NVIDIA Announces The GeForce GTX 1060, Linux Tests Happening

[sonnet]

GTX 1060: 1280 ALUs * 1.5GHz = 1900
RX 480: 2304 ALUs * 1.2GHz = 2700

Equal performance in Windows would mean GTX 1060 has 2700/1900=1.4 IPC advantage over RX 480.

Well rx 480 has about 55% performance compared to a 1080 . New drivers and fix should reduce the gap and bring the rx480 to deliver 60% of the 1080.
The 1060 has hald the cuda cores of the 1080. So it should deliver at least 50% the performances of the 1080. But given it also has more bandwith and rops per cuda core,
it should reasonably perform close to the rx 480

[dungeon]

Bleah... two posts eaten, third post auto-moderated. I bet this post goes through just fine since it has no useful content...

EDIT - yep

Ha, ha, i suspect that auto-moderator is extreme intelegent, as he eats your posts but passes the following:

I pity the morons who got out and bought a RX 480 from AMD. Didn't you idiots learn that AMD has fallen behind badly in the last few years in both CPUs and GPUs? Wasn't it obvious that nvidia will have launched the 1060 in a few weeks to counter the 'threat' from AMD and that the 1060 will shit on the RX 480? I guess ideology trumps reality for many of you...

Master5000, i guess you prefer GTA over Europa Universalis?

[bridgman]

Look at the block diagram for a compute unit, you can clearly see the execution path. There can be at least 4 in flight simd operations. And it certainly does look like they can be issued out of order. Several people here said that's not the case though, but it sure does look like it can.

When you say "4 in flight SIMD operations" are you talking about the 4 SIMDs in a CU ? If so yes there can be 4 operations going on in parallel, but each SIMD is executing instructions from a different instruction stream / program counter.

[bridgman]

Woo hoo, auto-moderated again.

[dungeon]

Woo hoo, auto-moderated again.

Just wait, it will appear tommorow... Civilization is banned here, only Saints Row works

edit: That reminds me something to only work is not enough, Virtual Programming recently released Saints Row 2, 3, 4(+1)... knowing VP's CPU cap i guess all these needs amdgpu-pro threadedGL profiles

[duby229]

They can not execute other instructions from the same instruction stream (the usual definition of out-of-order processing), but they can switch to another thread on the next clock and execute from that instruction stream instead. Each SIMD has 10 program counters associated with it (10 threads) for a total of 40 threads per CU.

What makes the terminology tricky is that the other thread may be executing the same shader program but with different data and a different program counter, but IMO that does not count as "out of order processing"... "block multithreading" is probably a good description.

Thank you, now the block diagram makes perfect sense to me.

[duby229]

When you say "4 in flight SIMD operations" are you talking about the 4 SIMDs in a CU ? If so yes there can be 4 operations going on in parallel, but each SIMD is executing instructions from a different instruction stream / program counter.

Ok, yeah, that makes perfect sense. The block diagram makes perfect sense to me now, thank you.

[Linuxhippy]

Well, since NVidia is clearly the GPU designer with the worst OSS drivers today (and the least amount of OSS commitment) the GTX 1060 is no option at all for true Linux/OSS users. In 3 years the binary drivery will be OOL the GTK1060 series, while you can count on the OSS support the RX480 will receice by paid AMD developers

[duby229]

Bleah... two posts eaten, third post auto-moderated. I bet this post goes through just fine since it has no useful content...

EDIT - yep

yeah, I replied to you twice and they got queued.

[vivnet]

Well rx 480 has about 55% performance compared to a 1080 . New drivers and fix should reduce the gap and bring the rx480 to deliver 60% of the 1080.
The 1060 has hald the cuda cores of the 1080. So it should deliver at least 50% the performances of the 1080. But given it also has more bandwith and rops per cuda core,
it should reasonably perform close to the rx 480

I fully expect it to outperform AMD's comparable hardware while being more efficient and having better driver support (given that marketing image is indeed from Nvidia).

So the marketing scheme is to make idiots see longer bars. That doesn't mean an intelligent person can't see the graph and interpret the numbers correctly.

Nvidia has no reason to lie; they've got them on the ropes and everyone knows it. I'm just waiting for that random haymaker to come out and clean it all up, if it hasn't been thrown already.