KitGuru tested the FX9590... it's pretty bad.

[guskline]

What kind of voltage you need for that 5GHz?

I used 1.515V in an Asus Sabertooth FX 990 Rev 1. mb (rig 2 below) I could boot in at 1.48 but upped it for stability. Gets hot quick.

 

[SlowSpyder]

I used 1.515V in an Asus Sabertooth FX 990 Rev 1. mb (rig 2 below) I could bott in at 1.48 but upped it for stability. Gets hot quick.

At the 1.515V / 5GHz setting, would you say that the CPU would overwhelm you're Corsair H100 at or near 100% load?

 

[BallaTheFeared]

The maximum overclocks argument is a bit lame though,especially when overclocking is a lottery. Factory pre-overclocks are also guaranteed.

a) You don't really need an OC for the i7, it's already faster in an overwhelming majority of apps.

b) They're guaranteed alright, for 30 days. 🙄

c) $900

 

[sushiwarrior]

I wouldn't count on it. Looking at the history of hyperthreading, it's been around for ages, but it's never really made a world of difference. I sort of see hyperthreading like Nvidia's PhysX; it has certain times where it makes a huge difference, but more often than not, it makes little to no difference (due to not being supported) as of now.

Games will definitely utilize more cores in the future, but historically, by that point, the other features of the CPU would be holding it back already. I wouldn't expect a 6 core X6 to suddenly catch up to an i5 4670K because of using more cores in the future.

Hyperthreading is one of the best inventions for CPUs in history. It's genius. It's similar to making a processor super OoO. Instead of sitting around doing nothing after a cache miss, it can execute a second thread. It's brilliant. Bulldozer happened because AMD wants to come up with their own comparable SMT architecture, but they probably are not able to license HT (I would think?).

Only cases where HT doesn't help are A. if it isn't threaded and B. if the operation is extremely operation heavy with what, maybe heavy L1 cache usage and light L2/L3 use? A very small operation set perhaps?

 

[galego]

a) You don't really need an OC for the i7, it's already faster in an overwhelming majority of apps.

People who OC the i7 think otherwise. No?

The point is that the 9590 is a choice for those who would buy a 8350 and OC it above 4.5Ghz. Intel does not have something similar because it cares little about overclockers

 

[Enigmoid]

Hyperthreading is one of the best inventions for CPUs in history. It's genius. It's similar to making a processor super OoO. Instead of sitting around doing nothing after a cache miss, it can execute a second thread. It's brilliant. Bulldozer happened because AMD wants to come up with their own comparable SMT architecture, but they probably are not able to license HT (I would think?).

Only cases where HT doesn't help are A. if it isn't threaded and B. if the operation is extremely operation heavy with what, maybe heavy L1 cache usage and light L2/L3 use? A very small operation set perhaps?

Not quite sure about this but does HT sometimes slow down an operation at the benefit of getting more operations done? For example, each calculation may take 10% longer but 30% more calculations can be done.

 

[TerryMathews]

Not quite sure about this but does HT sometimes slow down an operation at the benefit of getting more operations done? For example, each calculation may take 10% longer but 30% more calculations can be done.

It's less true today than it was in the P4 era but theoretically yes. The scheduler will prioritize towards maximum utilization, but that may come at the expense of individual threads.

 

[sushiwarrior]

Not quite sure about this but does HT sometimes slow down an operation at the benefit of getting more operations done? For example, each calculation may take 10% longer but 30% more calculations can be done.

Absolutely depends on the usage. How HT works is the architectural state resources are doubled, while the execution units stay the same. So the processor can be ordered to execute 2 instructions at once, using one execution block. Since a significant part of the CPU's time can be taken up waiting for values to be returned from the cache or something similar to that, instead the processor can now work on the second thread while before it would have been waiting for the value to be sent from the cache (although, to an extent, out of order execution already relieves this issue). Hyperthreading isn't about slowing anything down, it's about assuring that the processor's execution units are always working as hard as they could be, instead of sitting idle during certain periods.

 

[BallaTheFeared]

People who OC the i7 think otherwise. No?

The point is that the 9590 is a choice for those who would buy a 8350 and OC it above 4.5Ghz. Intel does not have something similar because it cares little about overclockers

Let me clarify, you don't need to OC the i7 to be faster than the 9590 in the majority of applications - it already is.

