AMD set to slash FX CPU pricing on September 1

[Sweepr]

Going to ditch my 4,5GHz FX6300 for Haswell pentium tomorrow. Thanks for the advice!

Except a Haswell Pentium (G3258 ~$70) is generally quite a bit cheaper than an ancient FX6300 (~$120), so not exactly the same price/segment. Nice try at trolling though.

 

[blake0812]

And I just bought the 6300 too!

 

[VirtualLarry]

For people out there that don't care about power consumption, this is good news, no? Cheaper 8-core chips? Too bad Intel didn't get that message.

 

[MiddleOfTheRoad]

Intel is generally a better buy for most people including the the typical computer gamer. And its not fair to say that Intel's advantage is due to "propaganda of misleading benchmarks" when Haswell's are generally faster in typical apps and games giving it better value for the same price. AMD only has an advantage in relatively niche cases.

If Intel is "generally a better buy" than why do they score so poorly on the Passmark Value Chart:
http://www.cpubenchmark.net/cpu_value_available.html

Sure, an Intel CPU is usually better than the equivalent AMD chip for gaming. But Intel definitely makes you pay a premium for that extra pop. For the "better buy," passmark weights performance related to street price -- only really the Celeron / Pentium dual cores are a compelling value from Intel on that chart (especially the G3258 -- that cpu is awesome). Notice that the FX chip do considerably better than i5/i7 for the VALUE metrics. I own both chips -- and there are NO games that I play on my i7 than I can't also play comfortably on my FX 8320. A lot of people skim over the fact that the GPU is just as (if not more) critical for gamers.

Niche cases? AMD does excellent on value/performance charts -- Only the overpriced Kaveri scores poorly in that regard. It's just the same tired "Intel Is Always Better" mantra -- but if you look at the Value charts, it's clear that it's not always true.

 

[blake0812]

For people out there that don't care about power consumption, this is good news, no? Cheaper 8-core chips? Too bad Intel didn't get that message.

yes that is good, i've been looking to get the 8320 but it was 125w and my motherboard only supported 95w.

 

[MiddleOfTheRoad]

For people out there that don't care about power consumption, this is good news, no? Cheaper 8-core chips? Too bad Intel didn't get that message.

I agree. I'd love a $150 Intel chip that does eight threads, even it is was a lower clock speed than an i7 -- Right now, AMD is the only one that sells one. I'll put that on my wish list with an unlocked i3.

Something I always wondered about the FX -- the cache is huge (and sluggish).... I wondered if it would be possible to build a reduced cache FX 8 core that drew maybe 65 watts. They did lower power variants of the APU's -- but they really never tried on the FX lineup. Maybe they were worried about overlapping the Opteron (which had 8 core 65 watt models)?

 

[BSim500]

If Intel is "generally a better buy" than why do they score so poorly on the Passmark Value Chart:
http://www.cpubenchmark.net/cpu_value_available.html

Because Passmark only shows how good CPU's are at synthetic Passmark benchmarks.

And according to that "CPU Mark per $" chart, everyone should rush out and buy a 2GHz AMD Turion Dual-Core RM-70 simply because it's "$2". Seriously...

 

[Enigmoid]

Something I always wondered about the FX -- the cache is huge (and sluggish).... I wondered if it would be possible to build a reduced cache FX 8 core that drew maybe 65 watts. They did lower power variants of the APU's -- but they really never tried on the FX lineup. Maybe they were worried about overlapping the Opteron (which had 8 core 65 watt models)?

Now AMD did state that there was little benefit in the L3 cache outside of server workloads but that may be because it is terribly slow. For the FX CPUs the L2 is not shared so L3 is a must.

It simply may be that it would be very difficult to get the bandwidth and latency from a shared L2 cache (eliminating L3 like the APUs) on a 4 module system.

Reducing L3 would certainty yield a very interesting chip. 8 MB L2 and 4 MB L3.

Reducing the amount of cache would certaintly increase yields, decrease die size and get costs under control. I can't help but think that the reason AMD is reducing prices is that sales of FX CPUs is dropping; while reducing the price will certaintly increase sales, AMD's profit margins will shrink even more. I don't expect black for their CPU division on the next quarterly.

 

[Phynaz]

More than likely, yes. Anyways, burning through 25% of an SSD's life in 1-2 months isn't my kind of fun.

At 5 - 9TB a month you aren't going to burn through 25% of it's life in 1 - 2 months.

By far the most telling takeaway thus far is the fact that all the drives have endured 600TB of writes without dying. That's an awful lot of data—well over 300GB per day for five years—and far more than typical PC users are ever likely to write to their drives. Even the most demanding power users would have a hard time pushing the endurance limits of these SSDs.

http://techreport.com/review/26058/the-ssd-endurance-experiment-data-retention-after-600tb/2

 

[MiddleOfTheRoad]

Because Passmark only shows how good CPU's are at synthetic Passmark benchmarks.

