The AMD Execution Thread [2007 - 2017]

Status
Not open for further replies.
Nah, it's the new tri-x.
This is with 70W more load than the dcu2 (which is usually better than the twin frozr for the highend card editions)
60576.png

http://www.anandtech.com/show/7601/sapphire-radeon-r9-290-review-our-first-custom-cooled-290/4
 
Nah, it's the new tri-x.
This is with 70W more load than the dcu2 (which is usually better than the twin frozr for the highend card editions)
60576.png

http://www.anandtech.com/show/7601/sapphire-radeon-r9-290-review-our-first-custom-cooled-290/4

We'll see about that when they review R9 290 with Twin Frozr IV - when I posted my post, that review wasn't out yet, and with R9 280X the Tri-X certainly isn't doing too well: http://www.anandtech.com/show/7406/the-sapphire-r9-280x-toxic-review/4
58759.png


(yes, Toxic, for whatever reason despite both using Tri-X cooler, R9 280X is called Toxic and R9 290(X) Tri-X)

edit:
TwinFrozr IV matches DCII on GTX 780's at least despite higher default clocks: http://www.techpowerup.com/reviews/ASUS/GeForce_GTX_780_Direct_Cu_II_OC/25.html http://www.techpowerup.com/reviews/MSI/GTX_780_TF_Gaming/25.html
 
Xbox One and PS4 combined Nov-Dec sales...7.2 million. That is at least 1 million ahead of estimates. Q4 numbers to be released on Jan 21 will be a BLOWOUT imho. Micron reported big numbers tonight and all the anecdotal evidence points to a much improved PC market and continued demand for AMD discrete GPUs.

I am sticking to my $1.65 billion revenue and .12 cents eps...long 25,000 shares
 
On this Kaveri launch...

I guess if I were AMD I would make sure the best dual graphics combo was out there at launch...to be uniformly benched...and supported with current drivers. Instead we get beta drivers and the possibility of compatibility with the full range of R7 discrete cards...nothing definitive Seems like AMD thinks through 90% of the details at launch and then discounts the remaining 10%..."they'll figure it out". Much like the reference cooler and throttling issues with R9. Sometimes when the review sites and consumers are left to "figure it out" they reach wrong or incomplete conclusions that may not paint their products in the best light.

The result is confusion as to what is possible...what is optimal...and what separates this APU from the competition or previous generation. The impression, as a result, is the 95 watt parts under-perform which may not be in fact true if matched with the right discrete gpu, drivers, and other optimizations. Shouldn't it be troubling to AMD that the general impression out there is the 45 watt part is the real game changer and not the flagship 95 watt part...specially since the 95 watt part are the only ones available?

I don't understand this final phase of execution that AMD has exhibited with its last 2 BIG launches...R9 and now Kaveri...it just seems incomplete and messy. I would encourage AMD to pay closer attention to crafting & driving their own narrative rather than leaving so much open to interpretation by others. Perhaps this doesn't effect OEM or partner design wins but I think it hurts them with the retail consumer. Mind share is critically important for a company that is restructuring and a consistent unambiguous message is needed. The first rule in PR 101 is stay in control of your message...never deviate...and stay ON message. You can't do this with a 90% launch...

Tell me what good a 7870k would do you right now without optimized drivers and a CLEAR idea as to which discrete card to get?

http://www.amd.com/us/products/technologies/dual-graphics/pages/dual-graphics.aspx#3

Not updated...

http://support.amd.com/en-us/download/desktop?os=Windows+7+-+64

Not updated...
 
On this Kaveri launch...

I guess if I were AMD I would make sure the best dual graphics combo was out there at launch...to be uniformly benched...and supported with current drivers.
Maybe February and some magical drivers will change things, but I'm going with history and assuming that the best dual graphics combo is not doing dual graphics. The sliver of use cases where the inconsistent benefit of what amounts to Crossfire with piddly GPUs in a configuration supported more poorly than how AMD supports discrete Crossfire (crappily, for those keeping track) is worth the headache of managing multi-GPU (a lot of it, for those keeping track) versus just getting a better card or getting a cheaper i3 and a better card is small. Better to save a few pennies a week and bite the marginally higher cost.

