AMD: Volcanic Islands R1100/1200 (8***/9*** series) Speculation/ Rumour Thread

[UniversalTruth]

This is the last time I'm telling you to stop posting stupid things outside of RPSC. Nobody on the tech side of the forums has any particular interest in wonky world views pouring onto everything.

AlexV, I think you are great guys here in this forum. If it was another one, I could have been expelled long time ago. You are good people.

Sorry, really. It was just a funny joke-ish type of comment to try to answer the other forum member attackings. No serious intentions whatsoever.

I know that perhaps I am not right for many things but... yeah...

I hope you understand

Cheers

 

[Pete]

UT, what part of boxleitnerb's post was an "attack" that you needed to respond to?

I also can't figure out the joke(s) in your post, but one puzzle at a time.

 

[Ailuros]

Oh come on he's nearly begging for a group hug, it's no time playing hard to get

 

[kyetech]

Well... that was awkward.

lol

 

[jaredpace]

What needs to change in the upcoming chip design in order for Tahiti's successor to have an increase in Mtris/sec Setup Rate, an increase in FP16 billinear texel rate, and an increase in Gtexels/sec ROP rate?


edit:

To stay grounded in reality & avoid disappointment, I shall guess that they stayed on 28nm and did a Bonaire-type upgrade to Tahiti resulting in ~30% performance increase.


Contrarily, here is my wishlist for upgrades & improvements:

- More focus on sustainable high clock states via TCP TDP control of voltage regulation
- Some finesse to their Powertune 2.0 C7 voltage/clock states using dedicated circuitry
- A general improvement to maximize "Boost state" performance relative to silicon quality
- Each ASIC sample achieves highest possible mhz depending on leakage (finer grained binning)
- An improved metal bracket (ie EVGA high-air-flow) exhaust from the blower / maximize airflow in shroud
- A redesigned and more efficient Heatsink & cooling solution. (Which I believe has been speculated)
- A "graphics-centric" core design with sacrificed DP throughput (and an alternate ASIC for compute)
- An increase in the number of "ACEs" or setup engines for the 'Front-End'
- A jump to 20nm & 4096+ ALU on a R600-or-greater sized die
- Increased Rasterizer Performance / ROP count & output
- Increase in Mtris/sec, FP16 billinear texel rate, and Gtexels/sec
- Bring back custom downsampling resolutions & scaling in CCC

 

[Albuquerque]

So, uh, when do I get more info on the 9700Pro? And the flashable, unlockable 9500Non-Pro?

I sooo miss my flashed, unlocked, overclocked 9500np from back in the day. I think my card still holds the #2 or #3 benchmark spot over on Overclockers.com for the 9500np unlocked category.

 

[LordEC911]

Trying to get things back on track...

I guess I'm having a hard time understand how people on other forums are implying that Bonaire has less/equal performance/w than 7770....

40% more SPs
33% more bandwidth
39% more transistors
30% larger die
~30% more performance
at roughly the same TDP ~5% diff.

 

[boxleitnerb]

I guess I'm having a hard time understand how people on other forums are implying that Bonaire has less/equal performance/w than 7770....

You probably mean me and have not read my post thoroughly enough to make that kind of statement. Maybe you should read again, more carefully. It depends on the specific SKU, on the review (I only use compilations of multiple values to make my points, I don't care for cherrypicking) and of course cooling (the 7790 often has better cooling, enhancing power efficiency through lower temps).

 

[Alexko]

What needs to change in the upcoming chip design in order for Tahiti's successor to have an increase in Mtris/sec Setup Rate, an increase in FP16 billinear texel rate, and an increase in Gtexels/sec ROP rate?


edit:

To stay grounded in reality & avoid disappointment, I shall guess that they stayed on 28nm and did a Bonaire-type upgrade to Tahiti resulting in ~30% performance increase.