And according to that "CPU Mark per $" chart, everyone should rush out and buy a 2GHz AMD Turion Dual-Core RM-70 simply because it's "$2". Seriously...

But synthetic benchmarks are likely the most accurate method of measuring actual processing power -- since most game developers optimize/compile for specific hardware, it produces a lot of misleading benchmarks. Look at what happened with Watch Dogs on Radeons -- that game choked on Radeons when it was released. Suddenly, a new catalyst driver shows up and the cards can suddenly play the game fluidly. But if you benchmarked these cards before the new drivers -- you'd think a Radeon R9 280x video card was unusably slow hardware. Game benchmarks can really distort the performance of the actual hardware.

I will admit it is really dumb for passmark to put some laptop chips on the list.... I still don't understand why they do it.

 

[Erenhardt]

Except a Haswell Pentium (G3258 ~$70) is generally quite a bit cheaper than an ancient FX6300 (~120), so not exactly the same price/segment. Nice try at trolling though.

*Checks the OP again*

Make sure its not you, who is trolling. You should have rolled out with $132 (lunch day MSRP) so the trollometers would go pop!

 

[Sweepr]

*Checks the OP again*

Yeah, I'm sure someone making fun of cheaper Haswell Pentiums has a lot to do with AMD FX price cuts. Bononos was making a point about Intel/AMD chips @ same prices yet you compared a more expensive FX chip to an entry-level Haswell chip. You are trying too hard.

 

[Erenhardt]

Ok, I will spoon-feed you.

AMD will introduce three new FX 8-series CPUs at the same time, with two running at 95W, while the price of the popular FX-6300 is to drop from £80 to around £70.

Which drops in into pentium price range

 

[MiddleOfTheRoad]

Ok, I will spoon-feed you.

Which drops in into pentium price range

If the FX 6300 does in fact drop down to being that cheap -- I really hope it prompts Intel to at least try to close the value gap with price cuts of their own.

 

[Sweepr]

Ok, I will spoon-feed you.



Which drops in into pentium price range

Except it doesn't (Amazon.uk prices):

- Intel Pentium G3258 CPU (Unlocked): £54.95
- Intel Pentium G3220 CPU: £38.59

That's closer to £79 Core i3s than Pentiums.

 

[monkeydelmagico]

It's about time. Fx 6300 has to undercut the i3 otherwise it's a no brainer to buy intel. Same for 8350; it should be at least $20.- less than a locked i5.

 

[SlowSpyder]

It's about time. Fx 6300 has to undercut the i3 otherwise it's a no brainer to buy intel. Same for 8350; it should be at least $20.- less than a locked i5.

Right now Newegg has the FX 6300 for $108 and the FX 8320 for $130. Hard choice for me. But those aren't bad prices for very capable CPU's. $108 for a 'hex core' that can easily clock upwards of 4GHz for barely more than one Benjamin.

 

[Fjodor2001]

Some years ago, Intel also made price cuts when a new CPU generation was about to be release, or was released. Nowadays that does not happen much anymore, at least not for desktop CPUs. I guess it's because e.g. SB is almost as fast as IB, and IB almost as fast as Haswell. So if they'd sell IB at a 20-40% discount compared to the release price, most people would get that instead of Haswell.

So it's interesting that we're seeing price cuts for AMD. Why are they following a different strategy than Intel?

 

[ElFenix]

gotta love how the thread went from 'AMD slashing prices' to a pissing contest of 'SSD vs Mechanical' harddrives... WOOT!

iknorite

 

[jhu]

I agree. I'd love a $150 Intel chip that does eight threads, even it is was a lower clock speed than an i7 -- Right now, AMD is the only one that sells one. I'll put that on my wish list with an unlocked i3.

Performance is what matters, not necessarily the number of threads. For example, a stock Core i5 4570 (quad core) is faster than a stock Phenom II x6 1090T. See here. The IPC of Intel's chips is enormous. The only way for AMD to compete with regard to performance is to increase clocks and/or add more computing units.

 

[MiddleOfTheRoad]

Performance is what matters, not necessarily the number of threads. For example, a stock Core i5 4570 (quad core) is faster than a stock Phenom II x6 1090T. See here. The IPC of Intel's chips is enormous. The only way for AMD to compete with regard to performance is to increase clocks and/or add more computing units.

It depends on the task. More threads (usually) = more projects simultaneously on World Community Grid for example.... Thus more points per day. My $140 FX 8320 averages roughly around the same amount of points as my $340 i7 3770k each day (I run both under Ubuntu). Its 8 Threads Vs. 8 Threads -- both running stock clocks of 3.5 Ghz. The IPC of the physical hardware is not as far apart as Windows mistakenly leads people to believe. My FX 8320 and 3770K run neck and neck under Linux, yet that same FX gets demolished under Windows 7 by the 3770k (which is why people mistakenly assume it is weaker hardware).

