• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

The Kaveri Pre-Launch Thread (A10-7800 and A10-6800k @3,5 Ghz)

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
Status
Not open for further replies.
Huh? Cape Verde is 1.5B. 1.4B definitely seems too high for the GPU portion of Kaveri, though. Cape Verde's 1.5B includes the busses and other junk. Hawaii is 6.2 billion transistors.

For 640SP yes. That would be 1.2B for 512SP. Again, if we include everything. 128bit controller, x16 PCIe.

I have a feeling AMD made another "Bulldozer" transistor count. And the transistor count will be revised down to below 2B later.
 
Would not be surprised if the transistor count (density) is at least mildly 'overestimated'. Same thing happened with Bulldozer. Imo, they try to match as closely as they can to GF's claimed node attributes even if the reality is a bit different.
 
For 640SP yes. That would be 1.2B for 512SP. Again, if we include everything. 128bit controller, x16 PCIe.

I have a feeling AMD made another "Bulldozer" transistor count. And the transistor count will be revised down to below 2B later.
Honestly, I'm not seeing how it could be anything else. If a dense GPU can't even account for 1 billion transistors at nearly half the area of the chip, there's no way it could be north of 2 billion.
 
On Haswell if you count the GPU and L3 cache you ll
end with more than 50% die area being devoted to high
density circuitry, a number that is about the same as
Kaveri s high density ratio.

amd-kaveri-1-614x345.jpg

AMD Really needs to pay more attention to this Nvidia Slide:

NVCES-025_575px.jpg
 
Last edited:
Quad-core w/ Leopard cores and 4 CUs for x86 phones, 20nm.
Quad-core w/ Cambridge cores and 4 CUs for ARMv8 phones, 20nm.

Spoilers from the future but seriously the GPU should be around ~1.4 billion transistors.
 
Kaveri iGPU has higher transistor count than Cape Verde HD7770 although it has 512 SPs vs 640. Kaveri is GCN 1.1 with more ACES in the frond end for higher Compute, it has True Audio that raises transistor count and it also has CrossFire XDMA. Those features raises transistor count from Cape Verde to Kaveri GCN 1.1.

amd_kaveri_tech_day_2014_020b.JPG



CapeVerde.png
 
Kaveri iGPU has higher transistor count than Cape Verde HD7770 although it has 512 SPs vs 640. Kaveri is GCN 1.1 with more ACES in the frond end for higher Compute, it has True Audio that raises transistor count and it also has CrossFire XDMA. Those features raises transistor count from Cape Verde to Kaveri GCN 1.1.

That must be the reason why the GCN 1.1 based 7790 with 896 cores has a lower transistor count per GCN 1.1 core (2080M/896 = 2.32 million transistors per core) than the GCN 1.0 based 7770 with 640 cores (1500M/640 = 2.34 million transistors per core.)
 
imho...tegra is no competition to amd. it's just another arm soc. amd is about x86 , nvidia can not touch that.

You're right. Tegra isn't competition to AMD. It is better for its intended market. High end tablets. You're comparing a desktop LGA CPU with a 65-100W TDP to a 2W mobile SOC. Do you realize how stupid that is?

Tegra fits in the smallest tablets you can buy, while AMD still has not won any meaningful design wins with their mobile APUs - because they are far worse than Bay Trail or even core CPUs in terms of performance at that power window. Same applies to Tegra. These are 5W TDP SOCs that will fit in tablets that AMD will never get into. The bottom line is that Kaveri is going to be a pretty good HTPC chip for desktop LGA. But as far as mobile goes? It won't have meaningful performance in a 2W-5W TDP window, Bay trail will still be the best chip for that market. And the successor to Bay Trail is coming this year as well.

Meanwhile, Kaveri is a desktop LGA chip - which when cut down for mobile, the performance will be castrated. I can't see mobile Kaveri doing any better than the A4-5000 did, which was a 15W TDP part that never won any meaningful design wins. The Bay Trail owns that market for Windows 8.x tablets right now. Mobile Kaveri will not change that by my predictions - AMD is competing in an x86 market and is in over its head. You can't compete with intel in the mobile x86 market because their process advantage is at least 3 years past TSMC right now.

I do feel that Kaveri will be a good desktop LGA HTPC chip. For that it could be better price/performance than the i3. Outside of that niche though ? I dunno. We'll see.
 
Last edited:
You're comparing a desktop LGA CPU with a 65-100W TDP to a 2W mobile SOC. Do you realize how stupid that is?.

just in case: i'm not comparing anything ...it was simply an answer to OP.

