Soldato
- Joined
- 25 Nov 2011
- Posts
- 20,676
- Location
- The KOP
I have no idea what happens behind the scene. The code from console wasn't used for a reason what reason that is remains to be debated.
-
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
I have no idea what happens behind the scene. The code from console wasn't used for a reason what reason that is remains to be debated.
No, yuo are missing the point. If you artificially limit so ething like 2 cars to drive at the same speed it doesn't tell you which car is the fastest.
Running the same clock speed indicates IPC differences, but as has been explained a thousand times on here IPC is actually comeptlely irrelevant. The most important measure is instructions per second. Pascal achieves far higher IPS.ascal might actually perform slower than maxwell at the same clock speed. That doesn't mean pascal s worse than maxwell, merely that the critical path optimization used to increase clock speed split some critical long instructions over more clock ticks.
Because async compute is hard to tune for different architectures. Async code for GCN 1.3 vs 1.2 vs 1.1 etc.
D.P., I think Graphics Card subforum needs a huge banner that says "Instructions per Clock isn't everything. Instructions per Second is." In big letters and easy to read font. Too many times I've read in this subforum people trying to compare clock-for-clock different architectures. Same architecture you can do, but not two different ones, people seem to have a difficulty understanding that.
I can now say 'yea okay, lets compare Fiji vs Maxwell (same 28nm at least) at same clock, lets say 980Ti or a TX vs FuryX @ 1500 MHz'. Just doesn't work that way...
Edit: why do I get the feeling I will get flamed by nVidia preachers 'cause Fiji can't do 1500MHz
I can now say 'yea okay, lets compare Fiji vs Maxwell (same 28nm at least) at same clock, lets say 980Ti or a TX vs FuryX @ 1500 MHz'. Just doesn't work that way...
Edit: why do I get the feeling I will get flamed by nVidia preachers 'cause Fiji can't do 1500MHz
Last edited:
D.P., I think Graphics Card subforum needs a huge banner that says "Instructions per Clock isn't everything. Instructions per Second is." In big letters and easy to read font. Too many times I've read in this subforum people trying to compare clock-for-clock different architectures. Same architecture you can do, but not two different ones, people seem to have a difficulty understanding that.
I can now say 'yea okay, lets compare Fiji vs Maxwell (same 28nm at least) at same clock, lets say 980Ti or a TX vs FuryX @ 1500 MHz'. Just doesn't work that way...
Edit: why do I get the feeling I will get flamed by nVidia preachers 'cause Fiji can't do 1500MHz
It's bizarre that people seem to card about a single metric like IPC. It just gets so frustrating when you explain the irrelevance that they just don't seem to understand such basic concepts.
Pascal in DX12 offers no improvement over DX11 is the point. This is in contrast to DX12 showing a sometimes significant rise in performance for AMD. It could be down to better driver overhead or it could be down to the fact that AMD simply have better DX12 hardware. You might try convincing yourself that DX11 in future will still be totally relevant but the fact is more and more games and AAA titles will adopt DX12. The facts show that AMD benefit from DX12, Nvidia do not and writing that off as irrelevant is potentially putting anyone choosing Pascal at a performance disadvantage going forward.
Last edited:
Soldato
- Joined
- 19 Feb 2011
- Posts
- 5,849
Can we get away from discussing Pascal / 1080 / 1070 in this thread please and try it keep it to the topic of Polaris and AMD information.
That Zen rumor looks to be good, the AIO solution for AMD cards is something i was hoping we would not see again
Still if it means it gives the card a ton of overclocking headroom (im not sure it will if its anything like the Fiji rubbish) then i suppose i can live with it.
Hopefully tomorrow we might have a better idea of what to expect, however i do think we will see some Vega stuff before the end of the year, if only the entry level Vega stuff to compete with whats available at the time.
That Zen rumor looks to be good, the AIO solution for AMD cards is something i was hoping we would not see again
Still if it means it gives the card a ton of overclocking headroom (im not sure it will if its anything like the Fiji rubbish) then i suppose i can live with it.Hopefully tomorrow we might have a better idea of what to expect, however i do think we will see some Vega stuff before the end of the year, if only the entry level Vega stuff to compete with whats available at the time.
Yes, it does. As even the other guy finally admits and all the benchmarks show. You're 100% wrong and we really need to move on now.
