AMD FX-8350 - "Piledriver" for AMD Socket AM3+

[btarunr]

There is a couple generations after this one.

Zambezi(2011) -> Vishera(2012) -> Vishera 2.0(2013)

All Orochi...eight cores

Unknown 1.0(2014) -> Unknown 1.1(2015)

All Viperfish...ten cores

^---
Quad-lane DDR4
20 MB L2 cache + 4 MB Northbridge Cache(IOMMU/HMMU Coherency)
So, far the socket is going to be AM3+ or the long thought dead Socket G3.
Other awesome-things come with Viperfish.

Most of what you're talking about is Opteron/enterprise/MCMs. Sure, there was a bulldozer>piledriver>steamroller roadmap, but that related to micro-architecture, not necessarily how AMD implemented it.

 

[seronx]

No client CPUs in the future. Most of what you're talking about is Opteron/enterprise/MCMs.

Vishera, Vishera 2.0, Viperfish, Trinity, Trinity 2.0, Kaveri say otherwise.

 

[darkangel0504]

will steamroller support AM3+ ?

 

[btarunr]

Vishera, Vishera 2.0, Viperfish, Trinity, Trinity 2.0, Kaveri say otherwise.

Trinity, Trinity 2.0, Kaveri are not CPUs, they're APUs. My original assertion was that I don't see AMD client CPUs in the future, I only see APUs and enterprise CPUs. Never heard of Vishera 2.0. The only client product roadmaps we're seeing relate to APUs.

 

[seronx]

will steamroller support AM3+ ?

Steamroller will support AM3+ in form of Orochi Rev E. much like Thuban in the Phenom II era. The true Steamroller CPU Viperfish won't be shown till 2014. The reasons I believe it was delayed is that it is going to use 22-nm node/28-nm interconnect FinFETs much like how 14-XM uses the 14-nm node/20-nm interconnect.

 

[Nordic]

Your link sends me back to Guru3d

Agreed you 2500K is great and not worth changing over to this CPU, its more so for us Phenom II users that wish to catch up to those SB users I think over all this new CPU is a good buy. For me personally its a good upgrade, ive had a 990FX board sitting here for almost 6 months waiting for a good CPU to drop into it and this for me is what ill be getting.

Fixed the link. I would not consider side/upgrading right now. My 2500k is plenty right now. I just am... I don't know... Just curious as to what amd has compared to what I got. I do the same comparisons to higher up intel cpus to.

TI think it is due to the faster cache and lower latency interconnect that Intel wins. The problem with Bulldozer is to focused on not making errors that it actually needs a higher clock rate. For AMD it takes to long to get the instructions to the cores and to get executed and get written since the L2 is so slow.

Thats because this base architecture was primarily designed for servers right? The way you put it, it sounds like an easy fix.

 

[TheLaughingMan]

AMD FX-8350 Vishera 4.0GHz (4.2GHz Tubo) Socket AM...

Retail set at $220. Not bad, but I think $5 too high. I expect a drop in price by Christmas

 

[johnnyfiive]

Its nice to finally see what PD can do.

 

[seronx]

That's because this base architecture was primarily designed for servers right? The way you put it, it sounds like an easy fix.

The fix was to get rid of the L3 all together and simply use the L2.

Trinity: 4MB of L2(2 * 2)
Kaveri: 8MB of L2(4 * 2)
Orochi: 8MB of L2 + 8MB of L3 = 16 MB of total cache(2 * 4 + 2 * 4)
Viperfish: 20 MB of L2(4 * 5)

So, far there is two modes based on the slides for Kaveri and Viperfish:
High Performance mode
Low Performance mode

HPM -> Full 20/8 MB of L2 @ 32 - 40 cycles of latency
LPM -> Half 10/4 MB of L2 @ 16 - 20 cycles of latency

This way it can appease both server workloads and mainstream workloads. The cache is completely unified as well so all cores can access the L2.

 

[TheLaughingMan]

You go on with the speculation about processors 2 and 3 years away Seronx. I am going to turn off my FX-8350 and go to sleep.

 

[cadaveca]

The fix was to get rid of the L3 all together and simply use the L2.

Trinity: 4MB of L2(2 * 2)
Kaveri: 8MB of L2(4 * 2)
Orochi: 8MB of L2 + 8MB of L3 = 16 MB of total cache(2 * 4 + 2 * 4)
Viperfish: 20 MB of L2(4 * 5)

So, far there is two modes based on the slides for Kaveri and Viperfish:
High Performance mode
Low Performance mode

HPM -> Full 20/8 MB of L2 @ 32 - 40 cycles of latency
LPM -> Half 10/4 MB of L2 @ 16 - 20 cycles of latency

This way it can appease both server workloads and mainstream workloads. The cache is completely unified as well so all cores can access the L2.

