Is Maxwell for soon ? Yes and no - Hardware.fr

[LostHawk]

So someone claiming THEY KNOW what and where is just pulling his info out their bumhole.

I don't know if sites like extremetech.com are showing fake graphics or not but before reading this whole thread, that's where I saw the AMD Out-of-stock/Mining relation first and it looked like a good theory for me:
http://www.extremetech.com/computing...adeon-shortage

But "nobody can know how much AMD Gpus are being bought for the sole purpose of mining": I agree. There is no exact science to prove all this... It's a theory.

/peace

[LordEC911]

I don't know if sites like extremetech.com are showing fake graphics or not but before reading this whole thread, that's where I saw the AMD Out-of-stock/Mining relation first and it looked like a good theory for me:
http://www.extremetech.com/computing...adeon-shortage

But "nobody can know how much AMD Gpus are being bought for the sole purpose of mining": I agree. There is no exact science to prove all this... It's a theory.

/peace

Just to give some perspective, from 4days ago.

Here's some numbers for you then. Data from litecoinpool.org and litecoin.info. Since Nov. 28, which is about when the price of Litecoin spiked, the total network hashrate has gone from about 45 GH/sec to about 90 GH/s. Given that an average 280X can run about 650-700 kH/s and the 290 runs about 800 kH/s, that's an increase in installed capacity of about 55,000 290's.

Litecoin network is at 103GH/s now.
And who is to say that some miners don't overlap into the gaming category?

[Tim]

This is a Maxwell thread people.

[Iconyu]

This is a Maxwell thread people.

I take it you didn't actually read the first post claiming that Nvidia didn't have to release Maxwell because the 700 series was selling so well vs the R9, no mention was made of Nvidia not wanting to use Risk wafers because of wafer prices and yields.

[AliG]

I take it you didn't actually read the first post claiming that Nvidia didn't have to release Maxwell because the 700 series was selling so well vs the R9, no mention was made of Nvidia not wanting to use Risk wafers because of wafer prices and yields.

I agree with Tim, even if that's true we should keep this discussion about Maxwell, regardless of whether the original post showed bias. A bit of self policing goes a long way towards a news section that is truly just news

[Olivon]

The original post has no bias at all. Hardware.fr is a respected and fair site. If you're not agree, DWI.

Now, Maxwell talk only on this thread. Thanks all.

[Heinz68]

Would have worked much better if the OP was about Maxwell only. See you later, next talk when Maxwell is ready.

[PedantOne]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

[kromosto]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

TY i missed your previous explanation on this. Now my expectations about 28nm Maxwell's are on track again.

[Olivon]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

You mean GM104 ? What about die size ?

[GoldenTiger]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

Interesting, thanks for this info.

[Tim]

Would have worked much better if the OP was about Maxwell only. See you later, next talk when Maxwell is ready.

Maxwell is ready.

[PedantOne]

of course IS ready, first QS ES boards were ready and fully working this SUMMER! Now we are in pre mass production phase.

[Tim]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

Still February, or even earlier?

of course IS ready, first QS ES boards were ready and fully working this SUMMER! Now we are in pre mass production phase.

Exactly, it seems some people find it hard to believe.

[iMacmatician]

i have seen this week working maxwell es sample card, desktop card with higher mainstream chip. Launch is relatively soon, sooner then everyone expected, including me.

As i said many months ago here, first maxwell is 28nm, has better performance then kepler, BUT not too much. Most important INITIAL 28NM maxwell thing is power consumption a efficiency.

Do you know the number of CCs, clock speeds, and memory bus width?

[zalbard]

Maxwell is ready.

What flavor exactly? Something like GM104 that is barely faster than GK110? If so, honestly not too exciting...

And I am pretty sure the real, "big" Maxwell is not ready yet.

[GoldenTiger]

What flavor exactly? Something like GM104 that is barely faster than GK110? If so, honestly not too exciting...

And I am pretty sure the real, "big" Maxwell is not ready yet.

Yet if it is better ocing or performs better at 4k or cheaper than 780ti by a good amount then it is exciting enough. Plus if it works nicely at 28nm.... can't wait to see what a full chip at 20 could do then.

[tajoh111]

Yet if it is better ocing or performs better at 4k or cheaper than 780ti by a good amount then it is exciting enough. Plus if it works nicely at 28nm.... can't wait to see what a full chip at 20 could do then.

I am predicting something a little more boring. 2304 shaders with performance somewhere around 30-40% better than a gtx 770 at around a die size around hawaii.

Being efficient with the die usage, i think they can pack the same performance as a gtx SLI 760 performance - 10% for lower clocks, similar to how AMD packed the performance of 7870 CF onto a 290x.

It depends how much efficiency the new architecture buy them for the power envelope.

I suspect it's going to be slower than 290x when the 290x is in uber mode but more power efficient and with better overclocking headroom. Nvidia is probably going to underclock again to leave gtx 780 ti owners happy.

I suspect its probably cheaper to make a 450mm2 chip on 28nm for nvidia, than a 20nm 300mm2 chip, so I wouldn't be surprised to see a big gm104 at 28nm.

[GoldenTiger]

Yeah, I picture a large 28nm gm104. Guess we'll know soon, your speculation is almost definitely closer to what we'll see. I just am playing the blind optimist :P. Although if priced right it could still be interesting if efficiency is good enough to provide high clocks and/or oc ability as I was saying. I had the same idea on cost for 20nm...

[Tim]

What flavor exactly? Something like GM104 that is barely faster than GK110? If so, honestly not too exciting...

And I am pretty sure the real, "big" Maxwell is not ready yet.

It doesn't have to be much faster, it just has to be faster then 780Ti, as far as I'm aware AMD doesn't have anything until 20nm? That's a long time. When mid range Maxwell on 28nm beats 780Ti with less power consumption and higher efficiency, it's exciting all right.

Obviously it's not a big Maxwell, we won't see that until 20nm I bet.

[stuffme]

If my memory serves me right AMD has always had the upper hand when it comes to power efficiency, but now things have changed. I wonder how AMD is going to make GCN based chips more efficient when we now have the second generation and they have as ty efficiency as the previous one. I hope AMD can make GCN less power hungry so the competition would tighten even more.

[iMacmatician]

It doesn't have to be much faster, it just has to be faster then 780Ti, as far as I'm aware

Well it doesn't have to be faster than the 780 Ti. Since that was released last month I guess that any Q1 2014 Maxwell would be slightly over 780-level performance at most.

[PedantOne]

What flavor exactly? Something like GM104 that is barely faster than GK110? If so, honestly not too exciting...

And I am pretty sure the real, "big" Maxwell is not ready yet.

of course, because 20nm proces is not ready yet

[tajoh111]

of course, because 20nm proces is not ready yet

Any idea of the size increase over gk104?

[snoid_zero]

Do you know the number of CCs, clock speeds, and memory bus width?

That is too much to ask at this stage.