Contrarily, here is my wishlist for upgrades & improvements:

- More focus on sustainable high clock states via TCP TDP control of voltage regulation
- Some finesse to their Powertune 2.0 C7 voltage/clock states using dedicated circuitry
- A general improvement to maximize "Boost state" performance relative to silicon quality
- Each ASIC sample achieves highest possible mhz depending on leakage (finer grained binning)
- An improved metal bracket (ie EVGA high-air-flow) exhaust from the blower / maximize airflow in shroud
- A redesigned and more efficient Heatsink & cooling solution. (Which I believe has been speculated)
- A "graphics-centric" core design with sacrificed DP throughput (and an alternate ASIC for compute)
- An increase in the number of "ACEs" or setup engines for the 'Front-End'
- A jump to 20nm & 4096+ ALU on a R600-or-greater sized die
- Increased Rasterizer Performance / ROP count & output
- Increase in Mtris/sec, FP16 billinear texel rate, and Gtexels/sec
- Bring back custom downsampling resolutions & scaling in CCC

For what it's worth, Charlie claims that the changes in Hawaii are major, compared to the minor tweaks in Bonaire.

 

[Kaotik]

You probably mean me and have not read my post thoroughly enough to make that kind of statement. Maybe you should read again, more carefully. It depends on the specific SKU, on the review (I only use compilations of multiple values to make my points, I don't care for cherrypicking) and of course cooling (the 7790 often has better cooling, enhancing power efficiency through lower temps).

Sapphire 7790 Dual-X has better perf/watt at least at TPU compared to Sapphire 7790 Vapor-X. Vapor-X is superior cooler compared to Dual-X (AFAIK)

 

[boxleitnerb]

7790 Vapor-X? I think that doesn't exist

 

[jimbo75]

You probably mean me and have not read my post thoroughly enough to make that kind of statement. Maybe you should read again, more carefully. It depends on the specific SKU, on the review (I only use compilations of multiple values to make my points, I don't care for cherrypicking) and of course cooling (the 7790 often has better cooling, enhancing power efficiency through lower temps).

http://hexus.net/tech/reviews/graph...ut-sapphire-vs-powercolor-vs-gigabyte/?page=4

We're not entirely sure where Gigabyte gets the Triangle Cool Technology name from; the heatsink is basic and previously used on the less-powerful HD 7770 GPU.

Doesn't seem to affect temps or power draw compared to the better dual-cooled cards.

http://hexus.net/tech/reviews/graph...t-sapphire-vs-powercolor-vs-gigabyte/?page=12

 

[Dave Baumann]

- An improved metal bracket (ie EVGA high-air-flow) exhaust from the blower / maximize airflow in shroud

Note, while you may view these things as improved, bear in mind that the IHV's are beholden to regulatory certs and requirements. The back brackets actually have a lot to do with EMI (and there is a structural/rigidity element to them as well).

 

[boxleitnerb]

http://hexus.net/tech/reviews/graph...ut-sapphire-vs-powercolor-vs-gigabyte/?page=4

Doesn't seem to affect temps or power draw compared to the better dual-cooled cards.

http://hexus.net/tech/reviews/graph...t-sapphire-vs-powercolor-vs-gigabyte/?page=12

Read again: I said that lower temps generally lead to lower power consumption. The comparison here is between Pitcairn and the Bonaire review samples. Posting two Bonaire cards and then even pointing out that their load temperature is equal, doesn't make any sense in this context.

 

[Silent_Buddha]

Read again: I said that lower temps generally lead to lower power consumption. The comparison here is between Pitcairn and the Bonaire review samples. Posting two Bonaire cards and then even pointing out that their load temperature is equal, doesn't make any sense in this context.

Your specific claim however was that 7790's in general have better cooling which is why they have lower temps and better power consumption.

He was pointing out that the "better cooling" has nothing to do with it as it doesn't change power consumption whether it has a better or worse cooling solution.

That isn't to say that better cooling can't result in better power efficiency, but if the difference is only a few degrees or tens of degrees it will be virtually unmeasurable with standard temp monitoring hardware.