I did mention this in a different thread, but take a look at the Phoronix site for linux benchmarks of an AMD FX 8350 processor -- I bet they'll surprise you. Most people are completely stunned that an FX 8350 is faster than a 3770k in 7 benchmarks. I wish Anand would test more under Linux, it would give a more accurate reflection of the full potential of the physical hardware. Game benchmarks alone can be incredibly misleading.

Even now, Windows 7 seems to struggle efficiently feeding CMT (Even after the Bulldozer patches / I haven't tested Windows 8) -- In my experience, linux is much better pushing the hardware to its full potential. So if you are planning to run Linux, these FX price drops may become extremely compelling.

 

[DrMrLordX]

As a pro AMD nerd I don't think there is that much inventory to push out the door.

Probably quite the opposite, dumping prices in order to not choke with inventory in Q4. Expect another drop in CPU revenues by that time.

Hmm. Well, whatever their reasons for the price cut, for the sake of AMD and the buying public, let us hope they have something interesting to take over the flagship spot from the aging FX CPUs.

It depends on the task. More threads (usually) = more projects simultaneously on World Community Grid for example.... Thus more points per day. My $140 FX 8320 averages roughly around the same amount of points as my $340 i7 3770k each day (I run both under Ubuntu). Its 8 Threads Vs. 8 Threads -- both running stock clocks of 3.5 Ghz. The IPC of the physical hardware is not as far apart as Windows mistakenly leads people to believe. My FX 8320 and 3770K run neck and neck under Linux, yet that same FX gets demolished under Windows 7 by the 3770k (which is why people mistakenly assume it is weaker hardware).

I did mention this in a different thread, but take a look at the Phoronix site for linux benchmarks of an AMD FX 8350 processor -- I bet they'll surprise you. Most people are completely stunned that an FX 8350 is faster than a 3770k in 7 benchmarks. I wish Anand would test more under Linux, it would give a more accurate reflection of the full potential of the physical hardware. Game benchmarks alone can be incredibly misleading.

Even now, Windows 7 seems to struggle efficiently feeding CMT (Even after the Bulldozer patches / I haven't tested Windows 8) -- In my experience, linux is much better pushing the hardware to its full potential. So if you are planning to run Linux, these FX price drops may become extremely compelling.

It has been observed that Kaveri also has problems under Windows 7 (not so much under Win8.1. Haven't heard much about it under Linux).

You do raise an interesting point; in fact, much of what I have been doing lately has been in Xubuntu. However, one of the major headaches involved with the octal-core FX chips in particular are the insane thermals/power draw, no matter what your operating system. Sure, the FX lineup will be dropping in price, but you still have to pay for:

1). A strong-enough PSU to handle the chip + GPU(s).
2). A board that can handle the chip without CPU throttling/risk of VRM burnout. Especially if you're going for an overclock to 9590 levels of performance.
3). Proper cooling for the chip.

Even with the price drop, are you saving any money versus going Intel by getting an 8350 and overclocking it to, say, 4.5 ghz (let's be conservative)?

Also, how much custom recompiling from source have you had to do to fully reap the rewards of using an FX chip under Linux? Obviously you aren't running Gentoo, so you didn't have to go that far.

 

[jhu]

It depends on the task. More threads (usually) = more projects simultaneously on World Community Grid for example.... Thus more points per day. My $140 FX 8320 averages roughly around the same amount of points as my $340 i7 3770k each day (I run both under Ubuntu). Its 8 Threads Vs. 8 Threads -- both running stock clocks of 3.5 Ghz. The IPC of the physical hardware is not as far apart as Windows mistakenly leads people to believe. My FX 8320 and 3770K run neck and neck under Linux, yet that same FX gets demolished under Windows 7 by the 3770k (which is why people mistakenly assume it is weaker hardware).

If you had looked at my link, you'll see that my results concur with Sandy/Ivy Bridge = FX Module at full threads. But more threads only means more performance for the same architecture and speed. I gave the example of quad-core Haswell (non-HT) being faster than hexcore K10h at the same clockspeed. But that's not true for Sandy/Ivy Bridge where the hexcore K10h just barely edges them out. That's not something that's obvious based on the core count and clock speed.

For low-threaded programs, FX is inferior hardware to Sandy/Ivy Bridge. And now with highly threaded programs, FX is inferior to Haswell.

 

[spaceduty]

Saw this on xbitlabs.

Link its rather strange they would offer the same chip in a lower wattage for the same price. Time will tell.

 

[SlowSpyder]

My FX 9370 under-volts really well, but I'm also using water cooling. The FX 8370E is an interesting part. Just messing around with my FX 9370 in Cinevench R15 I was able to run 4.4GHz / 4.7GHz turbo at 1.225v (FX 4xxx, 6xxx, 8xxx are 1.35v from the factory, I think). My CPU was pulling about ~120 watts. I haven't tried further under-volting yet, but maybe it'll go even further.

*edit - If the FX 8370E's are well binned parts, they could be great overclockers. Just speculating.