbers are not the correct way to prove a point in the face of hard numbers and evidence, Vagueness here is just wishful thinking wearing a mask.
Vega frame times look as much stable as the 1080 in HOCP testing.
Correcting a performance issue that made Ultra High end cards act like mid range ones is not driver magic, it's just getting things to their original level.
Anarchist4000
Veteran
What shaky points? The ones that were validated and left your argument flat on it's face? Anyone that looked at the figures could see this coming. I've never said drivers are needed to catch Pascal, just that certain features clearly aren't enabled. That's a rather simple fact. Forza 7 gave competitive performance even without the drivers.
So go and get some figures that support your argument. Empirical data was presented, I referenced that with simple facts and benchmarks seem to support my position. Hardly my fault a collection of marketing guys don't like the results. Frankly they should have seen it coming and now Nvidia is in damage control mode rushing out drivers to remedy the situation. I'd include the 1080 results, but they weren't included in the retesting.
So much for game ready drivers then. With multiple quick releases to fix major regressions with erratic performance that your competitor lacks. The whole magic deal was based on games not seeing huge increases from drivers. Here's a game with big increases from drivers, despite being a low level API where only scheduling should be an issue. The real discussion here is why the results are so erratic, not why this game is so challenging for Nvidia hardware. So if you just want to throw pies, please go elsewhere.