I dont think his point was that you'd see lower clock speeds than before. Just that there's still a tradeoff. Meaning you might see a smaller improvement in clock speeds in certain cases due to prioritizing IPC.
-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
AMD Polaris architecture – GCN 4.0
- Thread starter Boomstick777
- Start date
yeh i not sure if i understand very well lol
if you look at the tdp tho and the size of the chip doesnt seem like nvidia cared much about power usage?? unless the tdp is way off
if they made the 1080 the same size chip as a Ti it would have a 350tdp, the out going Ti being 250tdp
..its a smaller chip than the 980 and specs say it uses more power
that has to be mostly down to clockspeed?
if you look at the tdp tho and the size of the chip doesnt seem like nvidia cared much about power usage?? unless the tdp is way off
if they made the 1080 the same size chip as a Ti it would have a 350tdp, the out going Ti being 250tdp
..its a smaller chip than the 980 and specs say it uses more power
that has to be mostly down to clockspeed?
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
From what i have seen the gtx1080 has 7.2 billion transistors compared with a gtx980ti at 8 billion. If you ran them at gtx980ti clocks there would only be one winner. To me that tells me there was not much improvement in architecture if any. Check out Titan Black v gtx980 to see a real architecture improvement.
Pentium 4 to Core 2.
There is never a need to increase clock speed to gain performance. In general that is done but nothing stop you designing a processor with a much higher IPC and slower clocks, and that makes much more sense to do when moving to a smaller node where you have the transistor budget to increase IPC significantly.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
Again this is the Graphics forum
. I am asking about graphics architecture. Show me Nvidia taking your advice. P4 was trash and since then the trend is usually up on clock speeds. Amd have been similar apart for the odd stupidly priced 5.0ghz chip.980 runs at pretty decent clocks compared to a Titan Black, too.
Not saying it wasn't a good architectural improvement, it obviously was, but perhaps the architectural improvements are a notable part of what allows the higher clocks in the first place?
Lower density chips can have better electrical and thermal properties which can help but clock speed is largely determined by the critical path through a processor. If processing an instruction takes X amount of time then the clock speed is simply 1 / X. If you try to do more work in a single clock using more transistors then the critical path is longer and you have to lower clock speed.
The Intel Pentium 4 is a classic example of aiming to increase clock speed with a lower IPC.
Again this is the Graphics forum
Pascal is the first time Nvidia have moved in that direction.
Makes no difference if it is a GPU or CPU, the laws of Physics apply equally to both.
Adjusting the clock speed wont tell you anything about Nvidia's design.
Nvidia changed the architecture to facilitate faster clock speeds with equal or lower IPC so it is completely irrelevant what the performance of existing GPUs are at equivalent clocks peed. You can't compare apples to oranges.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
What if they both post the same or similar scores. Coincidence i take it. The thing is it's not a bad thing if these parts perform well. Being able to boost the clocks which boosts performance ain't bad if it still beats the opposition.
Last edited:
What if pascal is slower than Maxwell at equal clocks?
If performance is similar at equal clocks then it still means that nvidia engineered faster clocks instead of a higher IPC. Even if Pascal is faster than Maxwell at equal clocks you still don't know how much Nvidia balanced increased IPC against increased clocks.
The only way you will know is if nvidia tells you. We already have some strong indications that Nvidia increased clocks instead of IPC, we may very well see a lower IPC on pascal than Maxwell.
** edited *
Clock speed depends entirely on the architecture. If AMD decided to use the smaller process to increase IPC then the instructions per a second will go up at the same clock speed
Given particular fabrication process you can either design a processor with a higher IPC and lower clock speeds, or lower IPC and higher clock speeds. There is no definitive benefit to either approach.
Clock speed depends entirely on the architecture. If AMD decided to use the smaller process to increase IPC then the instructions per a second will go up at the same clock speed
Given particular fabrication process you can either design a processor with a higher IPC and lower clock speeds, or lower IPC and higher clock speeds. There is no definitive benefit to either approach.
Soldato
- Joined
- 28 May 2007
- Posts
- 10,222
Yes but if there are not big changes nothing much will change. Polaris is shaping up to be a big change where as Pascal is close to Maxwell if rumors are true. Polaris is more of an unknown at this minute. It's not so surprising as AMD have not made big changes for a while apart from HBM. Nvidia changed quite a bit with Maxwell.
I think it's becoming clear that pascal is just maxwell rehashed. If AMD do have a truly new architecture. Then this could well be their year. As annoying as it is right now. I think they have been smart letting Nvidia show its cards first. And they got the time to ensure they can beat it.
I dunno I just seemed to remember that card having a low clock speed and probably stand a chance of being lower than the one before it. Did have to verify the 7900 clocks though. Iv never forgotten that it had 768mb of memory also for some reason.
