As for gpus, i have high hopes for 5700. It has only 4 CUs less, same memory, same gpu,same bandwith and amount of vram, i hope they dont gimp the overclocking on it.If not, its gonna be in spitting distance to the 5700XT.
Total board power should be 225W for the 5700 XT and 180W for the 5700 non-XT. We asked if VBIOS would be locked and, due to security concerns, AMD is locking down custom VBIOS modifications. We then asked if powerplay table hacks will still work this generation, but we were told that we’d have to learn new tricks to bypass the power limits. That’s a little bit of a let-down from the fringe use case of power modding, but we’re hopeful that we can find new power bypasses.
3 Lefts? Hmm, starting to seem like P&N around here.
I think that's a flaw in the slide. The card she held up on stage didn't look warped like that. Also, if they fire him twice, does that mean he gets a promotion?
N7 process is not as good as everyone hoped for?A 2560 Shader 7nm chip is only using 24% less power than a 4096 Shader 14nm chip?
I'll used announced next time, as they were announced on May 6th. If you want to nitpick, the actual release of the 5700XT (7/7) will be more than three years after the release of the 1070. Does it really matter?It's actually exactly 3 years today. The 1070 was released June 10th, 2016.
It looks like all of the processes aren't that good for AMD's GPUs (blame GF etc.)N7 process is not as good as everyone hoped for?
Its not good for GPUs, really.
Or you can just get Radeon VII for less than twice as much, pull 90 MH/s, and only take up one slot.
Why get bent out of shape with this response, then? I just pointed out the coincidence.I'll used announced next time, as they were announced on May 6th. If you want to nitpick, the actual release of the 5700XT (7/7) will be more than three years after the release of the 1070. Does it really matter?
The non-XT 5700 should get almost 60MH/s for $379. That's not bad at all compared to the $699 VII that performs little more than 50% better.
@railven
I had a lame AP CS class in high school where we learned to program in Pascal on Macs of all things. We had some stupid robot named Carel (or whatever) that couldn't turn right. Communist plot, obviously.
Wish I could go back and watch the presentation. That shroud is a bit odd in all the stills.
The funny thing is that Im more interested in rumored 4-5 TFLOP Xbox IF it has optical drive. If it won't have - forget about my money .
[USER=359189]@Glo.
Good luck.
Buy hey, look at the bright side! MS is apparently going to offers us a 12 TFlop Navi, complete with 8 core Zen2 and 1 TB NVME drive next year for $499. Too bad you can't just rip one out of that machine, right?
Maybe, or maybe not. Nvidia appears to be skipping this node completely.It looks like all of the processes aren't that good for AMD's GPUs (blame GF etc.)
Gamernexus had this to say:
So not sure if this will affect the lucrative undervolting scene.
Remember, even RX Vega 64 with its whopping memory bandwidth advantage only pulled 40-45 MH/s consistently. 5700/5700XT won't be that much better.
I can't remember a time when the specs of two GPUs in rival architectures were basically mirror versions like this. It's going to make for an extremely interesting architectural comparison.
You have always a choice. You sacrifice power, performance or die area.Add that with modern semiconductors being heavily power limited, and everyone trying to extract much out of the hardware possible, it ends up being similar in design.
Lately with lower node you get worse clocks, in order to circumvent this, you need to lax the density.10,300 M transistors, 251 mm^2: 41 M transistor/mm^2
This is getting annoying. Is AMD not putting any effort in optimizing density? That figure is far below 7 nm theoretical specification.
I'm going to give AMD the benefit of the doubt here. I wouldn't be surprised if they have to, but I'm going to side on "it's not that bad" this time. I'll kill my own crow if I'm wrong.
Also that they are using a small chip to compete, I wonder what the relative CUDA only chip area is for RTX 2070 when scaled to 7nm.It means that AMD is clocking Navi way past it's perf/w curve to match Nvidia on performance.
That is what pluses in 14nm intels node means. Skylake has worse density than broadwell, and voila, higher clocks.
Transistor spacing to account for heating at high frequency maybe?This is getting annoying. Is AMD not putting any effort in optimizing density? That figure is far below 7 nm theoretical specification.