I don't think we are truly seeing what any of the cards whether Maxwell, Pascal, current GCN or Polaris, etc. can do (or can't do) when it comes to ground up written DX12 or Vulkan titles anyhow.
Soldato
- Joined
- 27 Dec 2009
- Posts
- 2,727
- Location
- Gillingham, Kent
Can we get away from discussing Pascal / 1080 / 1070 in this thread please and try it keep it to the topic of Polaris and AMD information.
That Zen rumor looks to be good, the AIO solution for AMD cards is something i was hoping we would not see again
Hopefully tomorrow we might have a better idea of what to expect, however i do think we will see some Vega stuff before the end of the year, if only the entry level Vega stuff to compete with whats available at the time.
It is getting difficult trawling through seemingly the same usernames having unrelated arguments in this thread to actually catch any news about Polaris, definitely agree with you on that point!
Seeing Zen would be nice, even if it's just to see that working samples exist and (hopefully) are pretty fast under whatever favourable conditions AMD are willing to show them. Understandable as you don't showcase something your new product is bad at - I don't ever remember Ferrari launching a new car and only letting journalists compare MPG figures.
I'm quite happy with the concept of AMD using an AIO cooler for it's dual GPU card. The picture posted earlier has the cooler on a card with 4 x PCI-E power connectors, so I am presuming it's a dual GPU card. Would prefer they keep using an AIO cooler rather than an air cooler that struggles to allow the card to run at full speed, I'm thinking the 7990 specifically in poor airflow cases/situations.
Definitely looking forward to more info about Polaris and others coming out. Need to know what my upgrade options and timescales are.
That Zen rumor looks to be good, the AIO solution for AMD cards is something i was hoping we would not see again
I would not see an AIO option as a sign of something bad. Look at how many AIO hybrid cards are coming out for the 1080 and 1070.
But I agree, it happens far too often so could people stop filling the thread with Pascal talk. And it is always the same hand full of people.
Yes indeedy. I said it earlier in the thread, I think the focus is providing chips for the big 3 console updates and for apple.
Why focus on providing a 1080 competitor right now when AMD has other strategic priorities. Many of my games currently run on a single card (980ti) at 1080p at 100fps+ so there is a big market for a slower, cooler card which can push 60hz out via laptops and TVs.
Throwing up benchmarks of how Nvidia beats AMD using the 1070 and 1080 is pointless as the companies currently have different strategic priorities to make money.
Usual suspects on the forum keep spamming NV posts in this thread
Last edited:
Here we go again. AMD supporter brings up Nvidia stuff.Usual suspects on the forum keep spamming NV posts in this thread
'Usual suspect' accusation ensues.
Last edited:
There are a lot of AIO's for CPUs that are a fashion statement, more than for performance or quietness. It seems that fashion is now transferring to GPUs. We've seen in the last couple of generations that air coolers can be every bit as good (and without leaks or pump whine).
http://www.hardocp.com/article/2016...x_1080_founders_edition_review/6#.V02WxrBRE-U
No difference between GTX1080 at DX11 vs DX12 in actual gameplay. Fury X shows an increase.
We can cherry pick all day but for me the facts point to GTX 1080 showing no real improvement going from DX11 to DX12.
Dunno if it has changed but apparently stuff like ASync support isn't even enabled in AoTS at the moment for Pascal.
The framerate changes in that benchmark are so small as to be within the realms even of margin of error lol 38.8-41.7 for the FX doesn't really indicate DX12 doing much at all.
Yes indeedy. I said it earlier in the thread, I think the focus is providing chips for the big 3 console updates and for apple.
Why focus on providing a 1080 competitor right now when AMD has other strategic priorities. Many of my games currently run on a single card (980ti) at 1080p at 100fps+ so there is a big market for a slower, cooler card which can push 60hz out via laptops and TVs.
Throwing up benchmarks of how Nvidia beats AMD using the 1070 and 1080 is pointless as the companies currently have different strategic priorities to make money.
Usual suspects on the forum keep spamming NV posts in this thread
I don't see how AMD's biggest, and only, competitor being mentioned in a thread about an upcoming AMD product would be irrelevant? We were discussing benefits gained from DX12, as that is an area where AMD have had the biggest advantage (and still do imo), and one way for Polaris to be more attractive than the GTX 1070 for customers. I think a lot of people are waiting for Polaris until deciding which product to settle for, so everything you'll be seeing tomorrow form AMD is going to be compared to their competitors' products (i.e. Intel and Nvidia).