You're right. Tegra isn't competition to AMD. It is better for its intended market. High end tablets.Tegra fits in the smallest tablets you can buy, while AMD still has not won any meaningful design wins with their mobile APUs - because they are far worse than Bay Trail or even core CPUs in terms of performance at that power window. Same applies to Tegra. These are 5W TDP SOCs that will fit in tablets that AMD will never get into. The bottom line is that Kaveri is going to be a pretty good HTPC chip for desktop LGA. But as far as mobile goes? It won't have meaningful performance in a 2W-5W TDP window, Bay trail will still be the best chip for that market. And the successor to Bay Trail is coming this year as well.

Meanwhile, Kaveri is a desktop LGA chip - which when cut down for mobile, the performance will be castrated. I can't see mobile Kaveri doing any better than the A4-5000 did, which was a 15W TDP part that never won any meaningful design wins. The Bay Trail owns that market for Windows 8.x tablets right now. Mobile Kaveri will not change that by my predictions - AMD is competing in an x86 market and is in over its head. You can't compete with intel in the mobile x86 market because their process advantage is at least 3 years past TSMC right now.

I do feel that Kaveri will be a good desktop LGA HTPC chip. For that it could be better price/performance than the i3. Outside of that niche though ? I dunno. We'll see.

that is just a way it is at the moment 😉

besides....intel is on rampage and out for a kill IMO, and the main prey is not amd....arm it is :biggrin:
 
Last edited:
I do feel that Kaveri will be a good desktop LGA HTPC chip. For that it could be better price/performance than the i3. Outside of that niche though ? I dunno. We'll see.

Intel market share is close to 70-80% up to Core i3 CPUs. If kaveri is better than Core i3 then it is better for 70-80% of the entire x86 market 🙄
 
Intel market share is close to 70-80% up to Core i3 CPUs. If kaveri is better than Core i3 then it is better for 70-80% of the entire x86 market 🙄

Maybe you missed the part about price/performance for HTPC use. I never said it was better than i3. Personally? I'd rather have an i3.

I see HTPC being the main niche for the Kaveri. That's about it. It won't make waves on the desktop and doubtful that it will make waves in mobile either - by the time Kaveri is "cut down" and is in an appropriate power window for tablets, the performance will reflect it. That was the case with Kabini, anyway.
 
Maybe you missed the part about price/performance for HTPC use. I never said it was better than i3. Personally? I'd rather have an i3.

I see HTPC being the main niche for the Kaveri. That's about it. It won't make waves on the desktop and doubtful that it will make waves in mobile either - by the time Kaveri is "cut down" and is in an appropriate power window for tablets, the performance will reflect it. That was the case with Kabini, anyway.

Quite the contrary , i dont see i3/i5 as valuable for desktops
unless they have a dGpu so Kaveri fits perfectly the bill for
mainstream offerings since it seems to perform better than
a GT630 + i5 4670K, think about it....
 
a GT630 + i5 4670K, think about it....

i'm sorry , but no one in his right mind will buy i5 4670K to pair it with GT630 , allright.
think about it ....HD4600 is already good enough for lite gaming etc...if one wants to game then choses a proper price video card....right?
 
I'm guessing that pricing is due to limited initial supply. It is a compelling choice versus Intel i3 and lower but not very attractive against i5 and higher except when dealing with Small Form Factor constraints.

Richland's pricing was much more realistic, imo.
 
Last edited:
i'm sorry , but no one in his right mind will buy i5 4670K to pair it with GT630 , allright.
think about it ....HD4600 is already good enough for lite gaming etc...if one wants to game then choses a proper price video card....right?

The problem is that people out there will buy a Core i5 with GTX620/630 OEM system. 🙄

People in forums like Anandtech are only close to 1-2% (perhaps lower) of the global PC market share. The rest goes and buy what each shop will sell them. 😛
 
Quite the contrary , i dont see i3/i5 as valuable for desktops
unless they have a dGpu so Kaveri fits perfectly the bill for
mainstream offerings since it seems to perform better than
a GT630 + i5 4670K, think about it....

Better than 4670k. You're a funny dude. 🙄
 
i'm sorry , but no one in his right mind will buy i5 4670K to pair it with GT630 , allright.
think about it ....HD4600 is already good enough for lite gaming etc...if one wants to game then choses a proper price video card....right?

I m talking of perfs , the market is different and consist mainly
of sub 100$ offerings.
 
Status
Not open for further replies.
Back
Top