Regards,
SB

 

[jaredpace]

Sounds like legitimate hurdles ahead. One of these designs aspirates much easier, runs cooler & lives longer, likely lowering power requirements. Even if it's just 1-2% improvement, this bracket idea is just one of many overlooked characteristics of a better 'final product'. If the competition can do it, it should be within rights. Fine grained DPM & 'boost' being even more significant. 1000-1200mhz range should have 40 DPM states with 5mhz steps.

 

[jimbo75]

Read again: I said that lower temps generally lead to lower power consumption. The comparison here is between Pitcairn and the Bonaire review samples. Posting two Bonaire cards and then even pointing out that their load temperature is equal, doesn't make any sense in this context.

Can you post some examples btw? I haven't really paid much attention to this card but looking over a bunch of reviews of it, it seems to draw very similar power to the 7770, regardless of cooler.

 

[gongo]

My 2 hopeful implementation for new AMD gpus...

1. independent audio and video hdmi output....finally multi-channel pcm audio done right...good bye 5.1 compressed encoding soundcards

2. easy heatsink removal/access to the gpu core.. so we can all attached an...AIO h2o cooler/full h2o cooler water block...to cool the hot die while the other parts like VRMs & RAMs are still secured and cool with default blower fan blowing air pass them....


oh yes btw Dave...with these new gpus to be sold...does it mean AMD will not be releasing the rumored memory-cum-multithreading DX11 improving drviers for HD7xxx ???

 

[boxleitnerb]

Your specific claim however was that 7790's in general have better cooling which is why they have lower temps and better power consumption.

He was pointing out that the "better cooling" has nothing to do with it as it doesn't change power consumption whether it has a better or worse cooling solution.

That isn't to say that better cooling can't result in better power efficiency, but if the difference is only a few degrees or tens of degrees it will be virtually unmeasurable with standard temp monitoring hardware.

Regards,
SB

No, I didn't claim that. I said, they "often" have better cooling. For example the 7790 Dual-X can have up to 15°C better temperatures under load than the 7870:
https://www.computerbase.de/artikel/grafikkarten/2013/amd-radeon-hd-7790-im-test/9/

I measured a ~15W difference on my GeForce GTX Titan when I lowered temps from 85°C to 65°C.

Just saying, it's not something you cannot ignore, even if it's just 2-3 W difference. Those can make or brake Bonaires Perf/W advantage:
Indexed efficiency values from 5+ reviews:
http://www.3dcenter.org/artikel/grafikkarten-marktueberblick-juli-2013
7790: 185% perf/81W = 2.28
7870: 260% perf/120W = 2.17
Add 2.5W to the 7790 and you get 185% perf/84W = 2.22 -> 2.3% more energy efficient than the 7870. That is groundbreaking!

Btw I've just seen, one card hat 1.13V and the other 1.17V under load. That should explain the difference in power consumption between both 7790 samples from TPU even better than temperature alone. So even less reason for generalizing. Who knows what the average 7790 and 7870 have as voltages.

 

[jimbo75]

http://www.tomshardware.com/reviews/radeon-hd-7790-bonaire-performance,3462-10.html
http://www.guru3d.com/articles_pages/msi_radeon_7790_oc_review,6.html
http://www.anandtech.com/show/6837/...eat-sapphire-the-first-desktop-sea-islands/16
http://www.vortez.net/articles_pages/his_hd_7790_turbo_review,9.html
http://www.bit-tech.net/hardware/graphics/2013/03/22/sapphire-radeon-hd-7790-1gb-review/8
http://www.hardwareheaven.com/revie...eview-power-temps-noise-and-overclocking.html

Those 6 reviews have the 7790's power consumption much closer to the 7770 than the 7850.

In the AT review, 16C temp difference between the 7790's makes zero difference to the cards power consumption (the OC'd card is 5% faster).

You may well have a point about different voltages, but that doesn't prove that Bonaire isn't more efficient than Pitcairn, more like AMD wasn't bothered about binning them.