Sometimes when the review sites and consumers are left to "figure it out" they reach wrong or incomplete conclusions that may not paint their products in the best light.
That AMD does this so consistently should allow the drawing of a more far-reaching conclusion of its own.

Shouldn't it be troubling to AMD that the general impression out there is the 45 watt part is the real game changer and not the flagship 95 watt part...specially since the 95 watt part are the only ones available?
AMD does not have a future in the upper power range with this architecture. It hasn't severly cut back on its desktop presence and just given up on the traditional server CPU market for no reason.
(edit: to be fair, that upper tier desktop market isn't making a good case for itself these days)
It has trimmed its design parameters such that the severe physical process deficit and design inferiority can be compensated by dropping out of the clock, performance, feature, and power ranges where you have to be at the top of your game to be compelling.

I would encourage AMD to pay closer attention to crafting & driving their own narrative rather than leaving so much open to interpretation by others. Perhaps this doesn't effect OEM or partner design wins but I think it hurts them with the retail consumer. Mind share is critically important for a company that is restructuring and a consistent unambiguous message is needed. The first rule in PR 101 is stay in control of your message...never deviate...and stay ON message. You can't do this with a 90% launch...
PR has to spin the base truth they have been given. Granted, AMD probably fired most of their PR, but even the best spin masters can only go so far with what they've been given.

Tell me what good a 7870k would do you right now without optimized drivers and a CLEAR idea as to which discrete card to get?
This is begging the question that a 7870K makes a strong case for itself even with those things.
 
Last edited by a moderator:
http://media.bestofmicro.com/Q/K/418268/original/61a.jpg

Notice on the bottom "On AMD A10-7850 APU + R9 290x"

I have modified my opinion of the Kaveri roll out. I think that the reason Kaveri hasn't been hyped yet with a specific discrete GPU combo is because the driver isn't ready. Part of the reason the driver isn't ready is Dice was late with their Mantle patch. One of the main reasons AMD has switched to this APU architecture was to push their GPU ip since they can't beat Intel on the CPU side. To push their GPU ip, they needed an in-house api to leverage it. Since they had a working relationship with the top game devs via the console business, it only made sense to partner with one of the bigs for the Mantle driver roll out...in this case Dice.

Dice threw a wrench into things last month when they announced they would be late with the Mantle update. AMD was committed to the roll out date due to ramp issues, OEM placements, earnings, and other issues...so they went without the Mantle driver. I am calling this a soft launch with the REAL launch coming with the Mantle drivers and specific discrete pairings that the review sites and others can sink their teeth into. I am giving AMD a 30 day pass on this launch...provided the Mantle driver launch is handled well.

Everybody who buys a $200 4670k would buy a 7850k IF the graphics pairing was superior on the major titles. I would argue that if this Mantle driver can show superior benchmarks with a 270x ($200) and above discrete card paired with a 7850k, this could be a game changer. I think this roll out was soft on purpose because the Mantle driver was late largely because of Dice.
 
Yeah I am sure AMD wanted to have Mantle and BF4 to show off with today's Kaveri launch. If it could make that CPU look more "equal" to Intel, it would be a clear benefit to building mind share among gamers.

I have doubts about it helping the IGP much though considering how extremely mem bandwidth bottlenecked it is. That aspect is curious too. The IGP could be making a much better showing and really blowing Intel away if only AMD had figured out some solution to memory bandwidth.
 
One of the main reasons AMD has switched to this APU architecture was to push their GPU ip since they can't beat Intel on the CPU side.
....
Everybody who buys a $200 4670k would buy a 7850k IF the graphics pairing was superior on the major titles.