This is not on topic. PLease feel free to start a new thread for your discussion outside of the FX-8350. Thanks.

 

[seronx]

This is not on topic. PLease feel free to start a new thread for your discussion outside of the FX-8350. Thanks.

FX-8350 does a good job it continues the trend.

 

[Melvis]

Fixed the link. I would not consider side/upgrading right now. My 2500k is plenty right now. I just am... I don't know... Just curious as to what amd has compared to what I got. I do the same comparisons to higher up intel cpus to.


Thats because this base architecture was primarily designed for servers right? The way you put it, it sounds like an easy fix.

Ta, yea i wouldn't either if i was in your shoes. I just did a number count using guru3d's review with the 2600K vs the FX-8350 (minus the gaming results) and it was a tie, so in my books that's a win, or should i say what Bulldozer should of been in the first place.

 

[Lionheart]

Ta, yea i wouldn't either if i was in your shoes. I just did a number count using guru3d's review with the 2600K vs the FX-8350 (minus the gaming results) and it was a tie, so in my books that's a win, or should i say what Bulldozer should of been in the first place.

Aaaww hells yeah bro looks like we both know what our new platform is going to be

Btw I was hoping you would be on steam so I could spam you with AMD news

 

[[H]@RD5TUFF]

I wish I could see some comparisons to the 2500k being they are the exact same price.

2500K is a dead ship and won't be sold for much longer so that's not a very good comparison you should compare with the 3570K.

 

[HumanSmoke]

The load power consumption figures seem all over the place.

PC Perspective: 15.3% worse than FX-8150
eTeknix: 5.9% worse than FX-8150
X-bit : 3.9% worse than FX-8150
TechSpot: 0.8% worse than FX-8150
Tech Report and Extreme Tech: 0% difference
Hardware Heaven: 0.4% better than FX-8150
Hexus: 1.6% better than FX-8150
Anandtech: 2% better than FX-8150
Hardware Canucks: 3.2% better than FX-8150
Hot Hardware: 4.7% better than FX-8150
[H]OCP: 4.8% better than FX-8150
Tom's Hardware: 5.8% better than FX-8150
Legit Reviews: 7.4% better than FX-8150
TechPowerUp: 20.9% better than FX-8150

The median/mean seem ballpark near enough the same for both SKU's, but the range is pretty damn loose.

 

[drdeathx]

In the summary it says 3770K costs almost twice as much. 8350 is pitted against the 3570K FYI not the 3770K.

 

[drdeathx]

Ta, yea i wouldn't either if i was in your shoes. I just did a number count using guru3d's review with the 2600K vs the FX-8350 (minus the gaming results) and it was a tie, so in my books that's a win, or should i say what Bulldozer should of been in the first place.

2600K is old news and going buh bye.

 

[cadaveca]

In the summary it says 3770K costs almost twice as much. 8350 is pitted against the 3570K FYI not the 3770K.

Explained above. But...meh.

The load power consumption figures seem all over the place.

PC Perspective: 15.3% worse than FX-8150
eTeknix: 5.9% worse than FX-8150
X-bit : 3.9% worse than FX-8150
TechSpot: 0.8% worse than FX-8150
Tech Report and Extreme Tech: 0% difference
Hardware Heaven: 0.4% better than FX-8150
Hexus: 1.6% better than FX-8150
Anandtech: 2% better than FX-8150
Hardware Canucks: 3.2% better than FX-8150
Hot Hardware: 4.7% better than FX-8150
[H]OCP: 4.8% better than FX-8150
Tom's Hardware: 5.8% better than FX-8150
Legit Reviews: 7.4% better than FX-8150
TechPowerUp: 20.9% better than FX-8150

The median/mean seem ballpark near enough the same for both SKU's, but the range is pretty damn loose.

are the scores for all sites relatively the same?

 

[cdawall]

The load power consumption figures seem all over the place.

