NVIDIA Pascal Thread

[happy medium]

When he says new king, does that mean faster than a new 1500$ AMD Duo card?

 

[tviceman]

that's boost clock, not manual overclocking

He didn't say for sure, but he mentioned overclocking and I highly, highly doubt it's boosting >20% above reference boost speeds. It's got to be overclocking. Also memory was overclocked from 10ghz to 11ghz, so more evidence the core was manually OC'd.

11ghz vram = 352 gb/s, up 57% over GM204.

 

[Head1985]

 

[96Firebird]

Memory was also 11Gbps, I think the 2.1GHz core clock was manually overclocked myself...

 

[happy medium]

He didn't say for sure, but he mentioned overclocking and I highly, highly doubt it's boosting >20% above reference boost speeds. It's got to be overclocking. Also memory was overclocked from 10ghz to 11ghz, so more evidence the core was manually OC'd.

11ghz vram = 352 gb/s, up 57% over GM204.

thanks for the info.:thumbsup:

 

[Mopetar]

Given the performance and specs, I guess we can now confirm that there is zero improvement in IPC from Maxwell to Pascal (on average)

Maxwell already had pretty good performance, so as long as they've shored up any of the DX12 shortcomings it should be a solid card.

If most of the cards can OC to 2.1 GHz boost, it's a worthwhile upgrade.

 

[iiiankiii]

Nvidia cards can automatically boost beyond the reference boost clock if thermal/power conditions allow. 2.1 GHz boost is super impressive but it was also on a single character render. It will be interesting to see if it can sustain that boost during normal gaming loads.

Well, I think as long as the silicone will hit 2.1Ghz with a bit of increase in power limits, voltage, and better cooling easily, I think it's pretty good. Imagine 2.3-2.4ghz on water with voltage bumping. That would be pretty sweet!!!

 

[SteveGrabowski]

I think you hope they don't give out the release date and pricing info so you can complain, and if they do give it out you'll complain all the same.

You don't think release date and price are important pieces of information when you announce a card?

 

[Hitman928]

He didn't say for sure, but he mentioned overclocking and I highly, highly doubt it's boosting >20% above reference boost speeds. It's got to be overclocking. Also memory was overclocked from 10ghz to 11ghz, so more evidence the core was manually OC'd.

11ghz vram = 352 gb/s, up 57% over GM204.

Good catch, looks like it was oc'd speeds. We'll see how the silicon lottery plays out.

 

[Cookie Monster]

The multi monitor projection thing looks interesting.

 

[Pneumothorax]

Damn my bro should've sold his Titan X while they were still worth something! Lol

 

[happy medium]

That looks like 75% faster than a gtx980 in Tomb Raider (direct x 12)

 

[iiiankiii]

Notice that they're comparing it to the GTX 980. That's a pretty good indicator that it would be priced at $549. They're trying to get GTX 980 guys to upgrade.

 

[kagui]

ok, that multi-projection
is pretty good, the kind of thing that is obvious now but nobody make before

 

[Mopetar]

The multi monitor projection thing looks interesting.

Yeah, that's actually something that I'm really interested in, because I run a 3 monitor setup.

 

[tviceman]

You don't think release date and price are important pieces of information when you announce a card?

I never said it wasn't; I said to slow your roll and let the presentation run it's course before you pull out your jump-to-conclusions board game.

 

[Saylick]

Hahaha, the banter between JHH and TAP is hilarious.

 

[tential]

Basically saying that maxwell sucks for vr

 

[Silverforce11]

Here's the analysis of data we have.

Ref card paper boost clock is 1733mhz. Claim observed boost clocks in 3dMark, 1860mhz.

~25% above 980Ti.

55% clock speed advantage vs reference 980Ti that boost to 1.2ghz.

10% less CC.

Think about a 980Ti at 1.5ghz, it will equal a 1080 @ 1860. The 1080 has a real 12% OC headroom based on this presentation.

Overall "IPC" or clock for clock per CC, performance regressed a lot. This I feel is indicative of their uarch evolved to handle graphics <-> compute switch instantly like GCN. Also it supports fine-grained preemption like GCN. There's a regression for adding these features, but in games that use it (heavy compute or VR), Pascal will fly. In many ways, it's much more future-proof.

 

[SteveGrabowski]

Notice that they're comparing it to the GTX 980. That's a pretty good indicator that it would be priced at $549. They're trying to get GTX 980 guys to upgrade.

Not too bad at $550 for a 22% gain on Titan X with 8GB VRAM.

 

[ThatBuzzkiller]

That's what I'm saying. Looks like they actually regressed per SP performance. Basically a shrunk Maxwell GM204 that clocks a lot higher while using more power.

So what if IPC regressed, it appears to be well worth it to trade it for higher clocks since it practically guarantees linear scaling ...

 

[Cookie Monster]

Guess Tom is going to get fired after the stream ends.

 

[tviceman]

I've got to hand it to Silverforce, he hit the nail on the head with regards to Pascal's gains over Maxwell. Much bigger gains in VR and in DX12.

 

[Saylick]

Basically saying that maxwell sucks for vr

Also shows how SLI scaling is really poor, if we are to assume that 1080 > 980 SLI.

 

[Spanners]

Looks like Tom is getting fired.