...exactly my thinking. Who cares of having 4 fast cores if most of your games are +20% faster with the APU?
 
http://media.bestofmicro.com/Q/K/418268/original/61a.jpg

Notice on the bottom "On AMD A10-7850 APU + R9 290x"

I have modified my opinion of the Kaveri roll out. I think that the reason Kaveri hasn't been hyped yet with a specific discrete GPU combo is because the driver isn't ready. Part of the reason the driver isn't ready is Dice was late with their Mantle patch. One of the main reasons AMD has switched to this APU architecture was to push their GPU ip since they can't beat Intel on the CPU side. To push their GPU ip, they needed an in-house api to leverage it. Since they had a working relationship with the top game devs via the console business, it only made sense to partner with one of the bigs for the Mantle driver roll out...in this case Dice.

Dice threw a wrench into things last month when they announced they would be late with the Mantle update. AMD was committed to the roll out date due to ramp issues, OEM placements, earnings, and other issues...so they went without the Mantle driver. I am calling this a soft launch with the REAL launch coming with the Mantle drivers and specific discrete pairings that the review sites and others can sink their teeth into. I am giving AMD a 30 day pass on this launch...provided the Mantle driver launch is handled well.

Everybody who buys a $200 4670k would buy a 7850k IF the graphics pairing was superior on the major titles. I would argue that if this Mantle driver can show superior benchmarks with a 270x ($200) and above discrete card paired with a 7850k, this could be a game changer. I think this roll out was soft on purpose because the Mantle driver was late largely because of Dice.

Mantle will help, but it will help just as much with Intel CPUs. I wouldn't expect Dual Graphics to have much of an impact (if any) with large discrete GPUs. Plus, Dual Graphics is not related to Mantle, which will be limited to certain games, so I don't see how this could radically change the picture.

I do agree, however, that AMD was most likely planning to rely on a Battlefield + Mantle benchmark in all Kaveri reviews, but it doesn't explain why Dual Graphics is half-baked.
 
Since nothing like this appears anywhere on AMD's roadmaps, even the leaked ones that extend to 2015, I'm guessing this design, whatever it was, was scrapped.
 
Here is where my lack of tech knowledge hurts me. That "45% faster" slide for BF4 shows the "dual graphics" counterpart to the 7850k as the 290x. It seems like there should be a huge bottleneck with the CPU portion of that APU with a card that powerful. Wouldn't a 270x, 280, or 280x show similar scores because of the CPU bottleneck? If not, then are games with Mantle optimizations "free" from CPU bottlenecks and able to improve and scale higher on better and better GPU processing power? If so, why would AMD show such an expensive combo to get those results?

Does this imply that Kaveri will work with any R7 or R9 card? I really hate the lack of clarity from AMD on which card(s)...what driver...etc. I still can't understand the 290x pairing...
 
AMD is blowing up ahead of earnings here...just hit $4.53 a share. The chart looks top heavy but this close to earnings it is sustainable..at least for a week or so. My position I started building on Oct 23 at $3.15 is looking solid right now...
 
There is no dual graphics involved there, it's only the 290x doing the graphics.
The mantle version is not free from cpu bottlenecks - bf4 (multiplayer) is a quite cpu heavy game, doing a lot of non-rendering stuff which already takes most of the cpu.
Mantle is likely helping more on cpu than gpu limits, so there's a point in running it on a quite cpu limited system ;)
However, they could probably have claimed even bigger difference if they used eg. a fx-8320 as it would be limited by the single directx render thread, while other cores are sitting idle - so it could maybe go from ~70% cpu utilization with directx to ~100% with mantle. The 7850k is like running at nearly 100% cpu in both dx and mantle version.
 
So the Mantle benchmarks will likely show Intel CPUs as superior but show AMD superior to Nvidia on the discrete GPU side of things due to the optimization? This is predicated on the "dual graphics" having little impact correct?

If so, good for discrete sales...not so much for Kaveri.
 
Not just that..
less cpu limit (and better use of gpu ofcourse) => higher performance than with directx (ie advantage against nvidia)
less cpu limit => cpu speed is less important (amd can better claim; "sure our cpus are slower, but they are fast enough")
better multicore utilization => advantage for amd's general approach of slower, but more cores for the price.
 