PC Perspective: 15.3% worse than FX-8150
eTeknix: 5.9% worse than FX-8150
X-bit : 3.9% worse than FX-8150
TechSpot: 0.8% worse than FX-8150
Tech Report and Extreme Tech: 0% difference
Hardware Heaven: 0.4% better than FX-8150
Hexus: 1.6% better than FX-8150
Anandtech: 2% better than FX-8150
Hardware Canucks: 3.2% better than FX-8150
Hot Hardware: 4.7% better than FX-8150
[H]OCP: 4.8% better than FX-8150
Tom's Hardware: 5.8% better than FX-8150
Legit Reviews: 7.4% better than FX-8150
TechPowerUp: 20.9% better than FX-8150

The median/mean seem ballpark near enough the same for both SKU's, but the range is pretty damn loose.

AMD Radeon HD 5870 1GB

2 x 4GB GSkill DDR-2 1866 memory @ 9.10.9.28 latencies

1 TB WD Caviar Black SATA-6G drive

Lite-On BDR Drive

Corsair TX750W Power Supply

Windows 7 Ultimate 64 Bit Edition

Asus Crosshair V Formula
AMD FX-8350
Corsair Vengeance 1866MHz 16GB
Corsair H80
Corsair HX1050
Kingston V+100 128GB SSD
Lian Li T60

AMD FX-8350 (Vishera, 8 cores, 4.0-4.2 GHz, 4 x 2 MB L2, 8 MB L3);
ASUS Crosshair V Formula (Socket AM3+, AMD 990FX + SB950);
Memory: 2 x 4 GB, DDR3-1866 SDRAM, 9-11-9-27 (Kingston KHX1866C9D3K2/8GX).
Graphics card: NVIDIA GeForce GTX 680 (2 GB/256-bit GDDR5, 1006/6008 MHz).
Disk drive: Intel SSD 520 240 GB (SSDSC2CW240A3K5).
Power supply unit: Corsair AX1200i (80 Plus Platinum, 1200 W).
Operating system: Microsoft Windows 7 SP1 Ultimate x64.

- x2 4GB G.Skill DDR3 PC3-14900 (CAS 8-9-8-24)
- Asrock Fatal1ty 990FX Professional (AMD 990FX)
- OCZ ZX Series 1250w
- Crucial m4 256GB (SATA 6Gb/s)
- Gigabyte GeForce GTX 580 SOC (1536MB)

Should I continue or do you get the picture? System power consumption is done at the wall so any variation in the system can give wildly different power consumption.

 

[[H]@RD5TUFF]

The load power consumption figures seem all over the place.

PC Perspective: 15.3% worse than FX-8150
eTeknix: 5.9% worse than FX-8150
X-bit : 3.9% worse than FX-8150
TechSpot: 0.8% worse than FX-8150
Tech Report and Extreme Tech: 0% difference
Hardware Heaven: 0.4% better than FX-8150
Hexus: 1.6% better than FX-8150
Anandtech: 2% better than FX-8150
Hardware Canucks: 3.2% better than FX-8150
Hot Hardware: 4.7% better than FX-8150
[H]OCP: 4.8% better than FX-8150
Tom's Hardware: 5.8% better than FX-8150
Legit Reviews: 7.4% better than FX-8150
TechPowerUp: 20.9% better than FX-8150

The median/mean seem ballpark near enough the same for both SKU's, but the range is pretty damn loose.

I'm not sure how they managed to get the power consumption up that much, to that means their is a flaw in their capacitors making they draw too much.

 

[cadaveca]

Should I continue or do you get the picture? System power consumption is done at the wall so any variation in the system can give wildly different power consumption.

Sure, but if all you are doing is swapping the CPU...

interesing, to say teh least.. I wonder how people are doing OC-wise and such too...

THe difference is probably in how those numbers were TESTED, not the CONFIGURATIONS.

 

[Melvis]

Aaaww hells yeah bro looks like we both know what our new platform is going to be

Btw I was hoping you would be on steam so I could spam you with AMD news

ive already got most of mine here, just waiting for the dam CPU

Yea i knew you would do that if i was on steam, to busy working to be able to play games today

 

[cdawall]

Sure, but if all you are doing is swapping the CPU...

interesing, to say teh least.. I wonder how people are doing OC-wise and such too...

Even within each CPU power consumption varies wildly. Look at your own Intel chips in comparison to anyone elses. A high volt 8350 will make a low volt 8150 look way better than two low volt chips.

 

[cadaveca]

Even within each CPU power consumption varies wildly. Look at your own Intel chips in comparison to anyone elses.

Sure, and again, I seem to be doing far better, I think.



Anyway, I just got one board here tested 8350, then tested the 8150, waited for BIOS, got BIOS, retested 8150, then retested 8350.