Yeah I forgot AMD doesn't offer warranty for OC, either on the 8350 or their 9590 (You can't call the typical newegg 30 day grace period a AMD warranty).

Seems risky to OC AMD at all, let alone buy a $900 processor with no warranty.

 

[sushiwarrior]

Let me clarify, you don't need to OC the i7 to be faster than the 9590 in the majority of applications - it already is.

Yeah I forgot AMD doesn't offer warranty for OC, either on the 8350 or their 9590 (You can't call the typical newegg 30 day grace period a AMD warranty).

Seems risky to OC AMD at all, let alone buy a $900 processor with no warranty.

Please, stop being a giant hypocrite. You make yourself look bad.

Intel has not tested, and does not warranty, the operation of the processor beyond its specifications. Intel has not tested, and does not warranty, the operation of other system components beyond their industry standard specifications. Intel assumes no responsibility that the processor and other system components, including if used with altered clock frequencies and/or voltages, will be fit for any particular purpose.

Seriously, you've got to be kidding me if that's the kind of hypocritical bull you want to start pulling.

 

[BallaTheFeared]

http://click.intel.com/tuningplan/

 

[aigomorla]

i learned the fastest way to get people to not listen to you, is to have the word GURU in your ID.
How many of u guys would of trusted me with watercooling, if i my name was WatercoolingGuru.

Like others have said... im finding some faults in this test..

Seriously, you've got to be kidding me if that's the kind of hypocritical bull you want to start pulling.

u owe balla an appology.

Intel does have a tuning plan warrenty which covers u from overclocking.

In other words, if it fails under normal usage, we will replace it under the standard warranty; if it fails while running outside of intel's specifications, we will replace it under the Performance Tuning Protection Plan.

 

[Chiropteran]

Hyperthreading is one of the best inventions for CPUs in history. It's genius. It's similar to making a processor super OoO. Instead of sitting around doing nothing after a cache miss, it can execute a second thread. It's brilliant. Bulldozer happened because AMD wants to come up with their own comparable SMT architecture, but they probably are not able to license HT (I would think?).

Not at all. It's been shown that, for example, an FX-8350 scales better going from 4 to 8 threads than an i7-3770. Now, the i7 has better single threaded performance to start, so even with the inferior scaling it can still be competitive in high-thread situations.

AMD's module approach is proven superior over hyper threading, it's just a matter of getting a single module core's performance up to par.

 

[TuxDave]

Not at all. It's been shown that, for example, an FX-8350 scales better going from 4 to 8 threads than an i7-3770. Now, the i7 has better single threaded performance to start, so even with the inferior scaling it can still be competitive in high-thread situations.

AMD's module approach is proven superior over hyper threading, it's just a matter of getting a single module core's performance up to par.

I don't see it as a mutually exclusive idea. You could put in more modules or have each module have hyperthreading or both! With the module/core only approach, if one of the module gets hit with a cache miss, it's going to be sitting there doing nothing.

 

[galego]

Let me clarify, you don't need to OC the i7 to be faster than the 9590 in the majority of applications - it already is.

Yeah I forgot AMD doesn't offer warranty for OC, either on the 8350 or their 9590 (You can't call the typical newegg 30 day grace period a AMD warranty).

Seems risky to OC AMD at all, let alone buy a $900 processor with no warranty.

Far from clarifying this borrows all, whereas ignoring my point.

Your opinion about OC the i7 is just an opinion. Others disagree, for example the author of the KitGuru review who OC the i7 a 33% before comparing it to the 9590 :whiste:

There is nothing risky (except with excessively high temperatures of Haswell chips). The build quality of an AMD chip is very superior. The worldwide records of overclocking go to AMD chips never to Intel.

But the point was not that, but the guaranty of obtaining a concrete clock. The 9590 is a choice for those who would buy a 8350 and overclock it beyond 4.5GHz, but cannot play to silicon lottery.

It is poor for Intel, which cannot guarantee you a given overclock capability. You could buy a Haswell k series and find that you only can OC a few percents. :whiste:

I note that you are locked in the $900 figure when it can be purchased by less or even by much less. Ask to one poster here how many he spent in the new chip.

 

[Accord99]

AMD's module approach is proven superior over hyper threading, it's just a matter of getting a single module core's performance up to par.