Not just that..
less cpu limit (and better use of gpu ofcourse) => higher performance than with directx (ie advantage against nvidia)
less cpu limit => cpu speed is less important (amd can better claim; "sure our cpus are slower, but they are fast enough")
better multicore utilization => advantage for amd's general approach of slower, but more cores for the price.

Got it...ty
 
So the Mantle benchmarks will likely show Intel CPUs as superior but show AMD superior to Nvidia on the discrete GPU side of things due to the optimization? This is predicated on the "dual graphics" having little impact correct?

If so, good for discrete sales...not so much for Kaveri.

In games where the CPU load is mostly rendering-related, AMD should pretty much reach the same level as Intel. In games that do a lot of physics, AI, etc., on the CPU, Intel should remain ahead, but AMD would be somewhat closer behind. Mantle also makes it easier to spread the rendering across many threads, which should be very helpful on AMD's FX line, but since the latter is apparently dead…

Mantle in itself does not enable Dual Graphics, but it makes it possible for developers to handle how they want to split up the rendering between different GPUs. I doubt many will bother to actually do it for asymmetrical setups (i.e. APU + dGPU) but AMD might offer to do (most of) the work for a few high-profile engines/games. I doubt it will amount to more than a bit of shiny material for a few marketing slides.

That said, Dual Graphics is currently limited by the low bandwidth of APUs. It's difficult to have good dual-GPU performance in general, and it only gets harder when they're mismatched. Kaveri can work well with a 7750 DDR3 (albeit very inconsistently) but it's much less helpful when paired with a 7750 GDDR5, let alone something faster.

Kaveri's successor, which is apparently named Carrizo, might have up to 16CUs on 20nm with HBM (pure speculation, but it sounds reasonable). This would put it somewhere between the Xbone and the PS4 in graphics computing power, which is certainly decent enough for Dual Graphics with mid-range graphics cards. But at this point, most people would probably be better off just sticking to the integrated graphics, while power users will prefer high-end cards that are unlikely to work well with Carrizo. Dual-graphics with a mid-range part could be an acceptable option for high performance on a budget, but unless it gets more reliable, I'd avoid it. Anyway, this last paragraph is 100% guesswork.
 
The inference in the "45% faster" slide is the compatibility of dual graphics with 7850k and 290x...as both are specifically cited. This is confusing, disingenuous, and flat out WRONG according to the 2 sources I just found:

http://news.softpedia.com/news/AMD-...TrueAudio-and-Fluid-Motion-Video-416847.shtml

"The performance difference (percentage-wise) won't be very large on R9 cards, but the improvements gained by systems with, say, a Radeon R7 240 are substantial.

According to AMD's tests, the Radeon R7 240, alone, runs Bioshock: Infinite at 20.4 FPS (frames per second) and Tomb Raider at 19.4 FPS.

In a Kaveri system, though, Dual Graphics tech pools the GPU resources together and increases the FPS.

The rate goes to 40.2 in the former (95% faster) and 38.1 in the latter game (again, 95% faster). So if you build a Kaveri-based PC, it actually pays to fork up $80 / €80 more for a low-end video board."

http://wccftech.com/amd-kaveri-dual-graphics-works-ddr3-memory-based-radeon-r7-gpus/

Update - We have just received information form an AMD representative that the AMD Kaveri APU officially supports the Radeon R7 250 and Radeon R7 240 with dual graphics technology. There’s no memory type restriction and the APU would be able to work with both DDR3 and GDDR5 variants of these cards. Do note that no other Radeon R200 series GPUs are supported by Kaveri for dual graphics compatibility. Following is the reply:

~ So out of the models you listed only the R7 240 and R7 250 are officially supported.
~ 7750 is not and neither is R7 260X
~ So this has nothing to do with DDR3 or GDDR5
~ R7 240 GDDR5 as well as R7 250 GDDR5 should work just fine.
~ In summary: R7 250 and R7 240 are the models that are supported for Dual Graphics with Kaveri.
 
Status
Not open for further replies.
Back
Top