Or you can look at it from reverse, both approaches end up with roughly maximum possible throughput with 8 "weak cores". Intel's approach is superior since the resources of each cores can be combined into one powerful core/module to give 60-70% better performance with a single thread. AMD's approach only gives a small boost.

Hyperthreading, therefore is also the elusive "Reverse Hyperthreading".

 

[galego]

http://click.intel.com/tuningplan/

i learned the fastest way to get people to not listen to you, is to have the word GURU in your ID.
How many of u guys would of trusted me with watercooling, if i my name was WatercoolingGuru.

Like others have said... im finding some faults in this test..

u owe balla an appology.

Intel does have a tuning plan warrenty which covers u from overclocking.

LOL he does not own an apology to balla, first because we are talking about normal purchases. Not about a special plan that you have to pay in addition to the chip purchase.

Second, that plan is not very different from buying an chip to a OC reseller who already gives you extra OC warranty.

Third, and more important, because this is omits the point. The tuning plan does not guarantize you a OC speed unlike AMD with the 9 series.

Fourth, Intel tunig plan is a joke. Imagine you buy an Intel failwell and OC, the chip fails. You receive a replacement chip, but Intel does not guarantee that this has the same thermal/electric tolerances that the one replaced. Then you try again, now with lower OC setting... and it fails again because was poor silicon. You will not receive a third one...

 

[BallaTheFeared]

Far from clarifying this borrows all, whereas ignoring my point.

Wut? You mean like ignoring how the i5 is faster in h264?

I'm not sure what can be more clear than the stock i7 is faster in more things than it's slower against the stock 9590.

Your opinion about OC the i7 is just an opinion. Others disagree, for example the author of the KitGuru review who OC the i7 a 33% before comparing it to the 9590 :whiste:

You mean in the review that showed the stock i7 winning more than losing, and the overclocked i7 walking away?

Is this the same review that was showing the 5GHz 9590 OC losing to the 4.5GHz i7 while drawing twice as much power?

It's like you're arguing to buy a video card that costs three times as much, and once overclocked isn't even as fast as the cheaper GPU, while it draws twice as much power.

I wonder how a product like that would float in a market that actually knows what competition is...

There is nothing risky (except with excessively high temperatures of Haswell chips). The build quality of an AMD chip is very superior. The worldwide records of overclocking go to AMD chips never to Intel.

Ohh, the no proof statement, right. Also my chip is on a $50 air cooler and was never delidded, no temp issues for me.

Yes yes, AMD makes quality tank like products. So robust and durable, they never ever fail, steady, slow, cumbersome, just like a tank, .01 miles to the gallon. Just like a tank.

Ironic statement since even the WR LN2 runs wouldn't beat a good clocking on air Haswell :|

Still none of that addresses the issue of no warranty for overclocking, nor does it address the 30 day warranty a $900 product comes with. Which makes you wonder, if AMD thought their uarch was so robust with excessive voltage and temps, why a 30 day warranty? Why not a year, why not a typical three year warranty?

Clearly not even AMD thinks these will last.

But the point was not that, but the guaranty of obtaining a concrete clock. The 9590 is a choice for those who would buy a 8350 and overclock it beyond 4.5GHz, but cannot play to silicon lottery.

Why would anyone buy a 8350 if they needed the performance it provides at 5GHz? Why wouldn't they buy a SBe, IBe, or Haswell? All of which offer better performance/$ and all of which even with the worst clockers are faster.

It is poor for Intel, which cannot guarantee you a given overclock capability. You could buy a Haswell k series and find that you only can OC a few percents. :whiste:

I see the the other way. It's poor of AMD, having to run their uarch to the brink with pretty much zero OC headroom, huge power budget, no warranty, and even then it fails. It seems even if AMD succeeded in hitting their clock goals they had for this terrible uarch for mainstream products it still would have failed just as miserably.

I note that you are locked in the $900 figure when it can be purchased by less or even by much less. Ask to one poster here how many he spent in the new chip.

I only purchase products from approved retailers, unless they're used. Though it's not like you have to worry about warranty, since AMD provides none anyways.

 

[galego]

Not at all. It's been shown that, for example, an FX-8350 scales better going from 4 to 8 threads than an i7-3770. Now, the i7 has better single threaded performance to start, so even with the inferior scaling it can still be competitive in high-thread situations.

AMD's module approach is proven superior over hyper threading, it's just a matter of getting a single module core's performance up to par.

The reason why the Intel 3770 scales so bad above 4 threads is because the 8 threads are associated to a mixture of real cores plus virtual cores. Unlike in the AMD module design where each one of the 8 threads is associated to a real core.

It depends what do you mean by "be competitive in high-thread situations". An 8350 can be a 20-40% faster than an i7-3770k in those situations. A 9590 a 40-60% faster.

 

[ShintaiDK]

The reason why the Intel 3770 scales so bad above 4 threads is because the 8 threads are associated to a mixture of real cores plus virtual cores. Unlike in the AMD module design where each one of the 8 threads is associated to a real core.

It depends what do you mean by "be competitive in high-thread situations". An 8350 can be a 20-40% faster than an i7-3770k in those situations. A 9590 a 40-60% faster.

8 FP threads would not get a "real core". And you make up more random nonsense numbers.

 

[guskline]

At the 1.515V / 5GHz setting, would you say that the CPU would overwhelm you're Corsair H100 at or near 100% load?

The H100 can handle it but the temps are @65C with both fans running full out. I think the FX 9590 might be a bit better. Remember though I was running ALL 8 cores at 5 Ghz. The 9590 runs all of them at 4.7 Ghz and some of them at 5 Ghz.

SlowSpyder I run my 8350 24/7 at 4.6 Ghz with Multiplier at 21 and FSB at 219. I use 1866 memory but set it at 1600 and allow the fsb increase to take it to @1750. My vcore is set to 1.462. I run much cooler and am plenty FAST for what I need.

Here are some benchmarks at the various speeds.

 

[guskline]

You can see in the 4.6Ghz column the 8350 gets great benchmarks AND runs cool.

 

[aigomorla]

LOL he does not own an apology to balla, first because we are talking about normal purchases. Not about a special plan that you have to pay in addition to the chip purchase.

Second, that plan is not very different from buying an chip to a OC reseller who already gives you extra OC warranty.

Third, and more important, because this is omits the point. The tuning plan does not guarantize you a OC speed unlike AMD with the 9 series.

Fourth, Intel tunig plan is a joke. Imagine you buy an Intel failwell and OC, the chip fails. You receive a replacement chip, but Intel does not guarantee that this has the same thermal/electric tolerances that the one replaced. Then you try again, now with lower OC setting... and it fails again because was poor silicon. You will not receive a third one...

no because you cant get anything close to that from the AMD side.

You pay the extra option, yet its an option.

Its also available only on the K line... but as i said if ur overclocking, your getting a K line anyhow.

and your last statement just shows how rooted u are with AMD.
The thing is , yeah u wont get the same overclocking chip, however u get a NEW luck at a new chip.
No questions asked minus physical damage on how the chip died.
Electron mitigation damage from long term overvolting... no problem.... Intel will replace it without batting an eye... or give u a cpu greater equal in value to it.
(and they are usually great in options of EQUAL OR GREATER)

With an AMD cpu... it dies.. its DEAD Jim, do not pass go.. do not collect 200 dollars... end of options for you period.



I still think that guy owes Balla an appology.
He said INTEL doesnt warrenty overclocking.
Balla showed it did.

 

[USER8000]

a) You don't really need an OC for the i7, it's already faster in an overwhelming majority of apps.

b) They're guaranteed alright, for 30 days. 🙄

c) $900

The cherry picked Core i5 CPUs which have been hand tested for good clocks(especially under exotic cooling) go for silly money. You forget the Intel P4 EE which were silly money too but never were any real value for money just like the FX9590. Moreover,at least looking in Europe retailers have 3 year warranties on the same CPUs,so I suspect that is a typo on Newegg just like with the clockspeeds you said were a typo.

 

[frozentundra123456]

The cherry picked Core i5 CPUs which have been hand tested for good clocks(especially under exotic cooling) go for silly money. You forget the Intel P4 EE which were silly money too but never were any real value for money just like the FX9590. Moreover,at least looking in Europe retailers have 3 year warranties on the same CPUs,so I suspect that is a typo on Newegg just like with the clockspeeds you said were a typo.

Well, at least you admit the 9590 is not a good value for the money. At least one part of your post is accurate and relevant.